GHSA-H36F-RQPX-J5WX
Vulnerability from github – Published: 2026-05-08 20:03 – Updated: 2026-05-15 23:53Unauthorized File and Knowledge Base Content Access via RAG Vector Search
Affected Component
RAG source resolution in chat completion pipeline:
- backend/open_webui/retrieval/utils.py (lines 963-965, 1063-1068, 1126-1131 in get_sources_from_items)
Affected Versions
Current main branch (commit 6fdd19bf1) and likely all versions with RAG functionality.
Description
The get_sources_from_items function resolves file and knowledge base references into vector search queries during chat completion. Three of the five code paths perform vector store queries without any authorization check, allowing users to extract content from files and knowledge bases they do not have access to.
| Path | Lines | Access Check |
|---|---|---|
type: "file", full-context |
1044-1050 | ✅ has_access_to_file |
type: "file", non-full-context (default) |
1063-1068 | ❌ None |
type: "collection" |
1070-1118 | ✅ Present |
type: "text" with collection_name |
963-965 | ❌ None |
Bare collection_name/collection_names |
1126-1131 | ❌ None |
The three unprotected paths pass user-supplied collection names directly to query_collection(), which queries the vector store without any authorization. Collection names follow predictable formats: file-<file_id> for files and the knowledge base UUID for knowledge bases.
CVSS 3.1 Breakdown
| Metric | Value | Rationale |
|---|---|---|
| Attack Vector | Network (N) | Exploited remotely via chat completion API |
| Attack Complexity | Low (L) | Single API call with a known resource ID |
| Privileges Required | Low (L) | Requires a valid user account |
| User Interaction | None (N) | No victim interaction required |
| Scope | Unchanged (U) | Impact within the application's data boundary |
| Confidentiality | High (H) | Full content of private files/knowledge bases extractable |
| Integrity | None (N) | No data modification |
| Availability | None (N) | No denial of service |
Attack Scenario
- User A uploads a private document and uses it in RAG (the document is embedded into the vector store as collection
file-<file_id>). - User A shares a chat or model referencing the file with User B, or User B otherwise obtains the file ID through a legitimate interaction.
- User A later revokes User B's access to the file.
- User B sends a chat completion request referencing the revoked file:
json POST /api/chat/completions { "model": "any-accessible-model", "messages": [{"role": "user", "content": "What does this document say about pricing?"}], "files": [{"type": "file", "id": "<revoked_file_id>"}] } - The non-full-context path (default) constructs collection name
file-<id>and queries the vector store with no access check. - Matching chunks are injected into the LLM context, and the response contains the victim's private file content.
The same attack works via {"type": "text", "collection_name": "<knowledge_base_id>"} for knowledge bases.
Impact
- Access revocation is ineffective for RAG content — users who previously had access can continue extracting file and knowledge base content indefinitely
- Private document content can be systematically extracted through targeted queries
- Breaks the access control model for files and knowledge bases at the RAG layer
Preconditions
- Attacker must know the file ID or knowledge base ID (UUID) of the target resource
- The target file/knowledge base must have been processed into the vector store
- Attacker must have a valid user account
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 0.8.12"
},
"package": {
"ecosystem": "PyPI",
"name": "open-webui"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "0.9.0"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-44560"
],
"database_specific": {
"cwe_ids": [
"CWE-862"
],
"github_reviewed": true,
"github_reviewed_at": "2026-05-08T20:03:09Z",
"nvd_published_at": "2026-05-15T20:16:47Z",
"severity": "MODERATE"
},
"details": "# Unauthorized File and Knowledge Base Content Access via RAG Vector Search\n\n## Affected Component\n\nRAG source resolution in chat completion pipeline:\n- `backend/open_webui/retrieval/utils.py` (lines 963-965, 1063-1068, 1126-1131 in `get_sources_from_items`)\n\n## Affected Versions\n\nCurrent main branch (commit `6fdd19bf1`) and likely all versions with RAG functionality.\n\n## Description\n\nThe `get_sources_from_items` function resolves file and knowledge base references into vector search queries during chat completion. Three of the five code paths perform vector store queries without any authorization check, allowing users to extract content from files and knowledge bases they do not have access to.\n\n| Path | Lines | Access Check |\n|------|-------|-------------|\n| `type: \"file\"`, full-context | 1044-1050 | \u2705 `has_access_to_file` |\n| `type: \"file\"`, non-full-context (default) | 1063-1068 | \u274c None |\n| `type: \"collection\"` | 1070-1118 | \u2705 Present |\n| `type: \"text\"` with `collection_name` | 963-965 | \u274c None |\n| Bare `collection_name`/`collection_names` | 1126-1131 | \u274c None |\n\nThe three unprotected paths pass user-supplied collection names directly to `query_collection()`, which queries the vector store without any authorization. Collection names follow predictable formats: `file-\u003cfile_id\u003e` for files and the knowledge base UUID for knowledge bases.\n\n## CVSS 3.1 Breakdown\n\n| Metric | Value | Rationale |\n|--------|-------|-----------|\n| Attack Vector | Network (N) | Exploited remotely via chat completion API |\n| Attack Complexity | Low (L) | Single API call with a known resource ID |\n| Privileges Required | Low (L) | Requires a valid user account |\n| User Interaction | None (N) | No victim interaction required |\n| Scope | Unchanged (U) | Impact within the application\u0027s data boundary |\n| Confidentiality | High (H) | Full content of private files/knowledge bases extractable |\n| Integrity | None (N) | No data modification |\n| Availability | None (N) | No denial of service |\n\n## Attack Scenario\n\n1. User A uploads a private document and uses it in RAG (the document is embedded into the vector store as collection `file-\u003cfile_id\u003e`).\n2. User A shares a chat or model referencing the file with User B, or User B otherwise obtains the file ID through a legitimate interaction.\n3. User A later revokes User B\u0027s access to the file.\n4. User B sends a chat completion request referencing the revoked file:\n ```json\n POST /api/chat/completions\n {\n \"model\": \"any-accessible-model\",\n \"messages\": [{\"role\": \"user\", \"content\": \"What does this document say about pricing?\"}],\n \"files\": [{\"type\": \"file\", \"id\": \"\u003crevoked_file_id\u003e\"}]\n }\n ```\n5. The non-full-context path (default) constructs collection name `file-\u003cid\u003e` and queries the vector store with no access check.\n6. Matching chunks are injected into the LLM context, and the response contains the victim\u0027s private file content.\n\nThe same attack works via `{\"type\": \"text\", \"collection_name\": \"\u003cknowledge_base_id\u003e\"}` for knowledge bases.\n\n## Impact\n\n- Access revocation is ineffective for RAG content \u2014 users who previously had access can continue extracting file and knowledge base content indefinitely\n- Private document content can be systematically extracted through targeted queries\n- Breaks the access control model for files and knowledge bases at the RAG layer\n\n## Preconditions\n\n- Attacker must know the file ID or knowledge base ID (UUID) of the target resource\n- The target file/knowledge base must have been processed into the vector store\n- Attacker must have a valid user account",
"id": "GHSA-h36f-rqpx-j5wx",
"modified": "2026-05-15T23:53:30Z",
"published": "2026-05-08T20:03:09Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/open-webui/open-webui/security/advisories/GHSA-h36f-rqpx-j5wx"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-44560"
},
{
"type": "PACKAGE",
"url": "https://github.com/open-webui/open-webui"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N",
"type": "CVSS_V3"
}
],
"summary": "Open WebUI has Unauthorized File and Knowledge Base Content Access via RAG Vector Search"
}
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.