LiteLLM Was Just Compromised on PyPI — Here's How to Detect Supply Chain Attacks

#security #python #javascript #discuss

What Happened

Today on Hacker News: LiteLLM versions 1.82.7 and 1.82.8 on PyPI are compromised.

This is a supply chain attack — malicious code injected into a legitimate package. If you installed these versions, your credentials may be exposed.

This Is Not New

Supply chain attacks happen constantly:

event-stream (2018) — 2M weekly downloads, maintainer handed off to attacker
ua-parser-js (2021) — cryptominer injected into 8M weekly download package
colors + faker (2022) — maintainer self-sabotaged in protest
LiteLLM (2026) — compromised PyPI release

How to Protect Yourself

Here's a practical detection script using only free APIs:

import requests
import json

def check_pypi_package(name):
    """Detect potential supply chain issues in a PyPI package."""
    data = requests.get(f"https://pypi.org/pypi/{name}/json").json()
    info = data.get("info", {})
    releases = data.get("releases", {})

    risks = []

    # Check 1: Recent maintainer email change
    author_email = info.get("author_email", "")

    # Check 2: Very recent release (within hours)
    versions = list(releases.keys())
    if versions:
        latest = versions[-1]
        files = releases[latest]
        if files:
            upload_time = files[0].get("upload_time", "")
            print(f"Latest: {latest} uploaded {upload_time}")

    # Check 3: Source repo mismatch
    project_urls = info.get("project_urls", {}) or {}
    home = info.get("home_page", "")
    if not home and not project_urls:
        risks.append("NO_SOURCE_REPO: no homepage or project URLs")

    # Check 4: Yanked versions (often indicates compromise)
    yanked = [v for v, files in releases.items() 
              if any(f.get("yanked", False) for f in files)]
    if yanked:
        risks.append(f"YANKED_VERSIONS: {yanked}")

    # Check 5: Suspiciously few maintainers
    maintainer = info.get("maintainer", "")
    author = info.get("author", "")
    if not maintainer and not author:
        risks.append("NO_MAINTAINER_INFO")

    return {
        "package": name,
        "version": info.get("version"),
        "risks": risks,
        "risk_level": "HIGH" if len(risks) >= 2 else "MEDIUM" if risks else "LOW"
    }

# Check your dependencies
for pkg in ["litellm", "requests", "flask", "django"]:
    result = check_pypi_package(pkg)
    print(f"\n{result['package']}: {result['risk_level']}")
    for risk in result["risks"]:
        print(f"  ⚠️ {risk}")

npm Version (Same Concept)

def check_npm_package(name):
    """Detect supply chain risks in npm packages."""
    data = requests.get(f"https://registry.npmjs.org/{name}").json()
    versions = list(data.get("versions", {}).keys())
    maintainers = data.get("maintainers", [])

    risks = []

    if len(maintainers) == 1:
        risks.append(f"SINGLE_MAINTAINER: {maintainers[0].get('name', 'unknown')}")

    # Check for maintainer change in latest version
    if len(versions) >= 2:
        latest = data["versions"][versions[-1]]
        prev = data["versions"][versions[-2]]
        if latest.get("_npmUser", {}) != prev.get("_npmUser", {}):
            risks.append("PUBLISHER_CHANGED: different user published latest")

    # Check GitHub advisories
    advisories = requests.get(
        f"https://api.github.com/advisories",
        params={"ecosystem": "npm", "package": name}
    ).json()
    if isinstance(advisories, list) and advisories:
        risks.append(f"KNOWN_CVES: {len(advisories)} advisories")

    return {"package": name, "risks": risks}

Prevention Checklist

✅ Pin exact versions (== not >=)
✅ Use lock files (poetry.lock, package-lock.json)
✅ Enable Dependabot or Renovate (free on GitHub)
✅ Review changelogs before updating major versions
✅ Check maintainer changes on every update
✅ Use pip install --require-hashes in production

Free Tools for Detection

Tool	What It Checks
`pip audit`	Known CVEs in Python deps
`npm audit`	Known CVEs in Node deps
npms.io API	Quality scores
GitHub Advisory DB	CVEs by ecosystem
Socket.dev	Supply chain analysis
PyPI JSON API	Package metadata