LightLLM through 1.2.0 mounts reinforcement learning control routes on the public HTTP API without authentication checks. Unauthenticated attackers can call endpoints like /pausegeneration, /abortrequest, /flushcache, and /initweightsupdategroup to disrupt inference operations and wedge workers on deployments started with --enablerl.

LightLLM through 1.2.0 fails to validate imageurl and audiourl parameters in multimodal endpoints, allowing unauthenticated attackers to perform server-side request forgery. Attackers can supply arbitrary URLs to fetch internal resources, with vision model processing disclosing content or error responses revealing internal network topology.

LightLLM through 1.2.0 contains a memory exhaustion vulnerability in the NCCL control channel when started with --pdtransmode nccl, allowing unauthenticated attackers to exhaust KV-transfer worker memory. Attackers can call the exposedsetvalue method to store unbounded key-value pairs without size limits, causing the worker process to crash and triggering node failure.

LightLLM through 1.2.0 contains a remote code execution vulnerability in the router profiler service when started with --enableprofiling flag. The service exposes an unauthenticated RPyC server with pickle deserialization enabled, allowing attackers to execute arbitrary code by sending crafted serialized objects to the profiler command queue.