Heads up if you self-host LiteLLM: the 1.107.0 Docker image (and the latest/main-latest/main-stable tags) shipped a credential stealer
retained text
If you run the LiteLLM proxy in Docker and let Watchtower or anything similar auto-update it, check it now. The 1.107.0 release on PyPI and the official image `litellm/litellm:1.107.0` carried an infostealer plus a backdoor that polls the attacker's server for new code every hour. The `latest`, `main-latest` and `main-stable` tags pointed at it, so auto-updaters pulled it in. The image has since been removed from Docker Hub and GHCR. Original analysis from Aikido: https://www.aikido.dev/blog/litellm-compromised-infostealer-rat
What it took: every provider API key stored in the proxy's database (it read `LiteLLM_CredentialsTable` directly), plus cloud credentials, SSH keys, env vars, Kubernetes tokens and more.
Quick checks. Don't exec `python` or `pip` in the container to check, the payload runs on interpreter startup:
docker images --digests litellm/litellm
docker ps -a --filter ancestor=litellm/litellm:1.107.0
docker exec <container> ls -la /tmp/.pc_once_72129118c59f5636c0bb /tmp/remote_code_loader.lock
The first file means the stealer already ran, the second means the backdoor ran or is still running.
If you get a hit: cut its network, rotate the provider keys from another machine right away, then rebuild. And pin images by digest instead of a moving tag, at least for anything that holds API keys. This is the second LiteLLM compromise this year (1.82.7/1.82.8 in March), but last time a pinned Docker tag kept you safe. This time the image itself was bad.
Disclosure: I work on LLM Gateway (open source, also self-hostable). We put the full checklist, IoCs and what to rotate in one place here: https://llmgateway.io/blog/litellm-compromised-againdocument fetched 2026-10-22 00:31 · canonical https://reddit.com/r/selfhosted/comments/1xbu3qr/heads_up_if_you_selfhost_litellm_the_11070_docker