scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-08 03:08:45 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 115

docs: add Supply Chain Security Audit guide with community scanner link

Browse source
This commit is contained in:
Atlas Agent 2026-03-27 00:17:47 +08:00
parent 25feae9f0f
commit 55ea662d45
1 changed files with 52 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

52
docs/my-website/docs/security_audit.md Normal file
View file

@ -0,0 +1,52 @@
---
title: Supply Chain Security Audit
---
# Supply Chain Security Audit
In light of the March 2026 supply chain incident, LiteLLM is committed to providing tools and guidance to help developers secure their environments.
## Local Environment Audit
If you suspect your local or server environment may have been exposed to compromised versions of LiteLLM (v1.82.7 or v1.82.8), you can use the following community-contributed tools to audit your filesystem.
### Agent Safe Check (Fingerprint Scanner)
`agent-safe-check` is a lightweight Python script designed to scan your `site-packages` for malicious `.pth` files and known attack fingerpints associated with the incident.
**Key Features:**
- **Fingerprint Matching**: Compares files against known malicious SHA256 hashes.
- **Malicious .pth Detection**: Identifies hidden execution triggers like `litellm_init.pth`.
**Usage:**
```bash
# Clone and run the auditor
git clone https://github.com/sayanget/lite-guard.git
python3 lite-guard/scan.py
```
**What it checks:**
1. Presence of `litellm_init.pth` in any directory in your `sys.path`.
2. File hash matching for the reported malicious payload (`ceNa7wMJnNHy1kRnNCcwJaFjWX3pORLfMh7xGL8TUjg`).
---
## CI/CD Security Best Practices
To prevent future supply chain attacks, we recommend the following configurations for your pipelines:
### 1. Pin Dependencies
Always use SHA hashes or specific versions for GitHub Actions and pinned versions in `requirements.txt`.
### 2. Use Trusted Publishers (OIDC)
If you are publishing your own forks or internal packages, migrate to [PyPI Trusted Publishers](https://docs.pypi.org/trusted-publishers/) to eliminate the need for long-lived API tokens.
### 3. Network Egress Filtering
Restrict your AI Gateway's outbound traffic to known providers. The March 2026 attack attempted to exfiltrate data to a non-official domain `models.litellm.cloud`. Monitoring for unexpected outbound DNS/HTTP requests is a strong secondary defense.
---
## Official Security Resources
- [March 2026 Incident Report](/blog/security-update-march-2026)
- [Security Policy](https://github.com/BerriAI/litellm/security/policy)