vLLM
Named by 21 kept items in this workspace.
14 items name it under affected products, 6 under tools and 1 under classification.technologies.
The items that name it 21
OpenAI-compatible servers; multimodal endpoint vulnerable to heap-address info leak (CVE-2026-22778)
AI backend serving infrastructure; disclosures reached 212 vulnerabilities in 2026, ~24% from unauthenticated API endpoints or SSRF.
latest version required, along with a 4-H200s node
serve the model via an OpenAI-compatible API
inference server used to serve the model via an OpenAI-compatible API
Serving framework; compromised instance can violate user-data isolation through shared GPU state, per GIFT paper
OpenAI-compatible local model server T3MP3ST can target for offline operation
OpenAI-compatible LLM server backend supported by OASIS via --provider openai
DoS confirmed, 3/3 SIGSEGV, via PyAV with libavcodec 4.4 on Ubuntu 22.04
Supported local model serving provider for Metis
Local model server exposing an OpenAI-compatible endpoint, supported by Shannon for local/self-hosted models
Inference-server attack surface; default configs frequently ship with no authentication
proxy exposed publicly
AI component with CVE rules added (v4.6.2); example target for AI Infra Scan (scanned via its API endpoint, e.g. http://127.0.0.1:8000)
AI component covered by expanded vuln library (v4.6.2); example target for AI Infra Scan (fingerprinted against 2000+ known CVEs)
Self-hosted inference server; does not enforce message-ordering restrictions by default
In every workspace you can read 126
AI
Security
All technologies in this workspace.