Security context
What an agent needs to avoid regressing past fixes and find the next vuln in this repo.
internlm/lmdeploy
main @ 5d4fac9
11
Fixes
4
CVEs
HIGH
Peak severity
45.5%
Coverage
Highlights
Denial of Service: 6 prior fixes. Scrutinize any change in this area.
src/turbomind/models/llama/LlamaBatch.cc: most-fixed (2 issues). Treat as high-risk during review.
5 high-severity fixes in this history; regressions here are high-impact.
Recurring patterns
The bug types that recur here, drawn from past fixes, not open vulnerabilities.
Remote Code Execution: Unsafe model loading practices allowed execution of arbitrary code via malicious model checkpoints. Mitigations required enforcing weights_only=True during serialization calls.
SSRF: Failure to validate IP addresses when fetching remote assets (e.g., images) allowed attackers to access internal networks, bypass isolation boundaries, and scan infrastructure.
Remote Code Execution: Loading untrusted remote weights or Hugging Face configurations with remote code execution enabled by default allowed arbitrary code execution during initialization.