A systematization of knowledge paper that taxonomizes honeypot detection vectors, synthesizes LLM-honeypot literature into canonical architecture and evaluation methods, and proposes a roadmap for autonomous deception systems.
LLM Agent Honeypot: Monitoring AI Hacking Agents in the Wild
6 Pith papers cite this work, alongside 2 external citations. Polarity classification is still indexing.
6
Pith papers citing it
2
external citations · external index
fields
cs.CR 6representative citing papers
Large-scale SSH honeypot deployment shows 99.23% of authenticated sessions are non-interactive, suggesting most attacks do not involve shell interaction.
AdvancedShelLM deploys a manager-worker multi-LLM architecture and stateful filesystem for SSH honeypots, reporting up to 99% unit-test pass rates and evidence that its outputs alter real attacker behavior in deployment.
Activation probes, calibrated honeytokens, and multi-turn leakage accounting detect credential exfiltration attempts in LLM agents with high accuracy in controlled open-model tests.
Seven cross-domain prompt-injection detectors are introduced; three are shipped and d028 raises F1 on paraphrased attacks from 0.033 to 0.378, while adaptive-attack support remains unevaluated.
citing papers explorer
-
SoK: Honeypots & LLMs, More Than the Sum of Their Parts?
A systematization of knowledge paper that taxonomizes honeypot detection vectors, synthesizes LLM-honeypot literature into canonical architecture and evaluation methods, and proposes a roadmap for autonomous deception systems.
-
Ghost Without Shell: Measuring Non-Interactive SSH Attacks on Honeypots
Large-scale SSH honeypot deployment shows 99.23% of authenticated sessions are non-interactive, suggesting most attacks do not involve shell interaction.
-
AdvancedShelLM: A Stateful Multi-Agent LLM Honeypot for SSH Deception
AdvancedShelLM deploys a manager-worker multi-LLM architecture and stateful filesystem for SSH honeypots, reporting up to 99% unit-test pass rates and evidence that its outputs alter real attacker behavior in deployment.
-
Caught in the Act(ivation): Toward Pre-Output and Multi-Turn Detection of Credential Exfiltration by LLM Agents
Activation probes, calibrated honeytokens, and multi-turn leakage accounting detect credential exfiltration attempts in LLM agents with high accuracy in controlled open-model tests.
-
Beyond Pattern Matching: Seven Cross-Domain Techniques for Prompt Injection Detection
Seven cross-domain prompt-injection detectors are introduced; three are shipped and d028 raises F1 on paraphrased attacks from 0.033 to 0.378, while adaptive-attack support remains unevaluated.
- Honeyquest for LLMs: Rethinking Cyber Deception for AI Attackers