FIX CARD
PRODUCT vLLM
SEVERITY HIGH
vLLM API open to the internet
Why it matters
Your model-serving API answers anyone: strangers can list your models and run unlimited inference on your GPU — you pay the compute bill, and they can read anything the model was prompted with through your app.
Fix it: 3 steps
- Restart vLLM with an API key: add
--api-key YOUR_LONG_RANDOM_KEYand require it from clients (Authorization: Bearer ...). - Better: bind to 127.0.0.1 (
--host 127.0.0.1) and front it with an authenticating reverse proxy or keep it on a private network. - Block the port at the firewall:
sudo ufw deny 8000.
Verify it’s fixed
curl -m 5 http://YOUR_SERVER_IP:8000/v1/models # should return 401/403 or time out, not a model list
References
Not sure if your stack is exposed?
Run the free check: usually under a minute, safe read-only probes.
Scan your stack