The Computer Incident Response Center Luxembourg (CIRCL) is a government-driven initiative designed to provide a systematic response facility to computer security threats and incidents.
- Website
- https://www.circl.lu
- info@circl.lu
- GNA identifier
- GNA-1 GCVE registry Recent publications
Recent vulnerabilities
355 GCVE records assigned by this organization as GNA-1GCVE-1-2026-20294 (CVE-2026-103321)
Vulnerability from gna-1 – Published: 2026-09-30 12:18 – Updated: 2026-09-30 12:18| URL | Tags |
|---|---|
| https://github.com/MISP/MISP/commit/92c7ccc43 | patch |
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-30 11:21 - Model
qwen3.8:27b- Input
-
https://github.com/MISP/MISP/commit/92c7ccc43.patch
0ecb893de300… - Confidence
- high
| Commit | Subject | Patch SHA-256 |
|---|---|---|
92c7ccc4398a
|
fix: [security] Validate the event graph preview and stop | 0ecb893de300… |
Fix summary
The vulnerability is remediated by enforcing strict server-side validation of the preview image field, restricting it to a well-formed base64-encoded PNG data URL, and by replacing the client-side string-concatenation rendering with DOM-based attribute assignment that does not interpret the value as HTML.
Patch summary
In app/Model/EventGraph.php, a new validation rule is added for the preview_img field requiring it to match the regex /^data:image\/png;base64,[A-Za-z0-9+\/]*={0,2}$/ (allowing empty). In app/webroot/js/event-graph.js, two occurrences of string-concatenated img tag construction (return '<img ... src="' + value + '" />') are replaced with jQuery DOM construction using $('<img ...>').prop('src', value), which sets the attribute safely without HTML parsing.
CVSS rationale
AV:N: exploited over the network via the MISP web interface. AC:L: no race conditions or special conditions required; storing a crafted value and viewing the graph is straightforward. AT:N: no manipulation of the attack target needed. PR:L: attacker needs a low-privilege authenticated MISP account to create/modify an event graph. UI:P: the victim passively triggers the XSS by viewing the event graph preview (hovering a button), a normal workflow action. VC/VI/VA:N: the MISP server itself is not compromised; the impact is in the victim's browser. SC:H: the attacker can read cookies, tokens, and data in the victim's session. SI:H: the attacker can perform authenticated actions as the victim. SA:N: no availability impact on the victim's system.
Weakness rationale
- CWE-79 The stored preview_img value was rendered into an HTML attribute via string concatenation without sufficient neutralization, enabling script injection in the victim's browser. This is a textbook stored XSS.
- CWE-20 The server accepted and persisted the preview_img field without any format validation, allowing arbitrary content to be stored and later rendered. The fix adds a strict regex validation rule.
Attack pattern rationale
- CAPEC-1 The patch directly addresses a stored XSS where attacker-controlled data is rendered into a web page without proper encoding or validation. CAPEC-1 is the canonical attack pattern for XSS and is the closest match. No uncertainty in this mapping; the commit message explicitly identifies the issue as stored XSS.
Assumptions to verify
- The affected version range is inferred from the tag_version_boundary (v2.5.48 with 43 commits after the fix); the exact fixed release version is not stated in the patch metadata.
- PR:L assumes the attacker needs at least a basic authenticated MISP account to create or modify an event graph entry; the patch does not specify the exact permission level required.
- UI:P assumes the victim triggers the XSS by viewing the event graph preview as part of normal workflow (hovering the plot button), which is a passive interaction rather than an active click on a crafted link.
- The CAPEC-1 mapping is direct and unambiguous given the explicit stored XSS identification in the commit message.
- The Co-Authored-By line referencing Claude Opus 4.8 is treated as a tool credit per the commit metadata; it is not a human contributor.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
6 | 9 | high | 5 |
{
"containers": {
"cna": {
"affected": [
{
"cpes": [
"cpe:2.3:a:misp:misp:*:*:*:*:*:*:*:*"
],
"modules": [
"EventGraph model",
"event-graph.js client-side rendering"
],
"product": "MISP",
"programFiles": [
"app/Model/EventGraph.php",
"app/webroot/js/event-graph.js"
],
"repo": "https://github.com/MISP/MISP",
"vendor": "MISP",
"versions": [
{
"lessThan": "2.5.48",
"status": "affected",
"version": "0",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "reporter",
"value": "Bastien Bossiroy of NCIA"
},
{
"lang": "en",
"type": "remediation developer",
"value": "iglocska"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Opus 4.8"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eMISP contains a stored cross-site script (XSS) vulnerability in the event graph preview feature.\u003c/p\u003e\u003cp\u003eThe event graph preview image field was accepted and stored without server-side validation. On the client side, the stored value was rendered into an HTML img element\u0027s src attribute via string concatenation, allowing a crafted value to break out of the attribute context and inject arbitrary script.\u003c/p\u003e\u003cp\u003ePreconditions:\u003c/p\u003e\u003cp\u003e- An authenticated MISP user with the ability to create or modify an event graph entry.\u003c/p\u003e\u003cp\u003e- A second user (the victim) who views the event graph and triggers the preview popover.\u003c/p\u003e\u003cp\u003eImpact:\u003c/p\u003e\u003cp\u003e- Execution of arbitrary JavaScript in the victim\u0027s browser within the MISP application context.\u003c/p\u003e\u003cp\u003e- Potential theft of session tokens, cookies, or sensitive data accessible to the victim\u0027s browser.\u003c/p\u003e\u003cp\u003e- Potential for performing actions on behalf of the victim within the MISP application.\u003c/p\u003e\u003cp\u003eAffected: MISP versions prior to the fix (commit applied after v2.5.48).\u003c/p\u003e"
}
],
"value": "MISP contains a stored cross-site script (XSS) vulnerability in the event graph preview feature.\n\nThe event graph preview image field was accepted and stored without server-side validation. On the client side, the stored value was rendered into an HTML img element\u0027s src attribute via string concatenation, allowing a crafted value to break out of the attribute context and inject arbitrary script.\n\nPreconditions:\n\n- An authenticated MISP user with the ability to create or modify an event graph entry.\n\n- A second user (the victim) who views the event graph and triggers the preview popover.\n\nImpact:\n\n- Execution of arbitrary JavaScript in the victim\u0027s browser within the MISP application context.\n\n- Potential theft of session tokens, cookies, or sensitive data accessible to the victim\u0027s browser.\n\n- Potential for performing actions on behalf of the victim within the MISP application.\n\nAffected: MISP versions prior to the fix (commit applied after v2.5.48)."
}
],
"impacts": [
{
"capecId": "CAPEC-1",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-1 Cross Site Scripting (XSS)"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 8.3,
"baseSeverity": "HIGH",
"privilegesRequired": "LOW",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "HIGH",
"subIntegrityImpact": "HIGH",
"userInteraction": "PASSIVE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:N/VI:H/VA:N/SC:H/SI:H/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "NONE",
"vulnIntegrityImpact": "HIGH",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-79",
"description": "CWE-79 Improper Neutralization of Input During Web Page Generation (\u0027Cross-site Scripting\u0027)",
"lang": "en",
"type": "CWE"
}
]
},
{
"descriptions": [
{
"cweId": "CWE-20",
"description": "CWE-20 Improper Input Validation",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/MISP/MISP/commit/92c7ccc43"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe vulnerability is remediated by enforcing strict server-side validation of the preview image field, restricting it to a well-formed base64-encoded PNG data URL, and by replacing the client-side string-concatenation rendering with DOM-based attribute assignment that does not interpret the value as HTML.\u003c/p\u003e"
}
],
"value": "The vulnerability is remediated by enforcing strict server-side validation of the preview image field, restricting it to a well-formed base64-encoded PNG data URL, and by replacing the client-side string-concatenation rendering with DOM-based attribute assignment that does not interpret the value as HTML."
}
],
"title": "MISP Stored Cross-Site Scripting (XSS) via Unvalidated Event Graph Preview Image",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The affected version range is inferred from the tag_version_boundary (v2.5.48 with 43 commits after the fix); the exact fixed release version is not stated in the patch metadata.",
"PR:L assumes the attacker needs at least a basic authenticated MISP account to create or modify an event graph entry; the patch does not specify the exact permission level required.",
"UI:P assumes the victim triggers the XSS by viewing the event graph preview as part of normal workflow (hovering the plot button), which is a passive interaction rather than an active click on a crafted link.",
"The CAPEC-1 mapping is direct and unambiguous given the explicit stored XSS identification in the commit message.",
"The Co-Authored-By line referencing Claude Opus 4.8 is treated as a tool credit per the commit metadata; it is not a human contributor."
],
"capecRationale": [
{
"capecId": "CAPEC-1",
"rationale": "The patch directly addresses a stored XSS where attacker-controlled data is rendered into a web page without proper encoding or validation. CAPEC-1 is the canonical attack pattern for XSS and is the closest match. No uncertainty in this mapping; the commit message explicitly identifies the issue as stored XSS."
}
],
"commit": "92c7ccc4398a64e61699bd79fb4010703616fc59",
"confidence": "high",
"credits": [
{
"lang": "en",
"type": "reporter",
"value": "Bastien Bossiroy of NCIA"
},
{
"lang": "en",
"type": "remediation developer",
"value": "iglocska"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Opus 4.8"
}
],
"cvssRationale": "AV:N: exploited over the network via the MISP web interface. AC:L: no race conditions or special conditions required; storing a crafted value and viewing the graph is straightforward. AT:N: no manipulation of the attack target needed. PR:L: attacker needs a low-privilege authenticated MISP account to create/modify an event graph. UI:P: the victim passively triggers the XSS by viewing the event graph preview (hovering a button), a normal workflow action. VC/VI/VA:N: the MISP server itself is not compromised; the impact is in the victim\u0027s browser. SC:H: the attacker can read cookies, tokens, and data in the victim\u0027s session. SI:H: the attacker can perform authenticated actions as the victim. SA:N: no availability impact on the victim\u0027s system.",
"fixSummary": "The vulnerability is remediated by enforcing strict server-side validation of the preview image field, restricting it to a well-formed base64-encoded PNG data URL, and by replacing the client-side string-concatenation rendering with DOM-based attribute assignment that does not interpret the value as HTML.",
"generatedAt": "2026-09-30T11:21:27.655462Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 5,
"confidence": "high",
"model": "qwen3.8:27b",
"score": 6
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "0ecb893de30056bc3f9609fcb8de43d60b1cedd6a1eea7d6aa53ef6d351d642a",
"patchSummary": "In app/Model/EventGraph.php, a new validation rule is added for the preview_img field requiring it to match the regex /^data:image\\/png;base64,[A-Za-z0-9+\\/]*={0,2}$/ (allowing empty). In app/webroot/js/event-graph.js, two occurrences of string-concatenated img tag construction (return \u0027\u003cimg ... src=\"\u0027 + value + \u0027\" /\u003e\u0027) are replaced with jQuery DOM construction using $(\u0027\u003cimg ...\u003e\u0027).prop(\u0027src\u0027, value), which sets the attribute safely without HTML parsing.",
"patchTruncated": false,
"patches": [
{
"commit": "92c7ccc4398a64e61699bd79fb4010703616fc59",
"patchSha256": "0ecb893de30056bc3f9609fcb8de43d60b1cedd6a1eea7d6aa53ef6d351d642a",
"source": "https://github.com/MISP/MISP/commit/92c7ccc43.patch",
"sourceUrl": "https://github.com/MISP/MISP/commit/92c7ccc43.patch",
"subject": "fix: [security] Validate the event graph preview and stop"
}
],
"source": "https://github.com/MISP/MISP/commit/92c7ccc43.patch",
"subject": "fix: [security] Validate the event graph preview and stop",
"tagVersionBoundary": {
"commits_after_fix": 43,
"repository": "https://github.com/MISP/MISP",
"tag": "v2.5.48",
"version": "2.5.48",
"version_type": "semver"
},
"weaknessRationale": [
{
"cweId": "CWE-79",
"rationale": "The stored preview_img value was rendered into an HTML attribute via string concatenation without sufficient neutralization, enabling script injection in the victim\u0027s browser. This is a textbook stored XSS."
},
{
"cweId": "CWE-20",
"rationale": "The server accepted and persisted the preview_img field without any format validation, allowing arbitrary content to be stored and later rendered. The fix adds a strict regex validation rule."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20294"
}
]
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-103321",
"datePublished": "2026-09-30T12:18:47.503336Z",
"dateReserved": "2026-09-30T12:18:54.232Z",
"dateUpdated": "2026-09-30T12:18:55.326462Z",
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20294"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20019 (CVE-2026-103239)
Vulnerability from gna-1 – Published: 2026-09-30 10:16 – Updated: 2026-09-30 10:16| URL | Tags |
|---|---|
| https://github.com/MISP/MISP/commit/96f735e7b | patch |
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-30 09:58 - Model
qwen3.8:27b- Input
-
https://github.com/MISP/MISP/commit/96f735e7b.patch
67f9b5741b88… - Confidence
- high
| Commit | Subject | Patch SHA-256 |
|---|---|---|
96f735e7b5d9
|
fix: [security] Tag collection saves no longer write sibling | 67f9b5741b88… |
Fix summary
The fix replaces the bulk-association save call with an explicit two-step process: first, only the TagCollection data is extracted from the request and saved via a plain save() operation that does not write belongsTo siblings; second, tag association rows are persisted individually in a controlled loop. This ensures that any User, Organisation, or other sibling model data present in the request payload is silently discarded and never reaches the database, eliminating the privilege escalation path.
Patch summary
In TagCollectionsController.php, both addWithTags() and editWithTags() were modified. The full $this->request->data is no longer passed to saveAssociated(). Instead, only the TagCollection key is extracted from the request. The saveAssociated() call is replaced with a plain save() for the collection, followed by an explicit loop that creates and saves each TagCollectionTag row individually. In editWithTags(), the tag rewrite logic is restructured to delete existing tag associations and re-insert them after the collection save. REST success/failure response handling is added to both methods.
CVSS rationale
The vulnerability is exploitable over the network (AV:N) with low complexity (AC:L) by simply adding extra fields to a legitimate tag collection request. No special attack conditions are required (AT:N). The attacker needs an authenticated account with the tag editor permission (PR:L). No user interaction is needed (UI:N). The primary impact is on integrity (VI:H) because the attacker can create or modify User and Organisation records to escalate to site admin. Modifications can be made so data can be see (VC:H) but no availability impact (VA:N). No subsequent component is affected (SC:N, SI:N, SA:N).
Weakness rationale
- CWE-284 A user with only tag editor permissions was able to write to User and Organisation records through the bulk-association save, bypassing the intended access control boundaries. The authorization model did not restrict which associated models could be persisted.
- CWE-862 The saveAssociated() call did not enforce per-model authorization checks, allowing any associated model data in the payload to be written regardless of the caller's permissions for those models.
Attack pattern rationale
- CAPEC-126 The attacker tampers with the HTTP request by injecting additional model fields (User, Organisation) into the tag collection payload. The application's bulk-save logic processes these unexpected parameters without filtering, leading to unauthorized record creation or modification. This is the closest CAPEC to the observed attack: adding extra parameters to a legitimate request to trigger unintended side effects.
- CAPEC-1 The application accepted the entire request body without validating or restricting which model keys were present before passing it to the persistence layer. This is a secondary mapping; CAPEC-126 is more specific to the parameter-injection attack vector observed here.
Assumptions to verify
- The affected version boundary is inferred from the tag_version_boundary metadata (v2.5.48 with 24 commits after fix), suggesting the fix is included in v2.5.48. No explicit version range is stated in the patch itself.
- The privilege level required is assumed to be 'perm_tag_editor' based on the commit message; the exact permission model and whether other roles are affected is not fully verifiable from the patch alone.
- CAPEC-126 (Parameter Tampering) is selected as the closest match; the attack is more precisely a mass-assignment / sibling-model injection via an ORM bulk-save, for which no exact CAPEC exists. CAPEC-126 is the best available approximation.
- The CVSS assumes the attacker can reach the MISP web interface over the network and holds a valid session with tag editor permissions. No multi-step or race-condition requirements are evident.
- The Co-Authored-By line references an AI assistant (Claude Opus 4.8); it is credited as a tool rather than a human remediation developer.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
6 | 9 | high | 5 |
{
"containers": {
"cna": {
"affected": [
{
"cpes": [
"cpe:2.3:a:misp:misp:*:*:*:*:*:*:*:*"
],
"modules": [
"TagCollectionsController"
],
"product": "MISP",
"programFiles": [
"app/Controller/TagCollectionsController.php"
],
"repo": "https://github.com/MISP/MISP",
"vendor": "MISP",
"versions": [
{
"lessThan": "2.5.48",
"status": "affected",
"version": "0",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "reporter",
"value": "Jeroen Pinoy"
},
{
"lang": "en",
"type": "remediation developer",
"value": "iglocska"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Opus 4.8"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eMISP contains a privilege escalation vulnerability in the tag collection creation and editing functionality. The affected actions accepted the full HTTP request payload and passed it to a bulk-association save operation, which writes not only the intended tag collection record but also any associated model data present in the payload.\u003c/p\u003e\u003cp\u003eA user holding the tag editor permission could craft a request that includes additional model data (such as User or Organisation records) alongside the tag collection fields. Because the save operation processed all associated models indiscriminately, the injected sibling records were written to the database, enabling the attacker to modify or create privileged accounts and escalate to site administrator.\u003c/p\u003e\u003cp\u003ePreconditions:\u003c/p\u003e\u003cp\u003e- An authenticated account with the tag editor permission (perm_tag_editor)\u003c/p\u003e\u003cp\u003e- Network access to the MISP instance\u003c/p\u003e\u003cp\u003eImpact:\u003c/p\u003e\u003cp\u003e- Unauthorized creation or modification of User and Organisation records\u003c/p\u003e\u003cp\u003e- Privilege escalation from tag editor to site administrator\u003c/p\u003e\u003cp\u003eAffected versions: \u0026lt; 2.5.48\u003c/p\u003e"
}
],
"value": "MISP contains a privilege escalation vulnerability in the tag collection creation and editing functionality. The affected actions accepted the full HTTP request payload and passed it to a bulk-association save operation, which writes not only the intended tag collection record but also any associated model data present in the payload.\n\nA user holding the tag editor permission could craft a request that includes additional model data (such as User or Organisation records) alongside the tag collection fields. Because the save operation processed all associated models indiscriminately, the injected sibling records were written to the database, enabling the attacker to modify or create privileged accounts and escalate to site administrator.\n\nPreconditions:\n\n- An authenticated account with the tag editor permission (perm_tag_editor)\n\n- Network access to the MISP instance\n\nImpact:\n\n- Unauthorized creation or modification of User and Organisation records\n\n- Privilege escalation from tag editor to site administrator\n\nAffected versions: \u003c 2.5.48"
}
],
"impacts": [
{
"capecId": "CAPEC-126",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-126 Parameter Tampering"
}
]
},
{
"capecId": "CAPEC-1",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-1 Improper Input Validation"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 8.6,
"baseSeverity": "HIGH",
"privilegesRequired": "LOW",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "NONE",
"subIntegrityImpact": "NONE",
"userInteraction": "NONE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "HIGH",
"vulnIntegrityImpact": "HIGH",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-284",
"description": "CWE-284 Improper Access Control",
"lang": "en",
"type": "CWE"
}
]
},
{
"descriptions": [
{
"cweId": "CWE-862",
"description": "CWE-862 Missing Authorization",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/MISP/MISP/commit/96f735e7b"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe fix replaces the bulk-association save call with an explicit two-step process: first, only the TagCollection data is extracted from the request and saved via a plain save() operation that does not write belongsTo siblings; second, tag association rows are persisted individually in a controlled loop. This ensures that any User, Organisation, or other sibling model data present in the request payload is silently discarded and never reaches the database, eliminating the privilege escalation path.\u003c/p\u003e"
}
],
"value": "The fix replaces the bulk-association save call with an explicit two-step process: first, only the TagCollection data is extracted from the request and saved via a plain save() operation that does not write belongsTo siblings; second, tag association rows are persisted individually in a controlled loop. This ensures that any User, Organisation, or other sibling model data present in the request payload is silently discarded and never reaches the database, eliminating the privilege escalation path."
}
],
"title": "MISP Tag Collection Save Allows Privilege Escalation via Sibling Model Injection",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The affected version boundary is inferred from the tag_version_boundary metadata (v2.5.48 with 24 commits after fix), suggesting the fix is included in v2.5.48. No explicit version range is stated in the patch itself.",
"The privilege level required is assumed to be \u0027perm_tag_editor\u0027 based on the commit message; the exact permission model and whether other roles are affected is not fully verifiable from the patch alone.",
"CAPEC-126 (Parameter Tampering) is selected as the closest match; the attack is more precisely a mass-assignment / sibling-model injection via an ORM bulk-save, for which no exact CAPEC exists. CAPEC-126 is the best available approximation.",
"The CVSS assumes the attacker can reach the MISP web interface over the network and holds a valid session with tag editor permissions. No multi-step or race-condition requirements are evident.",
"The Co-Authored-By line references an AI assistant (Claude Opus 4.8); it is credited as a tool rather than a human remediation developer."
],
"capecRationale": [
{
"capecId": "CAPEC-126",
"rationale": "The attacker tampers with the HTTP request by injecting additional model fields (User, Organisation) into the tag collection payload. The application\u0027s bulk-save logic processes these unexpected parameters without filtering, leading to unauthorized record creation or modification. This is the closest CAPEC to the observed attack: adding extra parameters to a legitimate request to trigger unintended side effects."
},
{
"capecId": "CAPEC-1",
"rationale": "The application accepted the entire request body without validating or restricting which model keys were present before passing it to the persistence layer. This is a secondary mapping; CAPEC-126 is more specific to the parameter-injection attack vector observed here."
}
],
"commit": "96f735e7b5d9e28ea2eaf3d8a02ab800bad5c9ed",
"confidence": "high",
"credits": [
{
"lang": "en",
"type": "reporter",
"value": "Jeroen Pinoy"
},
{
"lang": "en",
"type": "remediation developer",
"value": "iglocska"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Opus 4.8"
}
],
"cvssRationale": "The vulnerability is exploitable over the network (AV:N) with low complexity (AC:L) by simply adding extra fields to a legitimate tag collection request. No special attack conditions are required (AT:N). The attacker needs an authenticated account with the tag editor permission (PR:L). No user interaction is needed (UI:N). The primary impact is on integrity (VI:H) because the attacker can create or modify User and Organisation records to escalate to site admin. Modifications can be made so data can be see (VC:H) but no availability impact (VA:N). No subsequent component is affected (SC:N, SI:N, SA:N).",
"fixSummary": "The fix replaces the bulk-association save call with an explicit two-step process: first, only the TagCollection data is extracted from the request and saved via a plain save() operation that does not write belongsTo siblings; second, tag association rows are persisted individually in a controlled loop. This ensures that any User, Organisation, or other sibling model data present in the request payload is silently discarded and never reaches the database, eliminating the privilege escalation path.",
"generatedAt": "2026-09-30T09:58:24.508004Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 5,
"confidence": "high",
"model": "qwen3.8:27b",
"score": 6
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "67f9b5741b88384cd891423b2a8b9f5c8e071b5895b93eb337a16c7f780ee88c",
"patchSummary": "In TagCollectionsController.php, both addWithTags() and editWithTags() were modified. The full $this-\u003erequest-\u003edata is no longer passed to saveAssociated(). Instead, only the TagCollection key is extracted from the request. The saveAssociated() call is replaced with a plain save() for the collection, followed by an explicit loop that creates and saves each TagCollectionTag row individually. In editWithTags(), the tag rewrite logic is restructured to delete existing tag associations and re-insert them after the collection save. REST success/failure response handling is added to both methods.",
"patchTruncated": false,
"patches": [
{
"commit": "96f735e7b5d9e28ea2eaf3d8a02ab800bad5c9ed",
"patchSha256": "67f9b5741b88384cd891423b2a8b9f5c8e071b5895b93eb337a16c7f780ee88c",
"source": "https://github.com/MISP/MISP/commit/96f735e7b.patch",
"sourceUrl": "https://github.com/MISP/MISP/commit/96f735e7b.patch",
"subject": "fix: [security] Tag collection saves no longer write sibling"
}
],
"source": "https://github.com/MISP/MISP/commit/96f735e7b.patch",
"subject": "fix: [security] Tag collection saves no longer write sibling",
"tagVersionBoundary": {
"commits_after_fix": 24,
"repository": "https://github.com/MISP/MISP",
"tag": "v2.5.48",
"version": "2.5.48",
"version_type": "semver"
},
"weaknessRationale": [
{
"cweId": "CWE-284",
"rationale": "A user with only tag editor permissions was able to write to User and Organisation records through the bulk-association save, bypassing the intended access control boundaries. The authorization model did not restrict which associated models could be persisted."
},
{
"cweId": "CWE-862",
"rationale": "The saveAssociated() call did not enforce per-model authorization checks, allowing any associated model data in the payload to be written regardless of the caller\u0027s permissions for those models."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20019"
}
]
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-103239",
"datePublished": "2026-09-30T10:16:09.364718Z",
"dateReserved": "2026-09-30T10:16:15.941Z",
"dateUpdated": "2026-09-30T10:16:16.553111Z",
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20019"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20280 (CVE-2026-103237)
Vulnerability from gna-1 – Published: 2026-09-30 09:56 – Updated: 2026-09-30 09:56| URL | Tags |
|---|---|
| https://github.com/MISP/MISP/commit/9485ae40d | patch |
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-30 09:24 - Model
qwen3.8:27b- Input
-
https://github.com/MISP/MISP/commit/9485ae40d.patch
03af388a5ac0… - Confidence
- high
| Commit | Subject | Patch SHA-256 |
|---|---|---|
9485ae40d334
|
fix: [security] A nested model alias key no longer selects | 03af388a5ac0… |
Fix summary
The fix introduces a defensive save() override in the base model class that refuses to persist any record where the data array simultaneously contains a nested key matching the model alias and other top-level scalar fields, logging a warning and returning false. Additionally, all code paths that sanitize and save records (free-text import, module result processing, object delta merge, attribute bulk edit, sighting capture, shadow attribute proposal, event report creation) now explicitly unset the nested alias key from the data array before calling save(), ensuring the ORM cannot be redirected to an attacker-chosen row. Controller-level request reshaping was also corrected to avoid creating self-referencing data structures.
Patch summary
Added a save() override in AppModel.php that detects and rejects ambiguous payloads containing both a nested model-alias key and outer scalars. Added unset($data[$this->alias]) calls in Event.php (free-text, module results, object attribute save), MispAttribute.php (saveAttributes, captureAttribute, editAttributeBulk), MispObject.php (deltaMerge, editObject), ShadowAttribute.php (__preCaptureMassage), and Sighting.php (captureSightings) to strip the nested alias key before save. Changed six controller files to use $this->request->data = array('Model' => $this->request->data) instead of $this->request->data['Model'] = $this->request->data, preventing self-referencing structures. Added a new regression test suite (tests/testregressions.py) with cross-tenant attack scenarios and wired it into the CI workflow.
CVSS rationale
AV:N - MISP is a network-accessible web application. AC:L - The attack requires only crafting a request with a nested alias key; no race condition or complex state is needed. AT:N - No in-transit tampering required. PR:L - An authenticated user with basic write permission (perm_add) suffices; no admin or sync role needed. UI:N - No victim interaction required. VC:N - The attacker does not gain new read access; the impact is on data they can already partially see or infer. VI:H - The attacker can overwrite, re-parent, or soft-delete rows in the same instance, causing high integrity loss to the vulnerable component's data. VA:N - Soft-delete is a state change (integrity) rather than a service disruption. SC:N - No impact on separate components' confidentiality. SI:H - The cross-tenant nature means integrity of other organizations' data (a separate security scope) is compromised. SA:N - No security mechanism is weakened.
Weakness rationale
- CWE-639 The attacker controls the row identifier (id) inside the nested alias block, which the ORM uses to select the target row. The application's authorization and sanitization (id stripping, event_id pinning) is applied to the outer record and is bypassed because the ORM binds to the inner record. This is a direct case of a user-controlled key selecting an unauthorized resource.
- CWE-20 The application fails to validate or strip the nested model-alias key from user-supplied data before passing it to the ORM. The ORM's set() method interprets this key as the record to save, contradicting the application's intent. The absence of validation on this structural aspect of the input is the root cause.
Attack pattern rationale
- CAPEC-24 The attacker tampers with the structure of request parameters by injecting a nested key matching the model alias, causing the ORM to target a different row than the application intended. The outer parameters (sanitized id, pinned event_id) are effectively ignored in favor of the attacker-controlled inner parameters. This is a structural parameter tampering attack exploiting the ORM's data-binding semantics. The mapping is the closest available CAPEC; no more specific pattern for ORM-level key injection exists in the CAPEC catalog.
Assumptions to verify
- The affected version boundary is inferred from the tag_version_boundary metadata (v2.5.48); the exact last affected release and first fixed release are not explicitly stated in the patch.
- PR:L assumes the attacker needs only a basic authenticated account with perm_add; the regression tests confirm a non-admin, non-sync role suffices, but the minimum permission set is not exhaustively enumerated in the patch.
- VA:N assumes soft-delete is treated as an integrity impact (data state change) rather than an availability impact; if the organization considers soft-deleted records as unavailable, VA could be raised to L.
- CAPEC-24 (Parameter Tampering) is the closest available pattern; the specific ORM-level key-binding manipulation does not have a dedicated CAPEC entry, so the mapping is approximate.
- The Co-Authored-By line references an AI tool (Claude Opus 5); it is listed as a tool credit, not a human remediation developer, per CVE credit role semantics.
- The commit date is 2026-09-25; the vulnerability may have existed for an unknown duration prior to reporting.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
5 | 9 | high | 6 |
{
"containers": {
"cna": {
"affected": [
{
"cpes": [
"cpe:2.3:a:misp:misp:*:*:*:*:*:*:*:*"
],
"modules": [
"AttributesController",
"EventReportsController",
"EventsController",
"ObjectReferencesController",
"ShadowAttributesController",
"UsersController",
"AppModel",
"Event model",
"MispAttribute model",
"MispObject model",
"ShadowAttribute model",
"Sighting model"
],
"product": "MISP",
"repo": "https://github.com/MISP/MISP",
"vendor": "MISP",
"versions": [
{
"lessThan": "2.5.48",
"status": "affected",
"version": "0",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "reporter",
"value": "Jeroen Pinoy"
},
{
"lang": "en",
"type": "remediation developer",
"value": "iglocska"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Opus 5"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eMISP contains an improper input validation vulnerability in its ORM save path. When a user submits data through various endpoints (attribute add/edit, event edit, free-text import, sighting capture, shadow attribute proposal, event report creation, object reference add, user admin edit), the application sanitizes the flat record by stripping the primary key and pinning the event_id or object_id to the caller\u0027s context. However, the underlying ORM\u0027s set() method gives priority to a nested key whose name matches the model alias and discards the outer scalar fields.\u003c/p\u003e\u003cp\u003eAn authenticated user with basic write permissions can exploit this by embedding a nested block under the model alias key inside their request. The sanitization logic (id removal, event_id pinning) is applied to the outer record, but the ORM binds to the inner record instead, which carries an attacker-chosen id and event_id. This allows the attacker to overwrite, re-parent, or soft-delete rows belonging to other organizations or events they have no read access to.\u003c/p\u003e\u003cp\u003eImpact:\u003c/p\u003e\u003cp\u003e- Cross-tenant data integrity compromise (attribute values rewritten, objects re-parented to attacker events, rows soft-deleted)\u003c/p\u003e\u003cp\u003e- Affects multiple entity types: Attribute, Object, EventReport, Sighting, AttributeTag, ShadowAttribute\u003c/p\u003e\u003cp\u003e- Requires only a low-privilege authenticated account with perm_add\u003c/p\u003e\u003cp\u003eAffected versions: \u0026lt;2.5.48\u003c/p\u003e"
}
],
"value": "MISP contains an improper input validation vulnerability in its ORM save path. When a user submits data through various endpoints (attribute add/edit, event edit, free-text import, sighting capture, shadow attribute proposal, event report creation, object reference add, user admin edit), the application sanitizes the flat record by stripping the primary key and pinning the event_id or object_id to the caller\u0027s context. However, the underlying ORM\u0027s set() method gives priority to a nested key whose name matches the model alias and discards the outer scalar fields.\n\nAn authenticated user with basic write permissions can exploit this by embedding a nested block under the model alias key inside their request. The sanitization logic (id removal, event_id pinning) is applied to the outer record, but the ORM binds to the inner record instead, which carries an attacker-chosen id and event_id. This allows the attacker to overwrite, re-parent, or soft-delete rows belonging to other organizations or events they have no read access to.\n\nImpact:\n\n- Cross-tenant data integrity compromise (attribute values rewritten, objects re-parented to attacker events, rows soft-deleted)\n\n- Affects multiple entity types: Attribute, Object, EventReport, Sighting, AttributeTag, ShadowAttribute\n\n- Requires only a low-privilege authenticated account with perm_add\n\nAffected versions: \u003c2.5.48"
}
],
"impacts": [
{
"capecId": "CAPEC-24",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-24 Parameter Tampering"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 8.3,
"baseSeverity": "HIGH",
"privilegesRequired": "LOW",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "NONE",
"subIntegrityImpact": "HIGH",
"userInteraction": "NONE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:H/VA:N/SC:N/SI:H/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "NONE",
"vulnIntegrityImpact": "HIGH",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-639",
"description": "CWE-639 Authorization Bypass Through User-Controlled Key",
"lang": "en",
"type": "CWE"
}
]
},
{
"descriptions": [
{
"cweId": "CWE-20",
"description": "CWE-20 Improper Input Validation",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/MISP/MISP/commit/9485ae40d"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe fix introduces a defensive save() override in the base model class that refuses to persist any record where the data array simultaneously contains a nested key matching the model alias and other top-level scalar fields, logging a warning and returning false. Additionally, all code paths that sanitize and save records (free-text import, module result processing, object delta merge, attribute bulk edit, sighting capture, shadow attribute proposal, event report creation) now explicitly unset the nested alias key from the data array before calling save(), ensuring the ORM cannot be redirected to an attacker-chosen row. Controller-level request reshaping was also corrected to avoid creating self-referencing data structures.\u003c/p\u003e"
}
],
"value": "The fix introduces a defensive save() override in the base model class that refuses to persist any record where the data array simultaneously contains a nested key matching the model alias and other top-level scalar fields, logging a warning and returning false. Additionally, all code paths that sanitize and save records (free-text import, module result processing, object delta merge, attribute bulk edit, sighting capture, shadow attribute proposal, event report creation) now explicitly unset the nested alias key from the data array before calling save(), ensuring the ORM cannot be redirected to an attacker-chosen row. Controller-level request reshaping was also corrected to avoid creating self-referencing data structures."
}
],
"title": "MISP: Nested Model Alias Key Bypasses Sanitization to Modify Cross-Tenant Rows",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The affected version boundary is inferred from the tag_version_boundary metadata (v2.5.48); the exact last affected release and first fixed release are not explicitly stated in the patch.",
"PR:L assumes the attacker needs only a basic authenticated account with perm_add; the regression tests confirm a non-admin, non-sync role suffices, but the minimum permission set is not exhaustively enumerated in the patch.",
"VA:N assumes soft-delete is treated as an integrity impact (data state change) rather than an availability impact; if the organization considers soft-deleted records as unavailable, VA could be raised to L.",
"CAPEC-24 (Parameter Tampering) is the closest available pattern; the specific ORM-level key-binding manipulation does not have a dedicated CAPEC entry, so the mapping is approximate.",
"The Co-Authored-By line references an AI tool (Claude Opus 5); it is listed as a tool credit, not a human remediation developer, per CVE credit role semantics.",
"The commit date is 2026-09-25; the vulnerability may have existed for an unknown duration prior to reporting."
],
"capecRationale": [
{
"capecId": "CAPEC-24",
"rationale": "The attacker tampers with the structure of request parameters by injecting a nested key matching the model alias, causing the ORM to target a different row than the application intended. The outer parameters (sanitized id, pinned event_id) are effectively ignored in favor of the attacker-controlled inner parameters. This is a structural parameter tampering attack exploiting the ORM\u0027s data-binding semantics. The mapping is the closest available CAPEC; no more specific pattern for ORM-level key injection exists in the CAPEC catalog."
}
],
"commit": "9485ae40d334471882e3bd246651acaa68feef34",
"confidence": "high",
"credits": [
{
"lang": "en",
"type": "reporter",
"value": "Jeroen Pinoy"
},
{
"lang": "en",
"type": "remediation developer",
"value": "iglocska"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Opus 5"
}
],
"cvssRationale": "AV:N - MISP is a network-accessible web application. AC:L - The attack requires only crafting a request with a nested alias key; no race condition or complex state is needed. AT:N - No in-transit tampering required. PR:L - An authenticated user with basic write permission (perm_add) suffices; no admin or sync role needed. UI:N - No victim interaction required. VC:N - The attacker does not gain new read access; the impact is on data they can already partially see or infer. VI:H - The attacker can overwrite, re-parent, or soft-delete rows in the same instance, causing high integrity loss to the vulnerable component\u0027s data. VA:N - Soft-delete is a state change (integrity) rather than a service disruption. SC:N - No impact on separate components\u0027 confidentiality. SI:H - The cross-tenant nature means integrity of other organizations\u0027 data (a separate security scope) is compromised. SA:N - No security mechanism is weakened.",
"fixSummary": "The fix introduces a defensive save() override in the base model class that refuses to persist any record where the data array simultaneously contains a nested key matching the model alias and other top-level scalar fields, logging a warning and returning false. Additionally, all code paths that sanitize and save records (free-text import, module result processing, object delta merge, attribute bulk edit, sighting capture, shadow attribute proposal, event report creation) now explicitly unset the nested alias key from the data array before calling save(), ensuring the ORM cannot be redirected to an attacker-chosen row. Controller-level request reshaping was also corrected to avoid creating self-referencing data structures.",
"generatedAt": "2026-09-30T09:24:37.816321Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 6,
"confidence": "high",
"model": "qwen3.8:27b",
"score": 5
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "03af388a5ac0d93cebef902e0050aa8496aae8d88d1496b80abc0b9a929cb39a",
"patchSummary": "Added a save() override in AppModel.php that detects and rejects ambiguous payloads containing both a nested model-alias key and outer scalars. Added unset($data[$this-\u003ealias]) calls in Event.php (free-text, module results, object attribute save), MispAttribute.php (saveAttributes, captureAttribute, editAttributeBulk), MispObject.php (deltaMerge, editObject), ShadowAttribute.php (__preCaptureMassage), and Sighting.php (captureSightings) to strip the nested alias key before save. Changed six controller files to use $this-\u003erequest-\u003edata = array(\u0027Model\u0027 =\u003e $this-\u003erequest-\u003edata) instead of $this-\u003erequest-\u003edata[\u0027Model\u0027] = $this-\u003erequest-\u003edata, preventing self-referencing structures. Added a new regression test suite (tests/testregressions.py) with cross-tenant attack scenarios and wired it into the CI workflow.",
"patchTruncated": false,
"patches": [
{
"commit": "9485ae40d334471882e3bd246651acaa68feef34",
"patchSha256": "03af388a5ac0d93cebef902e0050aa8496aae8d88d1496b80abc0b9a929cb39a",
"source": "https://github.com/MISP/MISP/commit/9485ae40d.patch",
"sourceUrl": "https://github.com/MISP/MISP/commit/9485ae40d.patch",
"subject": "fix: [security] A nested model alias key no longer selects"
}
],
"source": "https://github.com/MISP/MISP/commit/9485ae40d.patch",
"subject": "fix: [security] A nested model alias key no longer selects",
"tagVersionBoundary": {
"commits_after_fix": 23,
"repository": "https://github.com/MISP/MISP",
"tag": "v2.5.48",
"version": "2.5.48",
"version_type": "semver"
},
"weaknessRationale": [
{
"cweId": "CWE-639",
"rationale": "The attacker controls the row identifier (id) inside the nested alias block, which the ORM uses to select the target row. The application\u0027s authorization and sanitization (id stripping, event_id pinning) is applied to the outer record and is bypassed because the ORM binds to the inner record. This is a direct case of a user-controlled key selecting an unauthorized resource."
},
{
"cweId": "CWE-20",
"rationale": "The application fails to validate or strip the nested model-alias key from user-supplied data before passing it to the ORM. The ORM\u0027s set() method interprets this key as the record to save, contradicting the application\u0027s intent. The absence of validation on this structural aspect of the input is the root cause."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20280"
}
]
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-103237",
"datePublished": "2026-09-30T09:56:28.525185Z",
"dateReserved": "2026-09-30T09:56:33.891Z",
"dateUpdated": "2026-09-30T09:56:33.959884Z",
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20280"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20227 (CVE-2026-103235)
Vulnerability from gna-1 – Published: 2026-09-30 09:09 – Updated: 2026-09-30 09:12| URL | Tags |
|---|---|
| https://github.com/MISP/MISP/commit/d1f5684f9 | patch |
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-30 07:37 - Model
qwen3.8:27b- Input
-
https://github.com/MISP/MISP/commit/d1f5684f9.patch
f478751165e6… - Confidence
- high
| Commit | Subject | Patch SHA-256 |
|---|---|---|
d1f5684f9a19
|
fix: [security] Delegation requests stay bound to the event | f478751165e6… |
Fix summary
The delegation record is now constructed from a strict allow-list of fields rather than persisting the raw user-submitted payload. The event_id is always derived from the authorized event in the URL, the requester_org_id is always taken from the authenticated session, and the primary key is never included in the saved data. Only message, distribution, and sharing_group_id are accepted from user input, eliminating the ability to retarget or overwrite existing delegation records.
Patch summary
In EventDelegationsController::delegateEvent(), the code previously saved $this->request->data['EventDelegation'] directly after setting a few fields. The fix replaces this with an explicit allow-list array containing only event_id (from the authorized URL event), requester_org_id (from the session), org_id (resolved from submitted UUID), message, distribution, and sharing_group_id. The primary key id is never included. A regression test class DelegationRequestRetargeting was added to verify that injecting a nested EventDelegation with a foreign id and event_id does not grant read access to the victim event.
CVSS rationale
Network vector: MISP is a web application accessible over HTTP. Low complexity: a single crafted POST request suffices. No attack target manipulation. Low privileges: requires an authenticated user with perm_delegate. No user interaction: read access is granted immediately upon creating the retargeted delegation. High confidentiality: grants read access to any event on the instance. High integrity: overwrites existing delegation records and can transfer event ownership. No availability impact without victim acceptance. Scope change (SC:H, SI:H): the vulnerability crosses organisational boundaries, affecting data owned by other organisations. No sub-system availability impact.
Weakness rationale
- CWE-915 The application persisted the entire user-submitted record including fields (id, event_id) that should not be attacker-controllable, allowing mass assignment of sensitive fields to retarget the delegation.
- CWE-639 The authorization check validated only the event in the URL, but the attacker-supplied primary key or event_id in the payload redirected the operation to a different record, bypassing the intended authorization boundary.
Attack pattern rationale
- CAPEC-12 The attacker manipulates hidden or additional fields in a form/API request (injecting id and event_id into the EventDelegation payload) to modify data beyond what the application intended to accept. This is the canonical mass assignment pattern: the server processes user-supplied fields it should have ignored. The mapping is direct and well-supported by the patch evidence.
Assumptions to verify
- The affected version boundary (< 2.5.48) is inferred from the tag_version_boundary metadata showing v2.5.48 as the nearest tag with 22 commits after the fix; the exact last affected release is not explicitly stated in the patch.
- PR:L assumes the attacker needs only the perm_delegate permission, which is a non-admin role; the exact role configuration may vary by deployment.
- VA:N assumes the availability impact (deletion of the original event) requires victim acceptance and is therefore not a direct availability impact of the vulnerability itself.
- The CAPEC-12 mapping is the closest standard pattern; the vulnerability also has IDOR characteristics (CWE-639) but CAPEC-12 best captures the mass-assignment mechanism demonstrated in the patch.
- The MISP.delegation server setting must be enabled for the vulnerability to be exploitable; this is a deployment configuration assumption.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
6 | 9 | high | 5 |
{
"containers": {
"cna": {
"affected": [
{
"cpes": [
"cpe:2.3:a:misp:misp:*:*:*:*:*:*:*:*"
],
"defaultStatus": "unaffected",
"modules": [
"EventDelegationsController"
],
"product": "MISP",
"programFiles": [
"app/Controller/EventDelegationsController.php"
],
"repo": "https://github.com/MISP/MISP",
"vendor": "MISP",
"versions": [
{
"lessThan": "2.5.48",
"status": "affected",
"version": "0",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "reporter",
"value": "Jeroen Pinoy"
},
{
"lang": "en",
"type": "remediation developer",
"value": "iglocska"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Opus 5"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eMISP contains a mass assignment vulnerability in the event delegation feature. When a user with delegation permission submits a delegation request, the application authorized the user against the event identified in the URL but then persisted the entire submitted record, including caller-supplied fields such as the primary key and event_id.\u003c/p\u003e\u003cp\u003eAn authenticated attacker could inject a primary key or event_id into the delegation payload to retarget an existing delegation record to any event on the instance. Because a delegation row grants the requesting organisation read access to the event it references, this effectively granted read access to arbitrary events belonging to other organisations. If the target organisation subsequently accepted the delegation, ownership of the event was transferred and the original record was deleted.\u003c/p\u003e\u003cp\u003ePreconditions:\u003c/p\u003e\u003cp\u003e- An authenticated user with the delegation permission (perm_delegate)\u003c/p\u003e\u003cp\u003e- The MISP.delegation server setting must be enabled\u003c/p\u003e\u003cp\u003eImpact:\u003c/p\u003e\u003cp\u003e- Confidentiality: read access to any event on the instance\u003c/p\u003e\u003cp\u003e- Integrity: overwriting existing delegation records and transferring event ownership\u003c/p\u003e\u003cp\u003eAffected versions: MISP \u0026lt; 2.5.48\u003c/p\u003e"
}
],
"value": "MISP contains a mass assignment vulnerability in the event delegation feature. When a user with delegation permission submits a delegation request, the application authorized the user against the event identified in the URL but then persisted the entire submitted record, including caller-supplied fields such as the primary key and event_id.\n\nAn authenticated attacker could inject a primary key or event_id into the delegation payload to retarget an existing delegation record to any event on the instance. Because a delegation row grants the requesting organisation read access to the event it references, this effectively granted read access to arbitrary events belonging to other organisations. If the target organisation subsequently accepted the delegation, ownership of the event was transferred and the original record was deleted.\n\nPreconditions:\n\n- An authenticated user with the delegation permission (perm_delegate)\n\n- The MISP.delegation server setting must be enabled\n\nImpact:\n\n- Confidentiality: read access to any event on the instance\n\n- Integrity: overwriting existing delegation records and transferring event ownership\n\nAffected versions: MISP \u003c 2.5.48"
}
],
"impacts": [
{
"capecId": "CAPEC-12",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-12 Mass Assignment"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 8.7,
"baseSeverity": "HIGH",
"privilegesRequired": "LOW",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "NONE",
"subIntegrityImpact": "NONE",
"userInteraction": "NONE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:L/SC:N/SI:N/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "LOW",
"vulnConfidentialityImpact": "HIGH",
"vulnIntegrityImpact": "HIGH",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-915",
"description": "CWE-915 Improperly Controlled Modification of Dynamically-Determined Object Attributes",
"lang": "en",
"type": "CWE"
}
]
},
{
"descriptions": [
{
"cweId": "CWE-639",
"description": "CWE-639 Authorization Bypass Through User-Controlled Key",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/MISP/MISP/commit/d1f5684f9"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe delegation record is now constructed from a strict allow-list of fields rather than persisting the raw user-submitted payload. The event_id is always derived from the authorized event in the URL, the requester_org_id is always taken from the authenticated session, and the primary key is never included in the saved data. Only message, distribution, and sharing_group_id are accepted from user input, eliminating the ability to retarget or overwrite existing delegation records.\u003c/p\u003e"
}
],
"value": "The delegation record is now constructed from a strict allow-list of fields rather than persisting the raw user-submitted payload. The event_id is always derived from the authorized event in the URL, the requester_org_id is always taken from the authenticated session, and the primary key is never included in the saved data. Only message, distribution, and sharing_group_id are accepted from user input, eliminating the ability to retarget or overwrite existing delegation records."
}
],
"source": {
"discovery": "UNKNOWN"
},
"title": "MISP Event Delegation Mass Assignment Allows Retargeting Delegation to Arbitrary Events",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The affected version boundary (\u003c 2.5.48) is inferred from the tag_version_boundary metadata showing v2.5.48 as the nearest tag with 22 commits after the fix; the exact last affected release is not explicitly stated in the patch.",
"PR:L assumes the attacker needs only the perm_delegate permission, which is a non-admin role; the exact role configuration may vary by deployment.",
"VA:N assumes the availability impact (deletion of the original event) requires victim acceptance and is therefore not a direct availability impact of the vulnerability itself.",
"The CAPEC-12 mapping is the closest standard pattern; the vulnerability also has IDOR characteristics (CWE-639) but CAPEC-12 best captures the mass-assignment mechanism demonstrated in the patch.",
"The MISP.delegation server setting must be enabled for the vulnerability to be exploitable; this is a deployment configuration assumption."
],
"capecRationale": [
{
"capecId": "CAPEC-12",
"rationale": "The attacker manipulates hidden or additional fields in a form/API request (injecting id and event_id into the EventDelegation payload) to modify data beyond what the application intended to accept. This is the canonical mass assignment pattern: the server processes user-supplied fields it should have ignored. The mapping is direct and well-supported by the patch evidence."
}
],
"commit": "d1f5684f9a193ea0a8a4f7709703011aa69d9682",
"confidence": "high",
"credits": [
{
"lang": "en",
"type": "reporter",
"value": "Jeroen Pinoy"
},
{
"lang": "en",
"type": "remediation developer",
"value": "iglocska"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Opus 5"
}
],
"cvssRationale": "Network vector: MISP is a web application accessible over HTTP. Low complexity: a single crafted POST request suffices. No attack target manipulation. Low privileges: requires an authenticated user with perm_delegate. No user interaction: read access is granted immediately upon creating the retargeted delegation. High confidentiality: grants read access to any event on the instance. High integrity: overwrites existing delegation records and can transfer event ownership. No availability impact without victim acceptance. Scope change (SC:H, SI:H): the vulnerability crosses organisational boundaries, affecting data owned by other organisations. No sub-system availability impact.",
"fixSummary": "The delegation record is now constructed from a strict allow-list of fields rather than persisting the raw user-submitted payload. The event_id is always derived from the authorized event in the URL, the requester_org_id is always taken from the authenticated session, and the primary key is never included in the saved data. Only message, distribution, and sharing_group_id are accepted from user input, eliminating the ability to retarget or overwrite existing delegation records.",
"generatedAt": "2026-09-30T07:37:05.841152Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 5,
"confidence": "high",
"model": "qwen3.8:27b",
"score": 6
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "f478751165e658f2b26822845ec034386d8a1740527cf30863792e9521f24611",
"patchSummary": "In EventDelegationsController::delegateEvent(), the code previously saved $this-\u003erequest-\u003edata[\u0027EventDelegation\u0027] directly after setting a few fields. The fix replaces this with an explicit allow-list array containing only event_id (from the authorized URL event), requester_org_id (from the session), org_id (resolved from submitted UUID), message, distribution, and sharing_group_id. The primary key id is never included. A regression test class DelegationRequestRetargeting was added to verify that injecting a nested EventDelegation with a foreign id and event_id does not grant read access to the victim event.",
"patchTruncated": false,
"patches": [
{
"commit": "d1f5684f9a193ea0a8a4f7709703011aa69d9682",
"patchSha256": "f478751165e658f2b26822845ec034386d8a1740527cf30863792e9521f24611",
"source": "https://github.com/MISP/MISP/commit/d1f5684f9.patch",
"sourceUrl": "https://github.com/MISP/MISP/commit/d1f5684f9.patch",
"subject": "fix: [security] Delegation requests stay bound to the event"
}
],
"source": "https://github.com/MISP/MISP/commit/d1f5684f9.patch",
"subject": "fix: [security] Delegation requests stay bound to the event",
"tagVersionBoundary": {
"commits_after_fix": 22,
"repository": "https://github.com/MISP/MISP",
"tag": "v2.5.48",
"version": "2.5.48",
"version_type": "semver"
},
"weaknessRationale": [
{
"cweId": "CWE-915",
"rationale": "The application persisted the entire user-submitted record including fields (id, event_id) that should not be attacker-controllable, allowing mass assignment of sensitive fields to retarget the delegation."
},
{
"cweId": "CWE-639",
"rationale": "The authorization check validated only the event in the URL, but the attacker-supplied primary key or event_id in the payload redirected the operation to a different record, bypassing the intended authorization boundary."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20227"
}
],
"x_generator": {
"engine": "Vulnogram 0.2.0"
}
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-103235",
"datePublished": "2026-09-30T09:09:00.000Z",
"dateReserved": "2026-09-30T09:09:00.000Z",
"dateUpdated": "2026-09-30T09:12:45.330040Z",
"requesterUserId": "00000000-0000-4000-9000-000000000000",
"serial": 1,
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20227"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20252 (CVE-2026-101041)
Vulnerability from gna-1 – Published: 2026-09-27 14:47 – Updated: 2026-09-27 14:47| Vendor | Product | Version | |
|---|---|---|---|
| vulnerability-lookup | vulnerability-lookup |
Affected:
0 , ≤ 6.2.0
(semver)
|
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-27 14:43 - Model
qwen3.8:27b- Input
-
patch set (2 sources)
ccb7572963e8… - Confidence
- medium
| Commit | Subject | Patch SHA-256 |
|---|---|---|
5462bab62d76
|
Fix recovery token consumption race | 75ed8b215e27… |
ad6f22882975
|
fix: [user] Validate the recovery form and harden the token | c1cecff5ac5f… |
Fix summary
The non-atomic check-then-consume pattern is replaced with a single conditional UPDATE statement (compare-and-set) that atomically verifies the stored SHA-256 nonce digest, updates the password hash, sets is_confirmed, and clears the token in one database operation. Only the first transaction to commit succeeds; all concurrent attempts match zero rows and are rejected with an error. The view now calls form.validate() before processing the password change, enforcing the form's length and equality validators. The dead consume_account_token() method is removed to eliminate the non-atomic consumption path entirely.
Patch summary
Added reset_password_with_account_token() to the User model performing a single conditional UPDATE (WHERE id = ? AND account_token_digest = sha256(nonce)) that sets pwdhash, is_confirmed=True, and account_token_digest=None atomically. Removed the separate consume_account_token() method. Changed the confirm_account view to call form.validate() instead of only comparing password1 and password2 manually. Introduced a _account_token_digest() static helper to centralize the SHA-256 computation (previously three inline calls). Removed synchronize_session=False from the UPDATE execution options so the ORM session stays consistent. Added regression tests for concurrent token consumption (two threads, barrier-synchronized) and for rejection of passwords that fail form validation (empty, three-character, mismatched).
CVSS rationale
AV:N: the recovery endpoint is reachable over the network. AC:H: exploiting the race condition requires precise timing of two concurrent requests, making the attack condition high complexity. AT:N: no active user interaction or deception is required beyond possessing a valid token. PR:N: no prior authentication is needed; the recovery token itself is the credential. UI:N: no user interaction is required. VC:N: the vulnerability does not directly expose confidential data. VI:H: the integrity of the user's password (a critical credential) is directly compromised. VA:N: no availability impact. SC:N: the impact is confined to the vulnerable component. SI:H: the integrity of the user's account (password, confirmation state) is compromised, affecting the user's ability to access their own account. SA:N: no security-authority impact.
Weakness rationale
- CWE-362 The primary defect is a TOCTOU race: the token nonce is verified and then consumed in separate operations, allowing two concurrent transactions to both pass verification before either commits. The fix replaces this with an atomic compare-and-set UPDATE, confirming the race condition as the root cause.
- CWE-20 The confirm_account view never called form.validate() on POST, so the form's minimum-length and equality validators were bypassed. A valid recovery link alone permitted setting an empty or three-character password. The fix adds the form.validate() call before processing.
Attack pattern rationale
- CAPEC-111 The attacker exploits the window between token verification and token consumption by issuing a concurrent request with the same token. Both requests pass the check; the last to commit overwrites the first's password. This is a textbook race-condition attack on a shared resource (the single-use token). CAPEC-111 is the closest available pattern; no CAPEC specifically covers TOCTOU on database-level token consumption, so this is the best match.
Assumptions to verify
- The product version and exact release boundary are not specified in the patch metadata; the affected version range is recorded as unspecified.
- The CAPEC-111 mapping is the closest available pattern for a database-level TOCTOU race on a single-use token; no CAPEC specifically describes concurrent compare-and-set exploitation, so this is the best available match.
- The CVSS AC:H reflects the race-condition aspect; the missing-validation aspect alone would be AC:L, but the patch set is treated as one vulnerability per the caller's instruction.
- The vendor is not explicitly identified in the metadata; the GitHub repository name 'vulnerability-lookup' is used as the product identifier.
- No specific authentication or deployment preconditions beyond possession of a valid recovery token are stated in the patch; the token is assumed to be delivered out-of-band (e.g., email).
- The Co-Authored-By line for Claude Fable 5.1 is treated as a remediation developer credit per the metadata's remediation_developers list; no finder or reporter is identified in the supplied evidence.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
4 | 9 | medium | 6 |
{
"containers": {
"cna": {
"affected": [
{
"modules": [
"website/models/user.py",
"website/web/views/user.py",
"website/lib/user_utils.py"
],
"product": "vulnerability-lookup",
"programFiles": [
"website/models/user.py",
"website/web/views/user.py",
"website/lib/user_utils.py"
],
"repo": "https://github.com/vulnerability-lookup/vulnerability-lookup",
"vendor": "vulnerability-lookup",
"versions": [
{
"lessThanOrEqual": "6.2.0",
"status": "affected",
"version": "0",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "Alexandre Dulaunoy"
},
{
"lang": "en",
"type": "remediation developer",
"value": "C\u00e9dric Bonhomme"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Fable 5.1"
},
{
"lang": "en",
"type": "finder",
"value": "avrlab233"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe account recovery (password reset) functionality in the vulnerability-lookup web application contains a time-of-check-to-time-of-use (TOCTOU) race condition in the consumption of single-use recovery tokens. The original implementation verified the token nonce against the stored digest and then consumed (cleared) it in separate database operations. Two concurrent HTTP requests presenting the same valid recovery token could both pass the verification check before either transaction committed, allowing both to set their own password on the target account. The last transaction to commit overwrites the first, enabling an attacker who possesses a valid recovery token to replace the legitimate user\u0027s password with one of their choosing.\u003c/p\u003e\u003cp\u003eA secondary defect in the same endpoint (confirm_account) allowed a valid recovery link to be used to set an empty or trivially short password (e.g., three characters). The view handler performed only a manual equality comparison between the two password fields and never invoked the form\u0027s validation logic, bypassing the intended minimum-length and complexity constraints.\u003c/p\u003e\u003cp\u003eThe affected component is the user account recovery endpoint (/user/confirm_account/\u0026lt;token\u0026gt;) and the associated token verification and consumption logic in the User model (website/models/user.py) and the view layer (website/web/views/user.py).\u003c/p\u003e"
}
],
"value": "The account recovery (password reset) functionality in the vulnerability-lookup web application contains a time-of-check-to-time-of-use (TOCTOU) race condition in the consumption of single-use recovery tokens. The original implementation verified the token nonce against the stored digest and then consumed (cleared) it in separate database operations. Two concurrent HTTP requests presenting the same valid recovery token could both pass the verification check before either transaction committed, allowing both to set their own password on the target account. The last transaction to commit overwrites the first, enabling an attacker who possesses a valid recovery token to replace the legitimate user\u0027s password with one of their choosing.\n\nA secondary defect in the same endpoint (confirm_account) allowed a valid recovery link to be used to set an empty or trivially short password (e.g., three characters). The view handler performed only a manual equality comparison between the two password fields and never invoked the form\u0027s validation logic, bypassing the intended minimum-length and complexity constraints.\n\nThe affected component is the user account recovery endpoint (/user/confirm_account/\u003ctoken\u003e) and the associated token verification and consumption logic in the User model (website/models/user.py) and the view layer (website/web/views/user.py)."
}
],
"impacts": [
{
"capecId": "CAPEC-111",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-111 Race Condition"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "HIGH",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 6.3,
"baseSeverity": "MEDIUM",
"privilegesRequired": "NONE",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "LOW",
"subIntegrityImpact": "NONE",
"userInteraction": "NONE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:H/AT:N/PR:N/UI:N/VC:L/VI:N/VA:N/SC:L/SI:N/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "LOW",
"vulnIntegrityImpact": "NONE",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-362",
"description": "CWE-362 Concurrency: Race Condition",
"lang": "en",
"type": "CWE"
}
]
},
{
"descriptions": [
{
"cweId": "CWE-20",
"description": "CWE-20 Improper Input Validation",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/vulnerability-lookup/vulnerability-lookup/commit/5462bab62d76df852619e01eb67da36c023c8c40"
},
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/vulnerability-lookup/vulnerability-lookup/commit/ad6f22882975516adf193a1a920aaa54025c71d4"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe non-atomic check-then-consume pattern is replaced with a single conditional UPDATE statement (compare-and-set) that atomically verifies the stored SHA-256 nonce digest, updates the password hash, sets is_confirmed, and clears the token in one database operation. Only the first transaction to commit succeeds; all concurrent attempts match zero rows and are rejected with an error. The view now calls form.validate() before processing the password change, enforcing the form\u0027s length and equality validators. The dead consume_account_token() method is removed to eliminate the non-atomic consumption path entirely.\u003c/p\u003e"
}
],
"value": "The non-atomic check-then-consume pattern is replaced with a single conditional UPDATE statement (compare-and-set) that atomically verifies the stored SHA-256 nonce digest, updates the password hash, sets is_confirmed, and clears the token in one database operation. Only the first transaction to commit succeeds; all concurrent attempts match zero rows and are rejected with an error. The view now calls form.validate() before processing the password change, enforcing the form\u0027s length and equality validators. The dead consume_account_token() method is removed to eliminate the non-atomic consumption path entirely."
}
],
"title": "Vulnerability-Lookup - Race Condition in Account Recovery Token Consumption Allows Password Takeover",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "partial",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The product version and exact release boundary are not specified in the patch metadata; the affected version range is recorded as unspecified.",
"The CAPEC-111 mapping is the closest available pattern for a database-level TOCTOU race on a single-use token; no CAPEC specifically describes concurrent compare-and-set exploitation, so this is the best available match.",
"The CVSS AC:H reflects the race-condition aspect; the missing-validation aspect alone would be AC:L, but the patch set is treated as one vulnerability per the caller\u0027s instruction.",
"The vendor is not explicitly identified in the metadata; the GitHub repository name \u0027vulnerability-lookup\u0027 is used as the product identifier.",
"No specific authentication or deployment preconditions beyond possession of a valid recovery token are stated in the patch; the token is assumed to be delivered out-of-band (e.g., email).",
"The Co-Authored-By line for Claude Fable 5.1 is treated as a remediation developer credit per the metadata\u0027s remediation_developers list; no finder or reporter is identified in the supplied evidence."
],
"capecRationale": [
{
"capecId": "CAPEC-111",
"rationale": "The attacker exploits the window between token verification and token consumption by issuing a concurrent request with the same token. Both requests pass the check; the last to commit overwrites the first\u0027s password. This is a textbook race-condition attack on a shared resource (the single-use token). CAPEC-111 is the closest available pattern; no CAPEC specifically covers TOCTOU on database-level token consumption, so this is the best match."
}
],
"commit": "ad6f22882975516adf193a1a920aaa54025c71d4",
"confidence": "medium",
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "Alexandre Dulaunoy"
},
{
"lang": "en",
"type": "remediation developer",
"value": "C\u00e9dric Bonhomme"
},
{
"lang": "en",
"type": "remediation developer",
"value": "Claude Fable 5.1"
}
],
"cvssRationale": "AV:N: the recovery endpoint is reachable over the network. AC:H: exploiting the race condition requires precise timing of two concurrent requests, making the attack condition high complexity. AT:N: no active user interaction or deception is required beyond possessing a valid token. PR:N: no prior authentication is needed; the recovery token itself is the credential. UI:N: no user interaction is required. VC:N: the vulnerability does not directly expose confidential data. VI:H: the integrity of the user\u0027s password (a critical credential) is directly compromised. VA:N: no availability impact. SC:N: the impact is confined to the vulnerable component. SI:H: the integrity of the user\u0027s account (password, confirmation state) is compromised, affecting the user\u0027s ability to access their own account. SA:N: no security-authority impact.",
"fixSummary": "The non-atomic check-then-consume pattern is replaced with a single conditional UPDATE statement (compare-and-set) that atomically verifies the stored SHA-256 nonce digest, updates the password hash, sets is_confirmed, and clears the token in one database operation. Only the first transaction to commit succeeds; all concurrent attempts match zero rows and are rejected with an error. The view now calls form.validate() before processing the password change, enforcing the form\u0027s length and equality validators. The dead consume_account_token() method is removed to eliminate the non-atomic consumption path entirely.",
"generatedAt": "2026-09-27T14:43:09.207406Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 6,
"confidence": "medium",
"model": "qwen3.8:27b",
"score": 4
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "ccb7572963e8b2435c2ee54cea019a60b1b70a5717cc3848ccd940232eae0b91",
"patchSummary": "Added reset_password_with_account_token() to the User model performing a single conditional UPDATE (WHERE id = ? AND account_token_digest = sha256(nonce)) that sets pwdhash, is_confirmed=True, and account_token_digest=None atomically. Removed the separate consume_account_token() method. Changed the confirm_account view to call form.validate() instead of only comparing password1 and password2 manually. Introduced a _account_token_digest() static helper to centralize the SHA-256 computation (previously three inline calls). Removed synchronize_session=False from the UPDATE execution options so the ORM session stays consistent. Added regression tests for concurrent token consumption (two threads, barrier-synchronized) and for rejection of passwords that fail form validation (empty, three-character, mismatched).",
"patchTruncated": false,
"patches": [
{
"commit": "5462bab62d76df852619e01eb67da36c023c8c40",
"patchSha256": "75ed8b215e2769b5d20b161ddfbde9e0b2975f836397979a94387340b0501ec1",
"source": "https://github.com/vulnerability-lookup/vulnerability-lookup/commit/5462bab62d76df852619e01eb67da36c023c8c40.patch",
"sourceUrl": "https://github.com/vulnerability-lookup/vulnerability-lookup/commit/5462bab62d76df852619e01eb67da36c023c8c40.patch",
"subject": "Fix recovery token consumption race"
},
{
"commit": "ad6f22882975516adf193a1a920aaa54025c71d4",
"patchSha256": "c1cecff5ac5f7804ff42f0583b156cefa2add775cb6f80506c917f84a26bf43d",
"source": "https://github.com/vulnerability-lookup/vulnerability-lookup/commit/ad6f22882975516adf193a1a920aaa54025c71d4.patch",
"sourceUrl": "https://github.com/vulnerability-lookup/vulnerability-lookup/commit/ad6f22882975516adf193a1a920aaa54025c71d4.patch",
"subject": "fix: [user] Validate the recovery form and harden the token"
}
],
"source": "patch set (2 sources)",
"subject": "fix: [user] Validate the recovery form and harden the token",
"weaknessRationale": [
{
"cweId": "CWE-362",
"rationale": "The primary defect is a TOCTOU race: the token nonce is verified and then consumed in separate operations, allowing two concurrent transactions to both pass verification before either commits. The fix replaces this with an atomic compare-and-set UPDATE, confirming the race condition as the root cause."
},
{
"cweId": "CWE-20",
"rationale": "The confirm_account view never called form.validate() on POST, so the form\u0027s minimum-length and equality validators were bypassed. A valid recovery link alone permitted setting an empty or three-character password. The fix adds the form.validate() call before processing."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "gcve-1-2026-20252"
}
]
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-101041",
"datePublished": "2026-09-27T14:47:33.302690Z",
"dateReserved": "2026-09-27T14:47:50.804Z",
"dateUpdated": "2026-09-27T14:47:50.875076Z",
"state": "PUBLISHED",
"vulnId": "gcve-1-2026-20252"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20183 (CVE-2026-100177)
Vulnerability from gna-1 – Published: 2026-09-25 13:35 – Updated: 2026-09-25 13:55| URL | Tags |
|---|---|
| https://github.com/ail-project/ail-framework/comm… | patch |
| Vendor | Product | Version | |
|---|---|---|---|
| ail project | ail framework |
Affected:
unspecified , < 7.1
(semver)
|
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-25 13:30 - Model
qwen3.8:27b- Input
-
https://github.com/ail-project/ail-framework/commit/3773ca36658c57ce592aebe74e27c855eb64f58a.patch
43e163a1b90d… - Confidence
- medium
| Commit | Subject | Patch SHA-256 |
|---|---|---|
3773ca36658c
|
chg: [security] crawler, enforce cookiejar ACL when creating | 43e163a1b90d… |
Fix summary
The cookiejar attachment path in api_add_crawler_task is now validated through the existing api_check_cookiejar_access_acl() function, which enforces organization, user, and role-based access control for the 'view' action on the referenced cookiejar. The user_role parameter is propagated from the API and UI entry points (api_rest.py, crawler_splash.py) into the ACL check. Separately, URL scheme validation is enforced unconditionally: only http and https schemes are accepted, and non-conforming URLs are rejected with a 400 error before any further processing.
Patch summary
In bin/lib/crawlers.py: (1) is_global_url() now returns False for any scheme other than http/https (previously returned True for data/file schemes) and handles URLs lacking a netloc by prepending http://. (2) api_validate_global_urls() now rejects non-HTTP/HTTPS schemes with a 400 error before the local-IP filter check. (3) The Cookiejar class gains a get_creator() alias. (4) The CrawlerTask.create() method validates the URL scheme. (5) api_add_crawler_task() replaces the inline level-0 user-ID comparison with a call to api_check_cookiejar_access_acl(cookiejar_uuid, user_org, user_id, user_role, action='view') and accepts a new user_role parameter. (6) api_add_crawler_capture() now calls api_validate_global_urls(). In var/www/blueprints/api_rest.py and crawler_splash.py: user_role is extracted from the authenticated session and passed to api_add_crawler_task().
CVSS rationale
The vulnerability is exploitable over the network (AV:N) with low complexity (AC:L) since the attacker only needs to substitute a known or guessed cookiejar UUID in an API request. No attack target manipulation is required (AT:N). An authenticated user with crawler task creation privileges is needed (PR:L). No user interaction is required (UI:N). The vulnerable component itself does not lose confidentiality, integrity, or availability (VC:N, VI:N, VA:N). However, the subsequent impact is high on confidentiality (SC:H) because the attacker gains access to another organization's stored cookies and session tokens via the crawler, which could be used to impersonate the victim on target sites. No subsequent integrity or availability impact is evident (SI:N, SA:N).
Weakness rationale
- CWE-639 The cookiejar UUID is a user-supplied identifier (key) that was not properly validated against the requesting user's organization and role. The old code only compared user IDs when level was 0 and skipped the check entirely otherwise, allowing cross-organization access to another user's cookiejar.
- CWE-20 The URL scheme validation was incorrect: data: and file: schemes were treated as valid global URLs, and the scheme check was skipped when the local-IP filter was disabled. The fix enforces that only http and https schemes are accepted unconditionally.
Attack pattern rationale
- CAPEC-1 The attacker tampers with the cookiejar UUID parameter in the crawler task creation API request to reference a resource (cookiejar) belonging to another organization. The system failed to validate that the referenced object was accessible to the requesting user's organization. This is the closest CAPEC pattern to an IDOR-style parameter substitution; no CAPEC specifically named 'Insecure Direct Object Reference' exists, so CAPEC-1 (modifying a request parameter to access unauthorized resources) is the best available match.
Assumptions to verify
- The api_check_cookiejar_access_acl() function is assumed to correctly enforce organization, user, and role-based access control; its implementation is not included in this patch.
- The cookiejar UUID is assumed to be guessable or discoverable by an authenticated user (e.g., via enumeration or information leakage); if UUIDs are truly random and unguessable, exploitability is reduced but the authorization flaw remains.
- The CVSS assumes the primary security impact is subsequent confidentiality (access to another org's cookies/sessions) rather than direct compromise of the Ail Framework itself.
- The CAPEC-1 mapping is the closest available pattern for an IDOR/parameter-substitution attack; no CAPEC specifically covers insecure direct object reference, so this mapping carries some uncertainty.
- The affected version boundary is inferred from the tag v7.1 (26 commits after the fix commit), suggesting the fix landed before the v7.1 release; exact affected version range is not confirmed.
- The URL scheme validation fix (rejecting data:/file: schemes) is treated as a secondary hardening change within the same patch set rather than a separate vulnerability.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
4 | 9 | medium | 6 |
{
"containers": {
"cna": {
"affected": [
{
"defaultStatus": "unaffected",
"modules": [
"bin/lib/crawlers.py",
"var/www/blueprints/api_rest.py",
"var/www/blueprints/crawler_splash.py"
],
"product": "ail framework",
"programFiles": [
"bin/lib/crawlers.py",
"var/www/blueprints/api_rest.py",
"var/www/blueprints/crawler_splash.py"
],
"repo": "https://github.com/ail-project/ail-framework",
"vendor": "ail project",
"versions": [
{
"lessThan": "7.1",
"status": "affected",
"version": "unspecified",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "terrtia"
},
{
"lang": "en",
"type": "finder",
"value": "Jeroen Pinoy"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe AIL Framework crawler task creation API (api_add_crawler_task) contained an insufficient authorization check when a user supplied a cookiejar UUID to attach to a one-shot or scheduled crawler task. The original code only verified that the cookiejar existed and, if its access level was 0, compared the cookiejar\u0027s owning user ID to the requesting user ID. It did not validate organizational boundaries, did not account for the requesting user\u0027s role. When the cookiejar level was not 0, no access check was performed at all. An authenticated user could therefore reference another organization\u0027s cookiejar by UUID and have the crawler use that organization\u0027s stored cookies (session tokens, authentication credentials) when performing web crawls, effectively leaking or exfiltrating the victim organization\u0027s session data.\u003cbr\u003e\u003c/p\u003e\u003cp\u003eThe vulnerability requires an authenticated user with the ability to create crawler tasks. The attacker must know or guess a valid cookiejar UUID belonging to another organization. The impact is unauthorized access to another organization\u0027s stored cookies and session data through the crawler infrastructure.\u003c/p\u003e"
}
],
"value": "The AIL Framework crawler task creation API (api_add_crawler_task) contained an insufficient authorization check when a user supplied a cookiejar UUID to attach to a one-shot or scheduled crawler task. The original code only verified that the cookiejar existed and, if its access level was 0, compared the cookiejar\u0027s owning user ID to the requesting user ID. It did not validate organizational boundaries, did not account for the requesting user\u0027s role. When the cookiejar level was not 0, no access check was performed at all. An authenticated user could therefore reference another organization\u0027s cookiejar by UUID and have the crawler use that organization\u0027s stored cookies (session tokens, authentication credentials) when performing web crawls, effectively leaking or exfiltrating the victim organization\u0027s session data.\n\n\nThe vulnerability requires an authenticated user with the ability to create crawler tasks. The attacker must know or guess a valid cookiejar UUID belonging to another organization. The impact is unauthorized access to another organization\u0027s stored cookies and session data through the crawler infrastructure."
}
],
"impacts": [
{
"capecId": "CAPEC-1",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-1 Tamper with URLs"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 6.3,
"baseSeverity": "MEDIUM",
"privilegesRequired": "LOW",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "HIGH",
"subIntegrityImpact": "NONE",
"userInteraction": "NONE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:N/SC:H/SI:N/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "NONE",
"vulnIntegrityImpact": "NONE",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-639",
"description": "CWE-639 Authorization Bypass Through User-Controlled Key",
"lang": "en",
"type": "CWE"
}
]
},
{
"descriptions": [
{
"cweId": "CWE-20",
"description": "CWE-20 Improper Input Validation",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/ail-project/ail-framework/commit/3773ca36658c57ce592aebe74e27c855eb64f58a"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe cookiejar attachment path in api_add_crawler_task is now validated through the existing api_check_cookiejar_access_acl() function, which enforces organization, user, and role-based access control for the \u0027view\u0027 action on the referenced cookiejar. The user_role parameter is propagated from the API and UI entry points (api_rest.py, crawler_splash.py) into the ACL check. Separately, URL scheme validation is enforced unconditionally: only http and https schemes are accepted, and non-conforming URLs are rejected with a 400 error before any further processing.\u003c/p\u003e"
}
],
"value": "The cookiejar attachment path in api_add_crawler_task is now validated through the existing api_check_cookiejar_access_acl() function, which enforces organization, user, and role-based access control for the \u0027view\u0027 action on the referenced cookiejar. The user_role parameter is propagated from the API and UI entry points (api_rest.py, crawler_splash.py) into the ACL check. Separately, URL scheme validation is enforced unconditionally: only http and https schemes are accepted, and non-conforming URLs are rejected with a 400 error before any further processing."
}
],
"source": {
"discovery": "UNKNOWN"
},
"title": "Ail Framework Crawler: Missing Cookiejar Authorization Check Allows Cross-Organization Cookiejar Attachment",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The api_check_cookiejar_access_acl() function is assumed to correctly enforce organization, user, and role-based access control; its implementation is not included in this patch.",
"The cookiejar UUID is assumed to be guessable or discoverable by an authenticated user (e.g., via enumeration or information leakage); if UUIDs are truly random and unguessable, exploitability is reduced but the authorization flaw remains.",
"The CVSS assumes the primary security impact is subsequent confidentiality (access to another org\u0027s cookies/sessions) rather than direct compromise of the Ail Framework itself.",
"The CAPEC-1 mapping is the closest available pattern for an IDOR/parameter-substitution attack; no CAPEC specifically covers insecure direct object reference, so this mapping carries some uncertainty.",
"The affected version boundary is inferred from the tag v7.1 (26 commits after the fix commit), suggesting the fix landed before the v7.1 release; exact affected version range is not confirmed.",
"The URL scheme validation fix (rejecting data:/file: schemes) is treated as a secondary hardening change within the same patch set rather than a separate vulnerability."
],
"capecRationale": [
{
"capecId": "CAPEC-1",
"rationale": "The attacker tampers with the cookiejar UUID parameter in the crawler task creation API request to reference a resource (cookiejar) belonging to another organization. The system failed to validate that the referenced object was accessible to the requesting user\u0027s organization. This is the closest CAPEC pattern to an IDOR-style parameter substitution; no CAPEC specifically named \u0027Insecure Direct Object Reference\u0027 exists, so CAPEC-1 (modifying a request parameter to access unauthorized resources) is the best available match."
}
],
"commit": "3773ca36658c57ce592aebe74e27c855eb64f58a",
"confidence": "medium",
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "terrtia"
}
],
"cvssRationale": "The vulnerability is exploitable over the network (AV:N) with low complexity (AC:L) since the attacker only needs to substitute a known or guessed cookiejar UUID in an API request. No attack target manipulation is required (AT:N). An authenticated user with crawler task creation privileges is needed (PR:L). No user interaction is required (UI:N). The vulnerable component itself does not lose confidentiality, integrity, or availability (VC:N, VI:N, VA:N). However, the subsequent impact is high on confidentiality (SC:H) because the attacker gains access to another organization\u0027s stored cookies and session tokens via the crawler, which could be used to impersonate the victim on target sites. No subsequent integrity or availability impact is evident (SI:N, SA:N).",
"fixSummary": "The cookiejar attachment path in api_add_crawler_task is now validated through the existing api_check_cookiejar_access_acl() function, which enforces organization, user, and role-based access control for the \u0027view\u0027 action on the referenced cookiejar. The user_role parameter is propagated from the API and UI entry points (api_rest.py, crawler_splash.py) into the ACL check. Separately, URL scheme validation is enforced unconditionally: only http and https schemes are accepted, and non-conforming URLs are rejected with a 400 error before any further processing.",
"generatedAt": "2026-09-25T13:30:48.386437Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 6,
"confidence": "medium",
"model": "qwen3.8:27b",
"score": 4
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "43e163a1b90d5926293d5dbd7982d46241f08d7a9d2cb0edf86c963f794afac5",
"patchSummary": "In bin/lib/crawlers.py: (1) is_global_url() now returns False for any scheme other than http/https (previously returned True for data/file schemes) and handles URLs lacking a netloc by prepending http://. (2) api_validate_global_urls() now rejects non-HTTP/HTTPS schemes with a 400 error before the local-IP filter check. (3) The Cookiejar class gains a get_creator() alias. (4) The CrawlerTask.create() method validates the URL scheme. (5) api_add_crawler_task() replaces the inline level-0 user-ID comparison with a call to api_check_cookiejar_access_acl(cookiejar_uuid, user_org, user_id, user_role, action=\u0027view\u0027) and accepts a new user_role parameter. (6) api_add_crawler_capture() now calls api_validate_global_urls(). In var/www/blueprints/api_rest.py and crawler_splash.py: user_role is extracted from the authenticated session and passed to api_add_crawler_task().",
"patchTruncated": false,
"patches": [
{
"commit": "3773ca36658c57ce592aebe74e27c855eb64f58a",
"patchSha256": "43e163a1b90d5926293d5dbd7982d46241f08d7a9d2cb0edf86c963f794afac5",
"source": "https://github.com/ail-project/ail-framework/commit/3773ca36658c57ce592aebe74e27c855eb64f58a.patch",
"sourceUrl": "https://github.com/ail-project/ail-framework/commit/3773ca36658c57ce592aebe74e27c855eb64f58a.patch",
"subject": "chg: [security] crawler, enforce cookiejar ACL when creating"
}
],
"source": "https://github.com/ail-project/ail-framework/commit/3773ca36658c57ce592aebe74e27c855eb64f58a.patch",
"subject": "chg: [security] crawler, enforce cookiejar ACL when creating",
"tagVersionBoundary": {
"commits_after_fix": 26,
"repository": "https://github.com/ail-project/ail-framework",
"tag": "v7.1",
"version": "7.1",
"version_type": "custom"
},
"weaknessRationale": [
{
"cweId": "CWE-639",
"rationale": "The cookiejar UUID is a user-supplied identifier (key) that was not properly validated against the requesting user\u0027s organization and role. The old code only compared user IDs when level was 0 and skipped the check entirely otherwise, allowing cross-organization access to another user\u0027s cookiejar."
},
{
"cweId": "CWE-20",
"rationale": "The URL scheme validation was incorrect: data: and file: schemes were treated as valid global URLs, and the scheme check was skipped when the local-IP filter was disabled. The fix enforces that only http and https schemes are accepted unconditionally."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20183"
}
],
"x_generator": {
"engine": "Vulnogram 0.2.0"
}
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-100177",
"datePublished": "2026-09-25T13:35:00.000Z",
"dateReserved": "2026-09-25T13:35:00.000Z",
"dateUpdated": "2026-09-25T13:55:39.228706Z",
"requesterUserId": "00000000-0000-4000-9000-000000000000",
"serial": 1,
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20183"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20279 (CVE-2026-100176)
Vulnerability from gna-1 – Published: 2026-09-25 13:29 – Updated: 2026-09-25 13:55- CWE-79 - Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting')
| URL | Tags |
|---|---|
| https://github.com/ail-project/ail-framework/comm… | patch |
| Vendor | Product | Version | |
|---|---|---|---|
| ail project | ail framework |
Affected:
unspecified , < 7.1
(semver)
|
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-25 13:24 - Model
qwen3.8:27b- Input
-
https://github.com/ail-project/ail-framework/commit/455dd92c4179c53204368d93681ded4b57e31643.patch
36262636a3a7… - Confidence
- medium
| Commit | Subject | Patch SHA-256 |
|---|---|---|
455dd92c4179
|
chg: [security] fix XSS in username timeline tooltip. | 36262636a3a7… |
Fix summary
The fix applies the existing sanitize_text() helper to the user-controlled username value (d.obj) immediately before it is interpolated into the D3 tooltip HTML string. This ensures that any HTML or script content embedded in the stored username is neutralized before DOM insertion, preventing script execution while preserving the tooltip's intended formatting (line breaks and date/time display).
Patch summary
In var/www/static/js/d3/timeline_basic.js, the single line constructing the tooltip HTML was changed from tooltip.html(`${d.obj}<br>...`) to tooltip.html(`${sanitize_text(d.obj)}<br>...`). The only modification is wrapping the untrusted d.obj variable with the sanitize_text() function call at the HTML sink, leaving the rest of the tooltip template (br tags, date formatting) unchanged.
CVSS rationale
AV:N: The vulnerability is in a web application accessible over the network. AC:L: The attacker only needs to craft a username with HTML/JS in a chat or forum that gets crawled; no race conditions or complex conditions are needed. AT:N: No attack on the target system is required; the malicious username is stored as part of normal data import. PR:N: The attacker does not need any account or privilege on the AIL Framework; they only need to be a user in a source chat/forum. UI:A: The victim (analyst) must actively hover over the specific timeline entry to trigger script execution. VC:H: XSS in an authenticated analyst session can expose sensitive intelligence data, session tokens, and credentials. VI:H: The attacker can perform actions as the analyst, modify data, or exfiltrate information. VA:N: No impact on availability. SC/SI/SA:N: No subsequent system impact is evidenced.
Weakness rationale
- CWE-79 The patch inserts a user-controlled value (username from imported chats/forums) into an HTML context via D3's html() method without sanitization. The fix adds sanitize_text() at the sink, confirming the root cause is missing output encoding/neutralization of untrusted data in an HTML context, which is the definition of CWE-79 (stored XSS variant).
Attack pattern rationale
- CAPEC-1 CAPEC-1 describes the general pattern of injecting client-side scripts into web pages via untrusted input. The patch evidence shows a stored username (untrusted input from chats/forums) being rendered into an HTML tooltip without sanitization, matching the stored-XSS sub-pattern of CAPEC-1. This is the closest and most direct CAPEC mapping for an XSS vulnerability where attacker-controlled data is persisted and later rendered in a victim's browser.
Assumptions to verify
- The UI:A (Active) user interaction metric assumes the analyst must deliberately hover over the specific timeline entry to trigger the XSS; passive rendering without hover is not evidenced in the patch.
- The affected version boundary is inferred from the tag_version_boundary metadata (v7.1, 29 commits after fix); the exact last affected and first fixed version numbers are not explicitly stated in the patch.
- PR:N assumes the attacker does not need an account on the AIL Framework itself, only a presence in a crawled chat or forum source; if the platform requires the attacker to be a registered user to have their username crawled, PR:L might be more appropriate.
- The CAPEC-1 mapping is the closest available pattern for stored XSS; no more specific CAPEC ID for 'stored XSS via imported data rendered in a tooltip' exists in the CAPEC catalog.
- The sanitize_text() helper is assumed to perform adequate HTML entity encoding or equivalent neutralization; its implementation is not shown in the patch.
- VC:H and VI:H reflect the sensitivity of an intelligence-analysis platform where an authenticated analyst session may contain classified or highly sensitive data; if the deployment context is less sensitive, lower impact ratings could apply.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
4 | 9 | medium | 6 |
{
"containers": {
"cna": {
"affected": [
{
"defaultStatus": "unaffected",
"modules": [
"username timeline (D3 timeline_basic)"
],
"product": "ail framework",
"programFiles": [
"var/www/static/js/d3/timeline_basic.js"
],
"repo": "https://github.com/ail-project/ail-framework",
"vendor": "ail project",
"versions": [
{
"lessThan": "7.1",
"status": "affected",
"version": "unspecified",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "Aurelien Thirion"
},
{
"lang": "en",
"type": "finder",
"value": "Jeroen Pinoy"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe AIL Framework\u0027s username timeline feature is vulnerable to stored cross-site scripting (XSS). Usernames imported from chats and crawled forums are stored without character restrictions. When an authenticated analyst views the username timeline, the application renders these stored usernames into the DOM using D3\u0027s html() method in the tooltip. Because the username value (d.obj) is interpolated directly into an HTML string without sanitization, a crafted username containing HTML event handlers (e.g., \u0026lt;img src=x onerror=alert(1)\u0026gt;) will execute arbitrary JavaScript in the analyst\u0027s browser when the analyst hovers over the corresponding timeline entry. The attack requires the victim to be an authenticated analyst with access to the timeline view and to interact with the malicious timeline entry (hover). Successful exploitation can lead to session hijacking, data exfiltration, or unauthorized actions performed within the analyst\u0027s authenticated session. The vulnerability resides in the client-side JavaScript file var/www/static/js/d3/timeline_basic.js.\u003c/p\u003e"
}
],
"value": "The AIL Framework\u0027s username timeline feature is vulnerable to stored cross-site scripting (XSS). Usernames imported from chats and crawled forums are stored without character restrictions. When an authenticated analyst views the username timeline, the application renders these stored usernames into the DOM using D3\u0027s html() method in the tooltip. Because the username value (d.obj) is interpolated directly into an HTML string without sanitization, a crafted username containing HTML event handlers (e.g., \u003cimg src=x onerror=alert(1)\u003e) will execute arbitrary JavaScript in the analyst\u0027s browser when the analyst hovers over the corresponding timeline entry. The attack requires the victim to be an authenticated analyst with access to the timeline view and to interact with the malicious timeline entry (hover). Successful exploitation can lead to session hijacking, data exfiltration, or unauthorized actions performed within the analyst\u0027s authenticated session. The vulnerability resides in the client-side JavaScript file var/www/static/js/d3/timeline_basic.js."
}
],
"impacts": [
{
"capecId": "CAPEC-1",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-1 Cross Site Scripting"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 8.5,
"baseSeverity": "HIGH",
"privilegesRequired": "NONE",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "NONE",
"subIntegrityImpact": "NONE",
"userInteraction": "ACTIVE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:A/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "HIGH",
"vulnIntegrityImpact": "HIGH",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-79",
"description": "CWE-79 Improper Neutralization of Input During Web Page Generation (\u0027Cross-site Scripting\u0027)",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/ail-project/ail-framework/commit/455dd92c4179c53204368d93681ded4b57e31643"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe fix applies the existing sanitize_text() helper to the user-controlled username value (d.obj) immediately before it is interpolated into the D3 tooltip HTML string. This ensures that any HTML or script content embedded in the stored username is neutralized before DOM insertion, preventing script execution while preserving the tooltip\u0027s intended formatting (line breaks and date/time display).\u003c/p\u003e"
}
],
"value": "The fix applies the existing sanitize_text() helper to the user-controlled username value (d.obj) immediately before it is interpolated into the D3 tooltip HTML string. This ensures that any HTML or script content embedded in the stored username is neutralized before DOM insertion, preventing script execution while preserving the tooltip\u0027s intended formatting (line breaks and date/time display)."
}
],
"source": {
"discovery": "UNKNOWN"
},
"title": "Stored Cross-Site Scripting (XSS) in AIL Framework Username Timeline Tooltip",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The UI:A (Active) user interaction metric assumes the analyst must deliberately hover over the specific timeline entry to trigger the XSS; passive rendering without hover is not evidenced in the patch.",
"The affected version boundary is inferred from the tag_version_boundary metadata (v7.1, 29 commits after fix); the exact last affected and first fixed version numbers are not explicitly stated in the patch.",
"PR:N assumes the attacker does not need an account on the AIL Framework itself, only a presence in a crawled chat or forum source; if the platform requires the attacker to be a registered user to have their username crawled, PR:L might be more appropriate.",
"The CAPEC-1 mapping is the closest available pattern for stored XSS; no more specific CAPEC ID for \u0027stored XSS via imported data rendered in a tooltip\u0027 exists in the CAPEC catalog.",
"The sanitize_text() helper is assumed to perform adequate HTML entity encoding or equivalent neutralization; its implementation is not shown in the patch.",
"VC:H and VI:H reflect the sensitivity of an intelligence-analysis platform where an authenticated analyst session may contain classified or highly sensitive data; if the deployment context is less sensitive, lower impact ratings could apply."
],
"capecRationale": [
{
"capecId": "CAPEC-1",
"rationale": "CAPEC-1 describes the general pattern of injecting client-side scripts into web pages via untrusted input. The patch evidence shows a stored username (untrusted input from chats/forums) being rendered into an HTML tooltip without sanitization, matching the stored-XSS sub-pattern of CAPEC-1. This is the closest and most direct CAPEC mapping for an XSS vulnerability where attacker-controlled data is persisted and later rendered in a victim\u0027s browser."
}
],
"commit": "455dd92c4179c53204368d93681ded4b57e31643",
"confidence": "medium",
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "terrtia"
}
],
"cvssRationale": "AV:N: The vulnerability is in a web application accessible over the network. AC:L: The attacker only needs to craft a username with HTML/JS in a chat or forum that gets crawled; no race conditions or complex conditions are needed. AT:N: No attack on the target system is required; the malicious username is stored as part of normal data import. PR:N: The attacker does not need any account or privilege on the AIL Framework; they only need to be a user in a source chat/forum. UI:A: The victim (analyst) must actively hover over the specific timeline entry to trigger script execution. VC:H: XSS in an authenticated analyst session can expose sensitive intelligence data, session tokens, and credentials. VI:H: The attacker can perform actions as the analyst, modify data, or exfiltrate information. VA:N: No impact on availability. SC/SI/SA:N: No subsequent system impact is evidenced.",
"fixSummary": "The fix applies the existing sanitize_text() helper to the user-controlled username value (d.obj) immediately before it is interpolated into the D3 tooltip HTML string. This ensures that any HTML or script content embedded in the stored username is neutralized before DOM insertion, preventing script execution while preserving the tooltip\u0027s intended formatting (line breaks and date/time display).",
"generatedAt": "2026-09-25T13:24:29.029450Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 6,
"confidence": "medium",
"model": "qwen3.8:27b",
"score": 4
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "36262636a3a75bbee1ac4f863ce7a5ce7939eb2e21fed5ede98403d38c29ec40",
"patchSummary": "In var/www/static/js/d3/timeline_basic.js, the single line constructing the tooltip HTML was changed from tooltip.html(`${d.obj}\u003cbr\u003e...`) to tooltip.html(`${sanitize_text(d.obj)}\u003cbr\u003e...`). The only modification is wrapping the untrusted d.obj variable with the sanitize_text() function call at the HTML sink, leaving the rest of the tooltip template (br tags, date formatting) unchanged.",
"patchTruncated": false,
"patches": [
{
"commit": "455dd92c4179c53204368d93681ded4b57e31643",
"patchSha256": "36262636a3a75bbee1ac4f863ce7a5ce7939eb2e21fed5ede98403d38c29ec40",
"source": "https://github.com/ail-project/ail-framework/commit/455dd92c4179c53204368d93681ded4b57e31643.patch",
"sourceUrl": "https://github.com/ail-project/ail-framework/commit/455dd92c4179c53204368d93681ded4b57e31643.patch",
"subject": "chg: [security] fix XSS in username timeline tooltip."
}
],
"source": "https://github.com/ail-project/ail-framework/commit/455dd92c4179c53204368d93681ded4b57e31643.patch",
"subject": "chg: [security] fix XSS in username timeline tooltip.",
"tagVersionBoundary": {
"commits_after_fix": 29,
"repository": "https://github.com/ail-project/ail-framework",
"tag": "v7.1",
"version": "7.1",
"version_type": "custom"
},
"weaknessRationale": [
{
"cweId": "CWE-79",
"rationale": "The patch inserts a user-controlled value (username from imported chats/forums) into an HTML context via D3\u0027s html() method without sanitization. The fix adds sanitize_text() at the sink, confirming the root cause is missing output encoding/neutralization of untrusted data in an HTML context, which is the definition of CWE-79 (stored XSS variant)."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20279"
}
],
"x_generator": {
"engine": "Vulnogram 0.2.0"
}
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-100176",
"datePublished": "2026-09-25T13:29:00.000Z",
"dateReserved": "2026-09-25T13:29:00.000Z",
"dateUpdated": "2026-09-25T13:55:11.866831Z",
"requesterUserId": "00000000-0000-4000-9000-000000000000",
"serial": 1,
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20279"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20269 (CVE-2026-100174)
Vulnerability from gna-1 – Published: 2026-09-25 13:21 – Updated: 2026-09-25 13:54| URL | Tags |
|---|---|
| https://github.com/ail-project/ail-framework/comm… | patch |
| Vendor | Product | Version | |
|---|---|---|---|
| ail project | ail framework |
Affected:
0 , < 7.1
(semver)
|
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-25 13:16 - Model
qwen3.8:27b- Input
-
https://github.com/ail-project/ail-framework/commit/b2d0ed05f6cf79b2312a3f1f923d0d1a5d7a9edb.patch
2d147dabcc9b… - Confidence
- medium
| Commit | Subject | Patch SHA-256 |
|---|---|---|
b2d0ed05f6cf
|
chg: [security] prevent stored XSS in | 2d147dabcc9b… |
Fix summary
The fix ensures that tag names and display values are rendered as plain text rather than as HTML when no custom renderer is explicitly configured. In the suggestion rendering path, the default (non-renderer) case now uses jQuery's .text() method, which HTML-escapes the content. In the selection rendering path, the default case uses document.createTextNode() to insert the value as a text node. Custom renderers, which are developer-supplied and trusted to produce safe HTML, continue to use .html() or .prepend() with raw HTML. The eval() call was also replaced with JSON.parse() to eliminate a code-injection vector.
Patch summary
In var/www/static/js/tags.js: (1) Replaced eval(data) with JSON.parse(data) for parsing array-like data strings. (2) In _renderComboItems, introduced a hasRenderer flag; when no custom renderer is set, the displayed value is inserted via resultItemEl.text(displayed) instead of the html property in the jQuery constructor. (3) In the selection rendering block, introduced a hasSelectionRenderer flag; when no custom selection renderer is set, the value is prepended as a text node via document.createTextNode(selectedItemHtml) instead of being set through the html property. (4) The comma separator for text-mode selections is now appended as a text node rather than concatenated into an HTML string.
CVSS rationale
AV:N: The vulnerability is exploitable over the network via a web application. AC:L: No race conditions or special timing are required; the payload is stored and triggers on page load. AT:N: No manipulation of the attack target is needed; the stored payload executes automatically. PR:L: The attacker needs a low-privilege authenticated account with permission to create custom tags. UI:P: The victim passively interacts by loading a page that contains the tag selector; no active click or download is required. VC:L: The XSS can read data from the victim's session and page context. VI:L: The XSS can modify page content and submit forms. VA:N: No availability impact is evident. SC:L: The compromised session can access other resources within the same origin. SI:L: The attacker can modify data through the victim's authenticated session. SA:N: No availability impact on subsequent components.
Weakness rationale
- CWE-79 The patch explicitly addresses stored XSS: user-controlled tag names were inserted into the DOM as HTML via jQuery's html property without escaping. The fix renders these values as text nodes, which is the canonical remediation for CWE-79. The commit subject line also explicitly states 'prevent stored XSS'.
- CWE-94 The replacement of eval(data) with JSON.parse(data) addresses a secondary code-injection weakness where a crafted data string could execute arbitrary JavaScript. This is a contributing factor but the primary vulnerability is the stored XSS in tag rendering.
Attack pattern rationale
- CAPEC-66 CAPEC-66 describes the general XSS attack pattern where attacker-controlled input is reflected or stored and later rendered as executable content in a victim's browser. This patch fixes a stored XSS variant where malicious tag names are persisted and rendered as HTML. The mapping is direct and well-supported by the commit message and code changes. No more specific CAPEC entry for stored XSS in a tag/selector widget exists in the CAPEC catalog, making CAPEC-66 the closest and most appropriate match.
Assumptions to verify
- The affected version range is inferred from the tag_version_boundary metadata indicating the fix commit is 31 commits before the v7.1 tag; exact version boundaries are not explicitly stated in the patch.
- The UI:P (Passive) rating assumes the victim only needs to load a page containing the tag selector; if the application requires the victim to explicitly open a specific dialog or widget, UI:A (Active) could be more appropriate.
- The CAPEC-66 mapping is the closest available pattern; no CAPEC entry specifically covers stored XSS in a tag-selector UI component, so the general XSS pattern is used.
- The PR:L rating assumes that creating a custom tag requires only a basic authenticated account; if tag creation requires elevated privileges, PR:H would be more appropriate.
- The eval() to JSON.parse() change is treated as a secondary hardening fix within the same patch set; the primary vulnerability addressed is the stored XSS in tag rendering.
- No specific reporter or finder is identified in the patch metadata or commit message; only the remediation developer (commit author) is credited.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
4 | 9 | medium | 6 |
{
"containers": {
"cna": {
"affected": [
{
"defaultStatus": "unaffected",
"modules": [
"tag selector (tags.js)"
],
"product": "ail framework",
"programFiles": [
"var/www/static/js/tags.js"
],
"repo": "https://github.com/ail-project/ail-framework",
"vendor": "ail project",
"versions": [
{
"lessThan": "7.1",
"status": "affected",
"version": "0",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "Aurelien Thirion"
},
{
"lang": "en",
"type": "finder",
"value": "Jeroen Pinoy"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe AIL Framework tag selector component (var/www/static/js/tags.js) is vulnerable to stored cross-site scripting (XSS). A user with the ability to create a custom tag could embed an HTML payload containing JavaScript event handlers (e.g., \u0026lt;img src=x onerror=alert(1)\u0026gt; or \u0026lt;svg onload=...\u0026gt;) in the tag name. When another authenticated user opened a page containing the tag selector, the malicious tag name was inserted into the DOM using jQuery\u0027s html property, causing the embedded script to execute in the victim\u0027s browser context.\u003c/p\u003e\u003cp\u003eThe vulnerable code paths affected both the suggestion/combo-item rendering (_renderComboItems) and the selected-tag rendering logic. In both cases, the display value derived from the tag\u0027s displayField was passed directly to the html property of a jQuery element constructor, which parses and inserts the string as raw HTML rather than as text.\u003c/p\u003e\u003cp\u003ePreconditions: the attacker must have an authenticated account with permission to create custom tags, and the victim must be an authenticated user who views a page that renders the tag selector with the attacker\u0027s stored tag. The attack is stored (persistent) and does not require the victim to perform any action beyond loading the page.\u003c/p\u003e\u003cp\u003eSecurity impact: successful exploitation allows arbitrary JavaScript execution in the victim\u0027s browser within the application\u0027s origin, potentially leading to session hijacking, unauthorized data access, form manipulation, or further client-side attacks.\u003c/p\u003e"
}
],
"value": "The AIL Framework tag selector component (var/www/static/js/tags.js) is vulnerable to stored cross-site scripting (XSS). A user with the ability to create a custom tag could embed an HTML payload containing JavaScript event handlers (e.g., \u003cimg src=x onerror=alert(1)\u003e or \u003csvg onload=...\u003e) in the tag name. When another authenticated user opened a page containing the tag selector, the malicious tag name was inserted into the DOM using jQuery\u0027s html property, causing the embedded script to execute in the victim\u0027s browser context.\n\nThe vulnerable code paths affected both the suggestion/combo-item rendering (_renderComboItems) and the selected-tag rendering logic. In both cases, the display value derived from the tag\u0027s displayField was passed directly to the html property of a jQuery element constructor, which parses and inserts the string as raw HTML rather than as text.\n\nPreconditions: the attacker must have an authenticated account with permission to create custom tags, and the victim must be an authenticated user who views a page that renders the tag selector with the attacker\u0027s stored tag. The attack is stored (persistent) and does not require the victim to perform any action beyond loading the page.\n\nSecurity impact: successful exploitation allows arbitrary JavaScript execution in the victim\u0027s browser within the application\u0027s origin, potentially leading to session hijacking, unauthorized data access, form manipulation, or further client-side attacks."
}
],
"impacts": [
{
"capecId": "CAPEC-66",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-66 Cross Site Scripting (XSS)"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 5.1,
"baseSeverity": "MEDIUM",
"privilegesRequired": "LOW",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "LOW",
"subIntegrityImpact": "LOW",
"userInteraction": "PASSIVE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:L/VI:L/VA:N/SC:L/SI:L/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "LOW",
"vulnIntegrityImpact": "LOW",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-79",
"description": "CWE-79 Improper Neutralization of Input During Web Page Generation (\u0027Cross-site Scripting\u0027)",
"lang": "en",
"type": "CWE"
}
]
},
{
"descriptions": [
{
"cweId": "CWE-94",
"description": "CWE-94 Improper Control of Generation of Code (\u0027Code Injection\u0027)",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/ail-project/ail-framework/commit/b2d0ed05f6cf79b2312a3f1f923d0d1a5d7a9edb"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe fix ensures that tag names and display values are rendered as plain text rather than as HTML when no custom renderer is explicitly configured. In the suggestion rendering path, the default (non-renderer) case now uses jQuery\u0027s .text() method, which HTML-escapes the content. In the selection rendering path, the default case uses document.createTextNode() to insert the value as a text node. Custom renderers, which are developer-supplied and trusted to produce safe HTML, continue to use .html() or .prepend() with raw HTML. The eval() call was also replaced with JSON.parse() to eliminate a code-injection vector.\u003c/p\u003e"
}
],
"value": "The fix ensures that tag names and display values are rendered as plain text rather than as HTML when no custom renderer is explicitly configured. In the suggestion rendering path, the default (non-renderer) case now uses jQuery\u0027s .text() method, which HTML-escapes the content. In the selection rendering path, the default case uses document.createTextNode() to insert the value as a text node. Custom renderers, which are developer-supplied and trusted to produce safe HTML, continue to use .html() or .prepend() with raw HTML. The eval() call was also replaced with JSON.parse() to eliminate a code-injection vector."
}
],
"source": {
"discovery": "UNKNOWN"
},
"title": "Stored Cross-Site Scripting (XSS) in AIL Framework Tag Selector via Unescaped Tag Names",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The affected version range is inferred from the tag_version_boundary metadata indicating the fix commit is 31 commits before the v7.1 tag; exact version boundaries are not explicitly stated in the patch.",
"The UI:P (Passive) rating assumes the victim only needs to load a page containing the tag selector; if the application requires the victim to explicitly open a specific dialog or widget, UI:A (Active) could be more appropriate.",
"The CAPEC-66 mapping is the closest available pattern; no CAPEC entry specifically covers stored XSS in a tag-selector UI component, so the general XSS pattern is used.",
"The PR:L rating assumes that creating a custom tag requires only a basic authenticated account; if tag creation requires elevated privileges, PR:H would be more appropriate.",
"The eval() to JSON.parse() change is treated as a secondary hardening fix within the same patch set; the primary vulnerability addressed is the stored XSS in tag rendering.",
"No specific reporter or finder is identified in the patch metadata or commit message; only the remediation developer (commit author) is credited."
],
"capecRationale": [
{
"capecId": "CAPEC-66",
"rationale": "CAPEC-66 describes the general XSS attack pattern where attacker-controlled input is reflected or stored and later rendered as executable content in a victim\u0027s browser. This patch fixes a stored XSS variant where malicious tag names are persisted and rendered as HTML. The mapping is direct and well-supported by the commit message and code changes. No more specific CAPEC entry for stored XSS in a tag/selector widget exists in the CAPEC catalog, making CAPEC-66 the closest and most appropriate match."
}
],
"commit": "b2d0ed05f6cf79b2312a3f1f923d0d1a5d7a9edb",
"confidence": "medium",
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "terrtia"
}
],
"cvssRationale": "AV:N: The vulnerability is exploitable over the network via a web application. AC:L: No race conditions or special timing are required; the payload is stored and triggers on page load. AT:N: No manipulation of the attack target is needed; the stored payload executes automatically. PR:L: The attacker needs a low-privilege authenticated account with permission to create custom tags. UI:P: The victim passively interacts by loading a page that contains the tag selector; no active click or download is required. VC:L: The XSS can read data from the victim\u0027s session and page context. VI:L: The XSS can modify page content and submit forms. VA:N: No availability impact is evident. SC:L: The compromised session can access other resources within the same origin. SI:L: The attacker can modify data through the victim\u0027s authenticated session. SA:N: No availability impact on subsequent components.",
"fixSummary": "The fix ensures that tag names and display values are rendered as plain text rather than as HTML when no custom renderer is explicitly configured. In the suggestion rendering path, the default (non-renderer) case now uses jQuery\u0027s .text() method, which HTML-escapes the content. In the selection rendering path, the default case uses document.createTextNode() to insert the value as a text node. Custom renderers, which are developer-supplied and trusted to produce safe HTML, continue to use .html() or .prepend() with raw HTML. The eval() call was also replaced with JSON.parse() to eliminate a code-injection vector.",
"generatedAt": "2026-09-25T13:16:06.156442Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 6,
"confidence": "medium",
"model": "qwen3.8:27b",
"score": 4
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "2d147dabcc9bebce7722917d80f13e49a10f9a2a4089ec6207cce9edb9187f77",
"patchSummary": "In var/www/static/js/tags.js: (1) Replaced eval(data) with JSON.parse(data) for parsing array-like data strings. (2) In _renderComboItems, introduced a hasRenderer flag; when no custom renderer is set, the displayed value is inserted via resultItemEl.text(displayed) instead of the html property in the jQuery constructor. (3) In the selection rendering block, introduced a hasSelectionRenderer flag; when no custom selection renderer is set, the value is prepended as a text node via document.createTextNode(selectedItemHtml) instead of being set through the html property. (4) The comma separator for text-mode selections is now appended as a text node rather than concatenated into an HTML string.",
"patchTruncated": false,
"patches": [
{
"commit": "b2d0ed05f6cf79b2312a3f1f923d0d1a5d7a9edb",
"patchSha256": "2d147dabcc9bebce7722917d80f13e49a10f9a2a4089ec6207cce9edb9187f77",
"source": "https://github.com/ail-project/ail-framework/commit/b2d0ed05f6cf79b2312a3f1f923d0d1a5d7a9edb.patch",
"sourceUrl": "https://github.com/ail-project/ail-framework/commit/b2d0ed05f6cf79b2312a3f1f923d0d1a5d7a9edb.patch",
"subject": "chg: [security] prevent stored XSS in"
}
],
"source": "https://github.com/ail-project/ail-framework/commit/b2d0ed05f6cf79b2312a3f1f923d0d1a5d7a9edb.patch",
"subject": "chg: [security] prevent stored XSS in",
"tagVersionBoundary": {
"commits_after_fix": 31,
"repository": "https://github.com/ail-project/ail-framework",
"tag": "v7.1",
"version": "7.1",
"version_type": "custom"
},
"weaknessRationale": [
{
"cweId": "CWE-79",
"rationale": "The patch explicitly addresses stored XSS: user-controlled tag names were inserted into the DOM as HTML via jQuery\u0027s html property without escaping. The fix renders these values as text nodes, which is the canonical remediation for CWE-79. The commit subject line also explicitly states \u0027prevent stored XSS\u0027."
},
{
"cweId": "CWE-94",
"rationale": "The replacement of eval(data) with JSON.parse(data) addresses a secondary code-injection weakness where a crafted data string could execute arbitrary JavaScript. This is a contributing factor but the primary vulnerability is the stored XSS in tag rendering."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20269"
}
],
"x_generator": {
"engine": "Vulnogram 0.2.0"
}
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-100174",
"datePublished": "2026-09-25T13:21:00.000Z",
"dateReserved": "2026-09-25T13:22:00.000Z",
"dateUpdated": "2026-09-25T13:54:48.431524Z",
"requesterUserId": "00000000-0000-4000-9000-000000000000",
"serial": 1,
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20269"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20239 (CVE-2026-100172)
Vulnerability from gna-1 – Published: 2026-09-25 13:13 – Updated: 2026-09-25 13:54- CWE-79 - Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting')
| URL | Tags |
|---|---|
| https://github.com/ail-project/ail-framework/comm… | patch |
| Vendor | Product | Version | |
|---|---|---|---|
| ail project | ail framework |
Affected:
unspecified , < 7.1
(semver)
|
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-25 12:56 - Model
qwen3.8:27b- Input
-
https://github.com/ail-project/ail-framework/commit/ee63a0a96646790255e038f5ca5dd0ab5fe98db2.patch
bd188338f882… - Confidence
- medium
| Commit | Subject | Patch SHA-256 |
|---|---|---|
ee63a0a96646
|
chg: [security] fix XSS in show item and message popover of | bd188338f882… |
Fix summary
The fix applies double HTML-escaping (Jinja2 forceescape|forceescape) to every dynamic value interpolated into the HTML-enabled data-content attribute of the popover elements in both templates. Double-escaping is required because the values must survive both the outer HTML attribute context and the inner HTML parsing performed by the popover plugin. This neutralizes any HTML or script injection from the extracted-match fields.
Patch summary
In block_message.html and show_item.html, the Jinja2 expressions for extracted_matches fields (icon color, icon style, icon, subtype, id, name, description) and the matched value r[1] are each suffixed with |forceescape|forceescape. No other logic, routing, or configuration changes are made. Two lines changed in total (one per file).
CVSS rationale
AV:N: web application reachable over the network. AC:L: crafting a message or item with a malicious extracted-match value requires no special race or timing. AT:N: no manipulation of the attack target is needed. PR:L: the attacker must be an authenticated user capable of creating or modifying chat messages or items. UI:P: the victim must passively interact (hover over the element) to trigger the popover rendering; no click or explicit action is required. VC:H: the injected script can read session cookies, DOM data, and API responses in the victim's browser. VI:H: the script can perform authenticated actions on behalf of the victim (create, modify, delete data). VA:N: no direct denial-of-service impact is evident. SC/SI/SA:N: the subcomponent (the AIL Framework server) is not directly compromised; the impact is confined to the victim's browser session.
Weakness rationale
- CWE-79 Dynamic values from extracted-match data are interpolated into an HTML-enabled attribute without output encoding, allowing script injection. The fix adds forceescape, confirming the root cause is missing output encoding in an HTML context. CWE-79 is the precise weakness.
Attack pattern rationale
- CAPEC-1 The patch directly addresses stored XSS: attacker-controlled content is persisted in application data (chat messages, items) and rendered as executable HTML in a victim's browser via an unescaped data-content attribute. CAPEC-1 is the canonical attack pattern for XSS and is the closest match. No uncertainty in this mapping.
Assumptions to verify
- The AIL Framework requires authenticated access to create or modify chat messages and items; PR:L is assumed based on the application being a threat-intelligence platform with user accounts.
- The extracted-match data (icon color, style, subtype, id, name, description) can be influenced by an authenticated user through normal application workflows (e.g., submitting content that the extraction pipeline processes).
- The popover is triggered on hover (data-trigger="hover"), classified as passive user interaction (UI:P).
- The tag v7.1 boundary with 143 commits after the fix suggests the fix landed before v7.1, but exact affected version ranges are not confirmed by the patch alone.
- CAPEC-1 is selected as the closest standard attack pattern; no more specific CAPEC for stored XSS in HTML attributes exists in the CAPEC catalog.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
5 | 9 | medium | 5 |
{
"containers": {
"cna": {
"affected": [
{
"defaultStatus": "unaffected",
"modules": [
"chats_explorer (block_message.html)",
"objects/item (show_item.html)"
],
"product": "ail framework",
"programFiles": [
"var/www/templates/chats_explorer/block_message.html",
"var/www/templates/objects/item/show_item.html"
],
"repo": "https://github.com/ail-project/ail-framework",
"vendor": "ail project",
"versions": [
{
"lessThan": "7.1",
"status": "affected",
"version": "unspecified",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "Aurelien Thirion"
},
{
"lang": "en",
"type": "finder",
"value": "Jeroen Pinoy"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe AIL Framework (ail-project/ail-framework) contains a stored cross-site scripting (XSS) vulnerability in two Jinja2 templates that render popovers for matched, tracked, or tagged content: var/www/templates/chats_explorer/block_message.html and var/www/templates/objects/item/show_item.html. In both templates, dynamic values associated with this content, including icon color, icon style, icon glyph, subtype, identifier, name, description, and matched value, are interpolated directly into the data-content HTML attribute of Bootstrap popover elements without appropriate output encoding. Because the popovers are configured with data-html=\"true\", the content is interpreted as HTML in the victim\u0027s browser. An authenticated attacker who can influence matched, tracked, or tagged content may inject arbitrary HTML or JavaScript into these values. When a victim displays the affected popover, the injected markup may execute in the victim\u0027s session, potentially enabling data exfiltration or actions with the victim\u0027s privileges. The vulnerability is classified as stored XSS because the malicious payload can persist in the affected match, tracking, or tag-related data and be delivered to users who view the affected content.\u003c/p\u003e"
}
],
"value": "The AIL Framework (ail-project/ail-framework) contains a stored cross-site scripting (XSS) vulnerability in two Jinja2 templates that render popovers for matched, tracked, or tagged content: var/www/templates/chats_explorer/block_message.html and var/www/templates/objects/item/show_item.html. In both templates, dynamic values associated with this content, including icon color, icon style, icon glyph, subtype, identifier, name, description, and matched value, are interpolated directly into the data-content HTML attribute of Bootstrap popover elements without appropriate output encoding. Because the popovers are configured with data-html=\"true\", the content is interpreted as HTML in the victim\u0027s browser. An authenticated attacker who can influence matched, tracked, or tagged content may inject arbitrary HTML or JavaScript into these values. When a victim displays the affected popover, the injected markup may execute in the victim\u0027s session, potentially enabling data exfiltration or actions with the victim\u0027s privileges. The vulnerability is classified as stored XSS because the malicious payload can persist in the affected match, tracking, or tag-related data and be delivered to users who view the affected content."
}
],
"impacts": [
{
"capecId": "CAPEC-1",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-1 Cross Site Scripting (XSS)"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 8.5,
"baseSeverity": "HIGH",
"privilegesRequired": "LOW",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "NONE",
"subIntegrityImpact": "NONE",
"userInteraction": "PASSIVE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:N/SC:N/SI:N/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "HIGH",
"vulnIntegrityImpact": "HIGH",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-79",
"description": "CWE-79 Improper Neutralization of Input During Web Page Generation (\u0027Cross-site Scripting\u0027)",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/ail-project/ail-framework/commit/ee63a0a96646790255e038f5ca5dd0ab5fe98db2"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe fix applies double HTML-escaping (Jinja2 forceescape|forceescape) to every dynamic value interpolated into the HTML-enabled data-content attribute of the popover elements in both templates. Double-escaping is required because the values must survive both the outer HTML attribute context and the inner HTML parsing performed by the popover plugin. This neutralizes any HTML or script injection from the extracted-match fields.\u003c/p\u003e"
}
],
"value": "The fix applies double HTML-escaping (Jinja2 forceescape|forceescape) to every dynamic value interpolated into the HTML-enabled data-content attribute of the popover elements in both templates. Double-escaping is required because the values must survive both the outer HTML attribute context and the inner HTML parsing performed by the popover plugin. This neutralizes any HTML or script injection from the extracted-match fields."
}
],
"source": {
"discovery": "UNKNOWN"
},
"title": "Stored XSS in AIL Framework extracted-match popovers via unescaped dynamic values in HTML-enabled data-content attributes",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The AIL Framework requires authenticated access to create or modify chat messages and items; PR:L is assumed based on the application being a threat-intelligence platform with user accounts.",
"The extracted-match data (icon color, style, subtype, id, name, description) can be influenced by an authenticated user through normal application workflows (e.g., submitting content that the extraction pipeline processes).",
"The popover is triggered on hover (data-trigger=\"hover\"), classified as passive user interaction (UI:P).",
"The tag v7.1 boundary with 143 commits after the fix suggests the fix landed before v7.1, but exact affected version ranges are not confirmed by the patch alone.",
"CAPEC-1 is selected as the closest standard attack pattern; no more specific CAPEC for stored XSS in HTML attributes exists in the CAPEC catalog."
],
"capecRationale": [
{
"capecId": "CAPEC-1",
"rationale": "The patch directly addresses stored XSS: attacker-controlled content is persisted in application data (chat messages, items) and rendered as executable HTML in a victim\u0027s browser via an unescaped data-content attribute. CAPEC-1 is the canonical attack pattern for XSS and is the closest match. No uncertainty in this mapping."
}
],
"commit": "ee63a0a96646790255e038f5ca5dd0ab5fe98db2",
"confidence": "medium",
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "terrtia"
}
],
"cvssRationale": "AV:N: web application reachable over the network. AC:L: crafting a message or item with a malicious extracted-match value requires no special race or timing. AT:N: no manipulation of the attack target is needed. PR:L: the attacker must be an authenticated user capable of creating or modifying chat messages or items. UI:P: the victim must passively interact (hover over the element) to trigger the popover rendering; no click or explicit action is required. VC:H: the injected script can read session cookies, DOM data, and API responses in the victim\u0027s browser. VI:H: the script can perform authenticated actions on behalf of the victim (create, modify, delete data). VA:N: no direct denial-of-service impact is evident. SC/SI/SA:N: the subcomponent (the AIL Framework server) is not directly compromised; the impact is confined to the victim\u0027s browser session.",
"fixSummary": "The fix applies double HTML-escaping (Jinja2 forceescape|forceescape) to every dynamic value interpolated into the HTML-enabled data-content attribute of the popover elements in both templates. Double-escaping is required because the values must survive both the outer HTML attribute context and the inner HTML parsing performed by the popover plugin. This neutralizes any HTML or script injection from the extracted-match fields.",
"generatedAt": "2026-09-25T12:56:24.137641Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 5,
"confidence": "medium",
"model": "qwen3.8:27b",
"score": 5
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "bd188338f8828ab59743776228d2991b51a3f83b870dedf073a54c6139ad9e26",
"patchSummary": "In block_message.html and show_item.html, the Jinja2 expressions for extracted_matches fields (icon color, icon style, icon, subtype, id, name, description) and the matched value r[1] are each suffixed with |forceescape|forceescape. No other logic, routing, or configuration changes are made. Two lines changed in total (one per file).",
"patchTruncated": false,
"patches": [
{
"commit": "ee63a0a96646790255e038f5ca5dd0ab5fe98db2",
"patchSha256": "bd188338f8828ab59743776228d2991b51a3f83b870dedf073a54c6139ad9e26",
"source": "https://github.com/ail-project/ail-framework/commit/ee63a0a96646790255e038f5ca5dd0ab5fe98db2.patch",
"sourceUrl": "https://github.com/ail-project/ail-framework/commit/ee63a0a96646790255e038f5ca5dd0ab5fe98db2.patch",
"subject": "chg: [security] fix XSS in show item and message popover of"
}
],
"source": "https://github.com/ail-project/ail-framework/commit/ee63a0a96646790255e038f5ca5dd0ab5fe98db2.patch",
"subject": "chg: [security] fix XSS in show item and message popover of",
"tagVersionBoundary": {
"commits_after_fix": 143,
"repository": "https://github.com/ail-project/ail-framework",
"tag": "v7.1",
"version": "7.1",
"version_type": "custom"
},
"weaknessRationale": [
{
"cweId": "CWE-79",
"rationale": "Dynamic values from extracted-match data are interpolated into an HTML-enabled attribute without output encoding, allowing script injection. The fix adds forceescape, confirming the root cause is missing output encoding in an HTML context. CWE-79 is the precise weakness."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20239"
}
],
"x_generator": {
"engine": "Vulnogram 0.2.0"
}
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-100172",
"datePublished": "2026-09-25T13:13:00.000Z",
"dateReserved": "2026-09-25T13:13:00.000Z",
"dateUpdated": "2026-09-25T13:54:03.667825Z",
"requesterUserId": "00000000-0000-4000-9000-000000000000",
"serial": 1,
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20239"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}
GCVE-1-2026-20276 (CVE-2026-100187)
Vulnerability from gna-1 – Published: 2026-09-25 13:40 – Updated: 2026-09-25 13:53- CWE-20 - Improper Input Validation
| URL | Tags |
|---|---|
| https://github.com/ail-project/ail-framework/comm… | patch |
| Vendor | Product | Version | |
|---|---|---|---|
| ail project | ail framework |
Affected:
unspecified , < 7.1
(semver)
|
qwen3.8:27b
advisory
bcp-05-x-01bcp-05-x-02
Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.
| Model | Source | Identifier |
|---|---|---|
| qwen3.8:27b | ollama | qwen3.8:27b |
- Generator
-
patch2vuln.pyon 2026-09-25 13:37 - Model
qwen3.8:27b- Input
-
https://github.com/ail-project/ail-framework/commit/5c8a68b3d8c7d7b03e5febe4ac5b5aca91826adb.patch
9fbf19112822… - Confidence
- medium
| Commit | Subject | Patch SHA-256 |
|---|---|---|
5c8a68b3d8c7
|
chg: [security] Onion module, prevent non-onion crawler task | 9fbf19112822… |
Fix summary
The length-based shortcut that bypassed onion-domain validation is removed. Every extracted URL is now parsed using the standard URL domain extraction function, and the resulting hostname is validated against the existing onion-domain validation routine before the URL is added to the crawler task set. This ensures that only URLs whose hostname is a genuine .onion domain can be queued as crawler tasks.
Patch summary
In bin/modules/Onion.py, the conditional block that checked len(url) == 69 and url.endswith('.onion') to shortcut domain extraction (url[7:]) without calling crawlers.is_valid_onion_domain() is removed. The code is replaced with an unconditional call to crawlers.get_url_domain(url) followed by a check that the returned domain is truthy and passes crawlers.is_valid_onion_domain(domain) before adding the domain and URL to the respective sets. Net change: 2 insertions, 9 deletions in a single file.
CVSS rationale
Attack Vector is Network because the attacker publishes crafted content on the web that the framework crawls. Attack Complexity is Low because the attacker only needs to construct a URL of exactly 69 characters ending in .onion with a non-onion hostname. No special Adversary Techniques are required. Privileges Required is None because the attacker is an unauthenticated content publisher. User Interaction is None because the crawler processes the content automatically. The primary impact is Low Integrity on the vulnerable component: the crawler task queue is polluted with non-onion URLs, violating the integrity of the framework's target selection. No confidentiality or availability impact is evident from the patch. No subsequent component impact is identified in the evidence.
Weakness rationale
- CWE-20 The URL input was validated using an insufficient heuristic (string length and suffix) that could be satisfied by non-onion URLs. The proper hostname parsing and onion-domain validation step was bypassed by the shortcut, allowing attacker-controlled URLs to be accepted as valid onion targets.
Attack pattern rationale
- CAPEC-126 The attacker tampers with URL parameters embedded in untrusted crawled content to bypass the application's validation logic and inject non-onion targets into the crawler task queue. CAPEC-126 is the closest available pattern describing manipulation of input parameters to achieve unintended application behavior. The mapping is approximate because the attack vector is content-based injection rather than direct request parameter manipulation, but the core mechanism (crafting input to bypass validation) aligns with this pattern.
Assumptions to verify
- The affected version range is inferred from the tag_version_boundary metadata indicating the fix commit is 24 commits before the v7.1 tag; the exact version number of the vulnerable release is not specified in the patch.
- The CVSS assumes the attacker can place crafted content in a location the AIL Framework crawler will process; the specific crawling scope and configuration are not detailed in the patch.
- CAPEC-126 (Parameter Tampering) is the closest available attack pattern; the actual attack is content-based URL injection rather than classic request parameter tampering, so the mapping is approximate.
- The impact is assessed as Low Integrity based on the evidence that non-onion URLs are queued as crawler tasks; broader downstream consequences (e.g., SSRF-like effects) are not evidenced in the patch and are not assumed.
- The commit date (24 Sep 2026) is taken as-is from the patch metadata; no independent verification of the timeline is performed.
Model comparison
Selected qwen3.8:27b
by deterministic-consensus-v1
The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required.
| Model | Score | Agreement | Confidence | Assumptions |
|---|---|---|---|---|
qwen3.8:27b |
5 | 9 | medium | 5 |
{
"containers": {
"cna": {
"affected": [
{
"defaultStatus": "unaffected",
"modules": [
"Onion module"
],
"product": "ail framework",
"programFiles": [
"bin/modules/Onion.py"
],
"repo": "https://github.com/ail-project/ail-framework",
"vendor": "ail project",
"versions": [
{
"lessThan": "7.1",
"status": "affected",
"version": "unspecified",
"versionType": "semver"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "Aurelien Thirion"
},
{
"lang": "en",
"type": "finder",
"value": "Jeroen Pinoy"
}
],
"descriptions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe Onion module in AIL Framework contained a performance shortcut in its URL extraction logic that accepted URLs as valid .onion targets based solely on a length check (exactly 69 characters) and a suffix check (ending in \".onion\"), without performing proper hostname parsing or onion-domain validation. An unauthenticated attacker who could publish or control web content crawled by the framework could embed a crafted URL containing an IP address or non-onion hostname with a path ending in \".onion\" that satisfied the length and suffix conditions. Such a URL would be extracted, its domain naively sliced from the string, and queued as a legitimate onion crawler task. This allowed unauthenticated content publishers to inject arbitrary non-onion targets into the crawler\u0027s task queue, influencing crawler behavior and potentially directing it toward unintended network resources. The vulnerability required no authentication, no user interaction, and only the ability to place crafted content in a location the framework would crawl. The security impact is a loss of integrity in the crawler\u0027s target selection: the framework processes and acts upon URLs that do not correspond to legitimate .onion services.\u003c/p\u003e"
}
],
"value": "The Onion module in AIL Framework contained a performance shortcut in its URL extraction logic that accepted URLs as valid .onion targets based solely on a length check (exactly 69 characters) and a suffix check (ending in \".onion\"), without performing proper hostname parsing or onion-domain validation. An unauthenticated attacker who could publish or control web content crawled by the framework could embed a crafted URL containing an IP address or non-onion hostname with a path ending in \".onion\" that satisfied the length and suffix conditions. Such a URL would be extracted, its domain naively sliced from the string, and queued as a legitimate onion crawler task. This allowed unauthenticated content publishers to inject arbitrary non-onion targets into the crawler\u0027s task queue, influencing crawler behavior and potentially directing it toward unintended network resources. The vulnerability required no authentication, no user interaction, and only the ability to place crafted content in a location the framework would crawl. The security impact is a loss of integrity in the crawler\u0027s target selection: the framework processes and acts upon URLs that do not correspond to legitimate .onion services."
}
],
"impacts": [
{
"capecId": "CAPEC-126",
"descriptions": [
{
"lang": "en",
"value": "CAPEC-126 Parameter Tampering"
}
]
}
],
"metrics": [
{
"cvssV4_0": {
"Automatable": "NOT_DEFINED",
"Recovery": "NOT_DEFINED",
"Safety": "NOT_DEFINED",
"attackComplexity": "LOW",
"attackRequirements": "NONE",
"attackVector": "NETWORK",
"baseScore": 6.9,
"baseSeverity": "MEDIUM",
"privilegesRequired": "NONE",
"providerUrgency": "NOT_DEFINED",
"subAvailabilityImpact": "NONE",
"subConfidentialityImpact": "NONE",
"subIntegrityImpact": "NONE",
"userInteraction": "NONE",
"valueDensity": "NOT_DEFINED",
"vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:L/VA:N/SC:N/SI:N/SA:N",
"version": "4.0",
"vulnAvailabilityImpact": "NONE",
"vulnConfidentialityImpact": "NONE",
"vulnIntegrityImpact": "LOW",
"vulnerabilityResponseEffort": "NOT_DEFINED"
},
"format": "CVSS",
"scenarios": [
{
"lang": "en",
"value": "GENERAL"
}
]
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-20",
"description": "CWE-20 Improper Input Validation",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"orgId": "00000000-0000-4000-9000-000000000000"
},
"references": [
{
"name": "Security patch",
"tags": [
"patch"
],
"url": "https://github.com/ail-project/ail-framework/commit/5c8a68b3d8c7d7b03e5febe4ac5b5aca91826adb"
}
],
"solutions": [
{
"lang": "en",
"supportingMedia": [
{
"base64": false,
"type": "text/html",
"value": "\u003cp\u003eThe length-based shortcut that bypassed onion-domain validation is removed. Every extracted URL is now parsed using the standard URL domain extraction function, and the resulting hostname is validated against the existing onion-domain validation routine before the URL is added to the crawler task set. This ensures that only URLs whose hostname is a genuine .onion domain can be queued as crawler tasks.\u003c/p\u003e"
}
],
"value": "The length-based shortcut that bypassed onion-domain validation is removed. Every extracted URL is now parsed using the standard URL domain extraction function, and the resulting hostname is validated against the existing onion-domain validation routine before the URL is added to the crawler task set. This ensures that only URLs whose hostname is a genuine .onion domain can be queued as crawler tasks."
}
],
"source": {
"discovery": "UNKNOWN"
},
"title": "AIL Framework Onion Module: Non-Onion URL Accepted as Crawler Task Due to Bypassed Domain Validation",
"x_gcve": [
{
"extensions": {
"bcp-05-x-01": {
"ai_annotations": [
{
"ai_level": "generated",
"description": "Draft vulnerability metadata was generated from a git-format patch using an Ollama-hosted language model. Human validation is required before publication.",
"gna_source": 1,
"models": [
{
"gna_source": 1,
"identifier": "qwen3.8:27b",
"name": "qwen3.8:27b",
"source": "ollama"
}
],
"review_status": "full",
"scope": "record",
"tags": [
"ai-computer-assisted:llm-generated",
"ai-computer-assisted:classification"
]
}
]
},
"bcp-05-x-02": {
"x_patch2vuln": {
"assumptions": [
"The affected version range is inferred from the tag_version_boundary metadata indicating the fix commit is 24 commits before the v7.1 tag; the exact version number of the vulnerable release is not specified in the patch.",
"The CVSS assumes the attacker can place crafted content in a location the AIL Framework crawler will process; the specific crawling scope and configuration are not detailed in the patch.",
"CAPEC-126 (Parameter Tampering) is the closest available attack pattern; the actual attack is content-based URL injection rather than classic request parameter tampering, so the mapping is approximate.",
"The impact is assessed as Low Integrity based on the evidence that non-onion URLs are queued as crawler tasks; broader downstream consequences (e.g., SSRF-like effects) are not evidenced in the patch and are not assumed.",
"The commit date (24 Sep 2026) is taken as-is from the patch metadata; no independent verification of the timeline is performed."
],
"capecRationale": [
{
"capecId": "CAPEC-126",
"rationale": "The attacker tampers with URL parameters embedded in untrusted crawled content to bypass the application\u0027s validation logic and inject non-onion targets into the crawler task queue. CAPEC-126 is the closest available pattern describing manipulation of input parameters to achieve unintended application behavior. The mapping is approximate because the attack vector is content-based injection rather than direct request parameter manipulation, but the core mechanism (crafting input to bypass validation) aligns with this pattern."
}
],
"commit": "5c8a68b3d8c7d7b03e5febe4ac5b5aca91826adb",
"confidence": "medium",
"credits": [
{
"lang": "en",
"type": "remediation developer",
"value": "terrtia"
}
],
"cvssRationale": "Attack Vector is Network because the attacker publishes crafted content on the web that the framework crawls. Attack Complexity is Low because the attacker only needs to construct a URL of exactly 69 characters ending in .onion with a non-onion hostname. No special Adversary Techniques are required. Privileges Required is None because the attacker is an unauthenticated content publisher. User Interaction is None because the crawler processes the content automatically. The primary impact is Low Integrity on the vulnerable component: the crawler task queue is polluted with non-onion URLs, violating the integrity of the framework\u0027s target selection. No confidentiality or availability impact is evident from the patch. No subsequent component impact is identified in the evidence.",
"fixSummary": "The length-based shortcut that bypassed onion-domain validation is removed. Every extracted URL is now parsed using the standard URL domain extraction function, and the resulting hostname is validated against the existing onion-domain validation routine before the URL is added to the crawler task set. This ensures that only URLs whose hostname is a genuine .onion domain can be queued as crawler tasks.",
"generatedAt": "2026-09-25T13:37:37.543009Z",
"generator": "patch2vuln.py",
"model": "qwen3.8:27b",
"modelComparison": {
"rankings": [
{
"agreementScore": 9,
"assumptionCount": 5,
"confidence": "medium",
"model": "qwen3.8:27b",
"score": 5
}
],
"selectedModel": "qwen3.8:27b",
"selectionMethod": "deterministic-consensus-v1",
"selectionNotice": "The selected result is closest to model consensus; this heuristic does not establish factual correctness and human review remains required."
},
"patchSha256": "9fbf191128229c1859f3643b5f50921cdf5f36fd70734e8f68a05e237b987d60",
"patchSummary": "In bin/modules/Onion.py, the conditional block that checked len(url) == 69 and url.endswith(\u0027.onion\u0027) to shortcut domain extraction (url[7:]) without calling crawlers.is_valid_onion_domain() is removed. The code is replaced with an unconditional call to crawlers.get_url_domain(url) followed by a check that the returned domain is truthy and passes crawlers.is_valid_onion_domain(domain) before adding the domain and URL to the respective sets. Net change: 2 insertions, 9 deletions in a single file.",
"patchTruncated": false,
"patches": [
{
"commit": "5c8a68b3d8c7d7b03e5febe4ac5b5aca91826adb",
"patchSha256": "9fbf191128229c1859f3643b5f50921cdf5f36fd70734e8f68a05e237b987d60",
"source": "https://github.com/ail-project/ail-framework/commit/5c8a68b3d8c7d7b03e5febe4ac5b5aca91826adb.patch",
"sourceUrl": "https://github.com/ail-project/ail-framework/commit/5c8a68b3d8c7d7b03e5febe4ac5b5aca91826adb.patch",
"subject": "chg: [security] Onion module, prevent non-onion crawler task"
}
],
"source": "https://github.com/ail-project/ail-framework/commit/5c8a68b3d8c7d7b03e5febe4ac5b5aca91826adb.patch",
"subject": "chg: [security] Onion module, prevent non-onion crawler task",
"tagVersionBoundary": {
"commits_after_fix": 24,
"repository": "https://github.com/ail-project/ail-framework",
"tag": "v7.1",
"version": "7.1",
"version_type": "custom"
},
"weaknessRationale": [
{
"cweId": "CWE-20",
"rationale": "The URL input was validated using an insufficient heuristic (string length and suffix) that could be satisfied by non-onion URLs. The proper hostname parsing and onion-domain validation step was bypassed by the shortcut, allowing attacker-controlled URLs to be accepted as valid onion targets."
}
]
}
}
},
"recordType": "advisory",
"vulnId": "GCVE-1-2026-20276"
}
],
"x_generator": {
"engine": "Vulnogram 0.2.0"
}
}
},
"cveMetadata": {
"assignerOrgId": "00000000-0000-4000-9000-000000000000",
"cveId": "CVE-2026-100187",
"datePublished": "2026-09-25T13:40:00.000Z",
"dateReserved": "2026-09-25T13:42:00.000Z",
"dateUpdated": "2026-09-25T13:53:04.756019Z",
"requesterUserId": "00000000-0000-4000-9000-000000000000",
"serial": 1,
"state": "PUBLISHED",
"vulnId": "GCVE-1-2026-20276"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.1"
}