HoneyRoute: Honeypot-Model Routing for Adversarial LLM Serving
HoneyRoute detects malicious LLM serving requests and diverts them to a honeypot model, reaching F1 0.911 with 38 ms median added latency.
HoneyRoute is an inference-serving layer pairing a streaming router (a frozen 0.8B embedding backbone with per-domain MLP heads) with a dual-implementation honeypot and an analysis loop that converts trapped interactions into attacker fingerprints for router retraining. On a production trace plus a seven-domain attack corpus it matches 96% of a two-tier guard-LLM cascade's F1 at 1/385th of its latency with 0% evasion under 13 adversarial transformations. Diverting malicious traffic cuts production token consumption under GCG-suffix flooding by 97.8%, and loop training raises detection F1 to 0.933.
- Retraining loop turns trapped attacker interactions into fingerprints
- Replica honeypot agrees with production model on 92.9% of benign holdout requests
- Naive unconditional bait injection collapses to 7.6% agreement; camouflaged selective injection recovers to 88.9%
- Correction head cuts misrouting of legitimate security research 9x
- Products
- HoneyRoute
- AI models
- HoneyRoute
From the original article
1 min read at arxiv.orgWe introduce HoneyRoute, an inference-serving layer that detects whether an incoming request is malicious and, if so, routes it to a dedicated honeypot model, shielding production while the adversary's interaction is continuously harvested for intelligence.
By Han Jin · 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.