Easily map the attack path and prioritize which CVEs are a threat to your organization
Miggo AIGenerate vendor-ready rules for the observed attack patterns, plus reasoning and safe deployment guidance
Miggo AIRoot Cause AnalysisThe vulnerability is an Insecure Direct Object Reference (IDOR) in the openai-assistants-vector-store endpoints. The core issue is that the application was not verifying that the credential ID provided in the request belongs to the user's active workspace. This allowed an attacker who knows the credential ID of another workspace to use it to perform operations (create, read, update, delete) on OpenAI assistant vector stores, leading to cross-workspace data access and unauthorized resource usage.
The analysis of the patch commit d81483b70c997ddf981acc9c49fbd9a02fa345cd confirms this. The changes in packages/server/src/controllers/openai-assistants-vector-store/index.ts show that a workspaceId check was added to all the functions handling the vector store operations: getAssistantVectorStore, listAssistantVectorStore, createAssistantVectorStore, updateAssistantVectorStore, and deleteAssistantVectorStore. Before the patch, these functions would call the corresponding service function with just the credential ID, but now they pass the workspaceId as well, allowing the service layer to properly scope the database queries and prevent the IDOR.
getAssistantVectorStorepackages/server/src/controllers/openai-assistants-vector-store/index.ts
listAssistantVectorStorepackages/server/src/controllers/openai-assistants-vector-store/index.ts
createAssistantVectorStorepackages/server/src/controllers/openai-assistants-vector-store/index.ts
updateAssistantVectorStorepackages/server/src/controllers/openai-assistants-vector-store/index.ts
deleteAssistantVectorStorepackages/server/src/controllers/openai-assistants-vector-store/index.ts
| Package Name | Ecosystem | Vulnerable Versions | First Patched Version |
|---|---|---|---|
| flowise | npm | <= 3.1.2 | 3.1.3 |