AI / ML CVE-2026-30762

LightRAG v1.4.10 with AUTH_ACCOUNTS configured but TOKEN_SECRET environment variable not set. Unauthenticated remote attacker forges JWT tokens using publicly known default secret 'lightrag-jwt-default-secret' to access protected endpoints. No user interaction required.

Real-world risk is HIGH for exposed installations despite moderate CVSS scoring. The CVSS vector AV:N/AC:L/PR:N/UI:N indicates network-accessible exploitation with low complexity and no authentication required, making this trivially exploitable. However, the confidentiality-only impact (C:H/I:N/A:N) limits the score to 7.5. The attack requires the specific misconfiguration where AUTH_ACCOUNTS is set but TOKEN_SECRET is not, which may be common in quick deployments or development environments exposed to networks. No EPSS score is available, and this CVE is not in CISA KEV, indicating no confirmed active exploitation. However, the barrier to exploitation is extremely low post-disclosure since the default secret is now public knowledge and JWT forging requires only basic Python skills. The GitHub Security Advisory publication significantly increases exploitation likelihood. Organizations using LightRAG with authentication enabled should treat this as an immediate priority regardless of the moderate CVSS score, as the gap between theoretical and practical exploitability is minimal for internet-facing instances.

An attacker discovers a publicly accessible LightRAG instance at api.example.com running with AUTH_ACCOUNTS configured. Using the publicly disclosed default secret 'lightrag-jwt-default-secret', the attacker runs a simple Python script with PyJWT to forge a token containing admin-level claims: jwt.encode({'sub': 'admin', 'role': 'admin', 'exp': time.time() + 3600}, 'lightrag-jwt-default-secret', algorithm='HS256'). They submit API requests with the Authorization: Bearer header containing this forged token, successfully authenticating as an administrator without knowing any legitimate credentials. The attacker can now access protected RAG query endpoints, exfiltrate indexed document data, or manipulate knowledge graph contents. The attack requires only network access and basic scripting knowledge, with no complex exploitation techniques or privilege escalation needed.