PYSEC-2026-2937
Vulnerability from pysec - Published: 2026-07-13 15:35 - Updated: 2026-07-13 16:05Summary
Type: Authorization bypass enabling workspace metadata + settings tampering. The PATCH /workspaces/{workspace_id} endpoint is gated only by require_workspace_member(workspace_id) (default min_role="member"). Any member can rewrite the workspace's name, description, and the settings JSON blob. The settings field is a free-form JSON object — depending on which downstream code reads it, this becomes a configuration-injection primitive for any setting the platform exposes there.
File: src/praisonai-platform/praisonai_platform/api/routes/workspaces.py, lines 63-74; services/workspace_service.py's update() method.
Root cause: Depends(require_workspace_member) resolves to default min_role="member". WorkspaceService.update(workspace_id, name, description, settings) writes the new fields to the workspace row without any caller-permission check. The role hierarchy (MemberService.has_role) is never consulted.
Affected Code
File: src/praisonai-platform/praisonai_platform/api/routes/workspaces.py, lines 63-74.
@router.patch("/{workspace_id}", response_model=WorkspaceResponse)
async def update_workspace(
workspace_id: str,
body: WorkspaceUpdate,
user: AuthIdentity = Depends(require_workspace_member), # <-- BUG: defaults to min_role="member"
session: AsyncSession = Depends(get_db),
):
ws_svc = WorkspaceService(session)
ws = await ws_svc.update(workspace_id, body.name, body.description, body.settings) # <-- writes any value
if ws is None:
raise HTTPException(status_code=404, detail="Workspace not found")
return WorkspaceResponse.model_validate(ws)
Why it's wrong: workspace name and settings are owner-tier fields. Renaming the workspace to a profanity is a low-impact griefing vector; rewriting the JSON settings blob is potentially a much higher-impact configuration injection (depending on what fields downstream code reads from settings, the attacker may flip feature flags, redirect webhook URLs, change LLM provider keys for shared configs, disable audit logging, etc.). The require_workspace_member(min_role) parameter is implemented and unused. This endpoint should require owner.
Exploit Chain
- Attacker is a member of workspace
Wwith role "member". State: attacker holds JWT. - Attacker sends
PATCH /workspaces/WwithAuthorization: Bearer <attacker_jwt>and body{"name": "Compromised", "description": "Owned by attacker", "settings": {"allow_public_invite": true, "ai_provider_url": "https://attacker.example/v1"}}. State: control flow entersupdate_workspace. require_workspace_member(W, attacker)passes.WorkspaceService.update(W, ...)writes the three fields. State: workspaceWnow has attacker-chosen name, description, and settings.- The settings JSON is read by any downstream code that consults workspace settings (LLM proxying, invite flows, webhook routing). If the deployment uses settings-keyed configuration overrides, those overrides now point at attacker-controlled endpoints.
- Final state: with one member-level token plus one PATCH, the attacker rewrites the workspace's metadata and settings, with effects ranging from cosmetic (rename) to substantive (settings-keyed config injection).
Security Impact
Severity: sec-moderate. CVSS 6.5: network attack, low complexity, low privileges, no user interaction, scope unchanged, no confidentiality directly (though settings rewrites may enable indirect data exfiltration via attacker-pointed integration URLs), high integrity, no availability claim.
Attacker capability: rewrite any workspace's name, description, and settings JSON. The actual blast radius depends on what fields the deployment reads from settings — but that field is documented as a free-form JSON blob, so any future configuration the platform adds there becomes attacker-tunable.
Preconditions: praisonai-platform is deployed multi-tenant; attacker has any membership token in the target workspace.
Differential: source-inspection-verified. With the suggested fix below, member-tier tokens fail the gate and the metadata rewrite is rejected with 403.
Suggested Fix
--- a/src/praisonai-platform/praisonai_platform/api/routes/workspaces.py
+++ b/src/praisonai-platform/praisonai_platform/api/routes/workspaces.py
@@ -63,11 +63,11 @@
@router.patch("/{workspace_id}", response_model=WorkspaceResponse)
async def update_workspace(
workspace_id: str,
body: WorkspaceUpdate,
- user: AuthIdentity = Depends(require_workspace_member),
+ user: AuthIdentity = Depends(_require_workspace_owner), # see member-update-role advisory for helper
session: AsyncSession = Depends(get_db),
):
ws_svc = WorkspaceService(session)
ws = await ws_svc.update(workspace_id, body.name, body.description, body.settings)
if ws is None:
raise HTTPException(status_code=404, detail="Workspace not found")
return WorkspaceResponse.model_validate(ws)
Defence-in-depth: validate the keys allowed in body.settings against an allowlist so the field cannot become an arbitrary config-injection primitive even for owners. The four companion workspace-mutation endpoints (add_member, update_member_role, remove_member, delete_workspace) exhibit the same default-min-role gap and are filed as their own advisories.
| Name | purl | praisonai-platform | pkg:pypi/praisonai-platform |
|---|
{
"affected": [
{
"package": {
"ecosystem": "PyPI",
"name": "praisonai-platform",
"purl": "pkg:pypi/praisonai-platform"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.1.4"
}
],
"type": "ECOSYSTEM"
}
],
"versions": [
"0.1.0",
"0.1.1",
"0.1.2",
"0.1.3"
]
}
],
"aliases": [
"CVE-2026-47411",
"GHSA-rcmc-q9rj-4wmq"
],
"details": "## Summary\n\n**Type:** Authorization bypass enabling workspace metadata + settings tampering. The `PATCH /workspaces/{workspace_id}` endpoint is gated only by `require_workspace_member(workspace_id)` (default `min_role=\"member\"`). Any member can rewrite the workspace\u0027s `name`, `description`, and the `settings` JSON blob. The settings field is a free-form JSON object \u2014 depending on which downstream code reads it, this becomes a configuration-injection primitive for any setting the platform exposes there.\n**File:** `src/praisonai-platform/praisonai_platform/api/routes/workspaces.py`, lines 63-74; `services/workspace_service.py`\u0027s `update()` method.\n**Root cause:** `Depends(require_workspace_member)` resolves to default `min_role=\"member\"`. `WorkspaceService.update(workspace_id, name, description, settings)` writes the new fields to the workspace row without any caller-permission check. The role hierarchy (`MemberService.has_role`) is never consulted.\n\n## Affected Code\n\n**File:** `src/praisonai-platform/praisonai_platform/api/routes/workspaces.py`, lines 63-74.\n\n```python\n@router.patch(\"/{workspace_id}\", response_model=WorkspaceResponse)\nasync def update_workspace(\n workspace_id: str,\n body: WorkspaceUpdate,\n user: AuthIdentity = Depends(require_workspace_member), # \u003c-- BUG: defaults to min_role=\"member\"\n session: AsyncSession = Depends(get_db),\n):\n ws_svc = WorkspaceService(session)\n ws = await ws_svc.update(workspace_id, body.name, body.description, body.settings) # \u003c-- writes any value\n if ws is None:\n raise HTTPException(status_code=404, detail=\"Workspace not found\")\n return WorkspaceResponse.model_validate(ws)\n```\n\n**Why it\u0027s wrong:** workspace name and settings are owner-tier fields. Renaming the workspace to a profanity is a low-impact griefing vector; rewriting the JSON `settings` blob is potentially a much higher-impact configuration injection (depending on what fields downstream code reads from `settings`, the attacker may flip feature flags, redirect webhook URLs, change LLM provider keys for shared configs, disable audit logging, etc.). The `require_workspace_member(min_role)` parameter is implemented and unused. This endpoint should require owner.\n\n## Exploit Chain\n\n1. Attacker is a member of workspace `W` with role \"member\". State: attacker holds JWT.\n2. Attacker sends `PATCH /workspaces/W` with `Authorization: Bearer \u003cattacker_jwt\u003e` and body `{\"name\": \"Compromised\", \"description\": \"Owned by attacker\", \"settings\": {\"allow_public_invite\": true, \"ai_provider_url\": \"https://attacker.example/v1\"}}`. State: control flow enters `update_workspace`.\n3. `require_workspace_member(W, attacker)` passes. `WorkspaceService.update(W, ...)` writes the three fields. State: workspace `W` now has attacker-chosen name, description, and settings.\n4. The settings JSON is read by any downstream code that consults workspace settings (LLM proxying, invite flows, webhook routing). If the deployment uses settings-keyed configuration overrides, those overrides now point at attacker-controlled endpoints.\n5. Final state: with one member-level token plus one PATCH, the attacker rewrites the workspace\u0027s metadata and settings, with effects ranging from cosmetic (rename) to substantive (settings-keyed config injection).\n\n## Security Impact\n\n**Severity:** sec-moderate. CVSS 6.5: network attack, low complexity, low privileges, no user interaction, scope unchanged, no confidentiality directly (though settings rewrites may enable indirect data exfiltration via attacker-pointed integration URLs), high integrity, no availability claim.\n**Attacker capability:** rewrite any workspace\u0027s name, description, and settings JSON. The actual blast radius depends on what fields the deployment reads from `settings` \u2014 but that field is documented as a free-form JSON blob, so any future configuration the platform adds there becomes attacker-tunable.\n**Preconditions:** `praisonai-platform` is deployed multi-tenant; attacker has any membership token in the target workspace.\n**Differential:** source-inspection-verified. With the suggested fix below, member-tier tokens fail the gate and the metadata rewrite is rejected with 403.\n\n## Suggested Fix\n\n```diff\n--- a/src/praisonai-platform/praisonai_platform/api/routes/workspaces.py\n+++ b/src/praisonai-platform/praisonai_platform/api/routes/workspaces.py\n@@ -63,11 +63,11 @@\n @router.patch(\"/{workspace_id}\", response_model=WorkspaceResponse)\n async def update_workspace(\n workspace_id: str,\n body: WorkspaceUpdate,\n- user: AuthIdentity = Depends(require_workspace_member),\n+ user: AuthIdentity = Depends(_require_workspace_owner), # see member-update-role advisory for helper\n session: AsyncSession = Depends(get_db),\n ):\n ws_svc = WorkspaceService(session)\n ws = await ws_svc.update(workspace_id, body.name, body.description, body.settings)\n if ws is None:\n raise HTTPException(status_code=404, detail=\"Workspace not found\")\n return WorkspaceResponse.model_validate(ws)\n```\n\nDefence-in-depth: validate the keys allowed in `body.settings` against an allowlist so the field cannot become an arbitrary config-injection primitive even for owners. The four companion workspace-mutation endpoints (`add_member`, `update_member_role`, `remove_member`, `delete_workspace`) exhibit the same default-min-role gap and are filed as their own advisories.",
"id": "PYSEC-2026-2937",
"modified": "2026-07-13T16:05:38.312807Z",
"published": "2026-07-13T15:35:23.490169Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/MervinPraison/PraisonAI/security/advisories/GHSA-rcmc-q9rj-4wmq"
},
{
"type": "PACKAGE",
"url": "https://github.com/MervinPraison/PraisonAI"
},
{
"type": "PACKAGE",
"url": "https://pypi.org/project/praisonai-platform"
},
{
"type": "ADVISORY",
"url": "https://github.com/advisories/GHSA-rcmc-q9rj-4wmq"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-47411"
}
],
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:N",
"type": "CVSS_V3"
}
],
"summary": "praisonai-platform: Any workspace member can rewrite workspace name, description, and settings via PATCH /workspaces/{id}"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
How durations are calculated
Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.
Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.