InferD honeypot answers AI-inference probes without executing models
Original title: InferD: A Honeypot Framework for AI Inference Infrastructure
Static response pools cover transport, protocol, vulnerability-surface and agentic-session deception planes.
InferD is a low-interaction honeypot for public LLM serving stacks, commercial inference APIs, vector databases and agent orchestration gateways. It reproduces model inventories, token formats, tool-call protocols and authorization flows from static pools, so submitted prompts and tool calls are never executed. A provenance plane links a planted credential to evidence of downstream model-mediated processing.
- The stated design principle is deception without inference.
- The paper says commodity scanners index exposed AI services within hours.
- Nothing an adversary submits is executed.
From the original article
1 min read at arxiv.orgLarge language model serving stacks, commercial inference APIs, vector databases, and agent orchestration gateways are now ex- posed on the public internet and indexed by commodity scanners within hours of deployment.
By Hicham Hammouchi · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.