Description
InternLM LMDeploy through 0.17.0 in DistServe prefill/decode disaggregation mode fails to release scheduler sessions because the proxy uses user-facing session IDs instead of internal scheduler keys. Unauthenticated attackers can send completion requests to the proxy endpoint that accumulate unreleased scheduler metadata and memory until the prefill worker is out-of-memory killed.
Affected products
Version statuses are supplied by the assigning organization. Consult the original record for detailed ranges and platform conditions.
InternLM lmdeploy
Default version status: unaffected
0 to ≤ 0.17.0affected
pkg:pypi/lmdeployReferences
- https://github.com/InternLM/lmdeploy/issues/4967 ↗
- https://github.com/InternLM/lmdeploy ↗
- https://github.com/InternLM/lmdeploy/blob/v0.17.0/lmdeploy/pytorch/disagg/conn/engine_conn.py#L93-L99 ↗
- https://www.vulncheck.com/advisories/internlm-lmdeploy-through-0.17.0-memory-exhaustion-via-session-id-mismatch ↗