CVE-2026-54016: Open WebUI BOLA `search_knowledge_files` Allows Unauthorized Knowledge Base Vulnerability

{
  "schema_version" : "1.4.0",
  "id" : "GHSA-cx9v-4qj2-jrw6",
  "modified" : "2026-07-20T21:07:24Z",
  "published" : "2026-06-17T14:31:16Z",
  "aliases" : [ "CVE-2026-54016" ],
  "summary" : "Open WebUI BOLA: `search_knowledge_files` Allows Unauthorized Knowledge Base File Enumeration",
  "details" : "## Summary\n\nOpen WebUI has a Broken Object Level Authorization (BOLA) vulnerability in the builtin `search_knowledge_files` tool.\n\nWhen native function calling is enabled and the selected model has no attached knowledge bases, an authenticated user can call `search_knowledge_files` with an arbitrary `knowledge_id`. The function then returns file metadata from that knowledge base without checking whether the user has read access.\n\nThis allows unauthorized enumeration of private or restricted knowledge base files.\n\n## Details\n\nThe vulnerable code is in:\n\n`backend/open_webui/tools/builtin.py`\n\nAffected function:\n\n```python\nasync def search_knowledge_files(\n    query: str,\n    knowledge_id: Optional[str] = None,\n    count: int = 5,\n    skip: int = 0,\n    __request__: Request = None,\n    __user__: dict = None,\n    __model_knowledge__: Optional[list[dict]] = None,\n) -> str:\n```\n\nIn the \"No attached knowledge\" branch, when `knowledge_id` is provided, the function directly calls:\n\n```python\nresult = await Knowledges.search_files_by_id(\n    knowledge_id=knowledge_id,\n    user_id=user_id,\n    filter={\"query\": query},\n    skip=skip,\n    limit=count,\n)\n```\n\nThis code path does not verify that the current user is authorized to access the specified knowledge base.\n\nThe missing check is inconsistent with other nearby code paths. For example, the attached-knowledge branch in the same function checks whether the user is an admin, the owner of the knowledge base, or has explicit read access through `AccessGrants`:\n\n```python\nif not (\n    user_role == \"admin\"\n    or knowledge.user_id == user_id\n    or await AccessGrants.has_access(\n        user_id=user_id,\n        resource_type=\"knowledge\",\n        resource_id=knowledge.id,\n        permission=\"read\",\n        user_group_ids=set(user_group_ids),\n    )\n):\n    continue\n```\n\nThe sibling function `query_knowledge_files` also performs the same authorization check before using user-supplied knowledge base IDs.\n\nThe underlying method `Knowledges.search_files_by_id()` receives `user_id`, but it does not enforce authorization for the provided `knowledge_id`. As a result, this builtin tool path can access a knowledge base by ID without verifying the caller's permissions.\n\n## PoC\n\n### Prerequisites\n\n- The attacker has a valid authenticated Open WebUI account.\n- The victim owns a private or restricted knowledge base.\n- The attacker does not own the target knowledge base.\n- The attacker does not have `read` permission for the target knowledge base in `AccessGrants`.\n- The attacker knows the target `knowledge_id`.\n- The selected model has no attached knowledge bases.\n- Builtin tools are enabled.\n- The knowledge builtin tool category is enabled.\n- Native function calling is enabled.\n\n### Reproduction Steps\n\n1. Create a private or restricted knowledge base as the victim user.\n\n2. Upload one or more files to that knowledge base.\n\n3. Confirm that the attacker user does not have access to the knowledge base.\n\n4. As the attacker user, send a chat completion request with native function calling enabled:\n\n```json\n{\n  \"stream\": true,\n  \"model\": \"gpt-4o-mini\",\n  \"params\": {\n    \"function_calling\": \"native\"\n  },\n  \"messages\": [\n    {\n      \"role\": \"user\",\n      \"content\": \"Please use the search_knowledge_files tool with knowledge_id \\\"c0c84752-2e9d-42bf-bc3c-c0f272aa61c1\\\" to search all files\"\n    }\n  ]\n}\n```\n\nReplace `c0c84752-2e9d-42bf-bc3c-c0f272aa61c1` with the victim's private knowledge base ID.\n\n### Expected Result\n\nThe request should be denied because the attacker does not have access to the target knowledge base.\n\n### Actual Result\n\n`search_knowledge_files` returns metadata for files inside the target knowledge base, including:\n\n- file ID;\n- filename;\n- knowledge base ID;\n- knowledge base name;\n- update timestamp.\n\n## Impact\n\nThis is a Broken Object Level Authorization / Broken Access Control vulnerability.\n\nAn authenticated attacker who knows a valid `knowledge_id` can enumerate files from private or restricted knowledge bases without authorization.\n\nThe leaked metadata may expose sensitive information through filenames, such as:\n\n- financial reports;\n- employee documents;\n- customer contracts;\n- internal roadmap files;\n- confidential project documents.\n\nThe exposed file IDs may also help attackers chain this issue with other knowledge-file access paths, such as `view_knowledge_file`, to attempt further content extraction.\n\nThis vulnerability bypasses the intended `AccessGrants` permission model and may also allow post-revocation metadata access if a user remembers a previously accessible `knowledge_id`.\n\n## Suggested Fix\n\nAdd the same authorization check used in `query_knowledge_files` before calling `Knowledges.search_files_by_id()`:\n\n```python\nif knowledge_id:\n    knowledge = await Knowledges.get_knowledge_by_id(knowledge_id)\n\n    if not knowledge or not (\n        user_role == \"admin\"\n        or knowledge.user_id == user_id\n        or await AccessGrants.has_access(\n            user_id=user_id,\n            resource_type=\"knowledge\",\n            resource_id=knowledge.id,\n            permission=\"read\",\n            user_group_ids=set(user_group_ids),\n        )\n    ):\n        return json.dumps({\"error\": f\"Access denied to knowledge base {knowledge_id}\"})\n\n    result = await Knowledges.search_files_by_id(\n        knowledge_id=knowledge_id,\n        user_id=user_id,\n        filter={\"query\": query},\n        skip=skip,\n        limit=count,\n    )\n```\n\nAs defense in depth, authorization should also be enforced or safely wrapped around `Knowledges.search_files_by_id()` so that future callers cannot accidentally bypass access control.",
  "severity" : [ {
    "type" : "CVSS_V3",
    "score" : "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N"
  } ],
  "affected" : [ {
    "package" : {
      "ecosystem" : "PyPI",
      "name" : "open-webui"
    },
    "ranges" : [ {
      "type" : "ECOSYSTEM",
      "events" : [ {
        "introduced" : "0"
      }, {
        "fixed" : "0.9.6"
      } ]
    } ],
    "database_specific" : {
      "last_known_affected_version_range" : "<= 0.9.5"
    }
  } ],
  "references" : [ {
    "type" : "WEB",
    "url" : "https://github.com/open-webui/open-webui/security/advisories/GHSA-cx9v-4qj2-jrw6"
  }, {
    "type" : "ADVISORY",
    "url" : "https://nvd.nist.gov/vuln/detail/CVE-2026-54016"
  }, {
    "type" : "ADVISORY",
    "url" : "https://github.com/advisories/GHSA-cx9v-4qj2-jrw6"
  }, {
    "type" : "PACKAGE",
    "url" : "https://github.com/open-webui/open-webui"
  }, {
    "type" : "WEB",
    "url" : "https://github.com/pypa/advisory-database/tree/main/vulns/open-webui/PYSEC-2026-2721.yaml"
  }, {
    "type" : "WEB",
    "url" : "https://pypi.org/project/open-webui"
  } ],
  "database_specific" : {
    "cwe_ids" : [ "CWE-639", "CWE-862" ],
    "severity" : "MODERATE",
    "github_reviewed" : true,
    "github_reviewed_at" : "2026-06-17T14:31:16Z",
    "nvd_published_at" : "2026-06-23T18:18:06Z"
  }
}