Executive brief
LightRAG, an open-source knowledge graph library, contains a security flaw that allows unauthorized individuals to bypass its API key protection. When configured in its standard API-key mode, the system uses a publicly known secret key to sign security tokens, allowing anyone to create their own valid credentials. An attacker can use this to read, modify, or delete sensitive documents and knowledge graphs, potentially leading to data loss or unauthorized use of expensive AI services.
Technical details
LightRAG is vulnerable to an authentication bypass when deployed with LIGHTRAG_API_KEY enabled but AUTH_ACCOUNTS disabled. The vulnerability stems from three factors: the use of a hardcoded DEFAULT_TOKEN_SECRET in the source code, an unauthenticated /auth-status endpoint that issues guest JWTs signed with this secret, and a short-circuit in the combined_dependency function that accepts any valid guest token before checking the X-API-Key header. A remote attacker can either obtain a guest token from the server or mint one offline using the known secret to gain full access to the API. This allows for unauthorized document deletion, data ingestion, and knowledge graph manipulation. The issue is resolved in version 1.5.4.
Affected products
- HKUDS lightrag-hku < 1.5.4
Timeline
- 2026-06-24: disclosed: Initial disclosure to HKUDS/LightRAG
- 2026-07-15: advisory: NVD publication date
- 2026-07-20: patched: GitHub Advisory published and fix confirmed in v1.5.4