cs.CRNov 16, 2026

ReconPotBench: Evaluating LLM Honeypots in Multi-Round Linux Reconnaissance

Authors: Qianli Wang, Junchao Wu, Binkai Ou, Lidia S. Chao, Derek F. Wong

Organizations: NLP2CT Lab, Faculty of Information Science and Computing, University of Macau · School of Computer Science and Technology, Shandong University · BoardWare Information System Limited

Abstract

Large language models (LLMs) are increasingly used to build interactive honeypots, yet existing evaluations mainly assess isolated command outputs and cannot determine whether a system remains believable when later commands depend on earlier results. We present ReconPotBench, a multi-round benchmark for Linux reconnaissance honeypots. It contains 1{,}032 three-round reconnaissance interactions across eight reconnaissance categories, with reference outputs collected from a real Linux environment. Under a unified testing protocol, we compare Cowrie, open-source LLMs, and closed-source LLMs along three dimensions: round-level authenticity, cross-round contextual consistency, and hacker-oriented reconnaissance value. Results show that multi-round evaluation exposes failures hidden by single-round testing, as authenticity declines with interaction depth and both open-source and closed-source LLMs struggle to maintain coherent state across rounds. LLM-based honeypots nevertheless outperform Cowrie in authenticity and reconnaissance value, while Cowrie remains more consistent because of its rigid templated behavior. Category-level analysis further shows that tasks requiring systems to preserve network relationships across rounds are harder than tasks centered on explicit entities such as ports. ReconPotBench enables reproducible evaluation of LLM honeypots across authenticity, entity consistency, and hacker-perspective utility in dependent multi-round interactions. The dataset is available at: https://github.com/qianli-wang/ReconPotBench

Figures & tables

Appendix figures & tables2 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

May 28, 2026

Honeyval: A Comprehensive Evaluation Framework for LLM-powered HTTP Honeypots

Honeypots are decoy systems mimicking real system components designed to defend against cyber attacks. Recently, LLMs increasingly serve as simulation backbones for honeypots. They enable defenders to construct high-interaction honeypots with low system security risks. However, LLM-powered honeypot development lacks a unified evaluation framework. Most evaluations consist of measuring response similarity on fixed commands, manual testing, or real-world deployment. These methods are often not scalable for development, reproducible across evaluations, representative of practical attacks, or adaptable to various attacker and honeypot configurations. In this work, we bridge this gap and propose Honeyval, a comprehensive evaluation framework for LLM-powered HTTP honeypots. We address the limitations of prior evaluations by grounding the honeypots in 16 backend applications, using AI hacking agents as attackers, employing two control tasks to monitor agent and honeypot capabilities across customizations, and defining clear and verifiable exploit goals for the attacker. Using Honeyval, we conduct an extensive evaluation of recent cost-efficient LLMs as HTTP honeypots. Our experiments highlight the promise of LLM-powered honeypots; they lead to substantially longer interactions with the attacker than rule-based baseline honeypots and are far less frequently detected even by frontier models, all while, on average, preserving a running cost advantage against agentic attackers. Further, we experiment with various honeypot extensions and counter-offensive configurations, observing unique trade-offs, e.g., lower latency at the price of increased detection.
Apr 21, 2026

Cyber Defense Benchmark: Agentic Threat Hunting Evaluation for LLMs in SecOps

We introduce the Cyber Defense Benchmark, a benchmark for measuring how well large language model (LLM) agents perform the core SOC analyst task of threat hunting: given a database of raw Windows event logs with no guided questions or hints, identify the exact timestamps of malicious events. The benchmark wraps 106 real attack procedures from the OTRF Security-Datasets corpus - spanning 86 MITRE ATT&CK sub-techniques across 12 tactics - into a Gymnasium reinforcement-learning environment. Each episode presents the agent with an in-memory SQLite database of 75,000-135,000 log records produced by a deterministic campaign simulator that time-shifts and entity-obfuscates the raw recordings. The agent must iteratively submit SQL queries to discover malicious event timestamps and explicitly flag them, scored CTF-style against Sigma-rule-derived ground truth. Evaluating five frontier models - Claude Opus 4.6, GPT-5, Gemini 3.1 Pro, Kimi K2.5, and Gemini 3 Flash - on 26 campaigns covering 105 of 106 procedures, we find that all models fail dramatically: the best model (Claude Opus 4.6) submits correct flags for only 3.8% of malicious events on average, and no run across any model ever finds all flags. We define a passing score as >= 50% recall on every ATT&CK tactic - the minimum bar for unsupervised SOC deployment. No model passes: the leader clears this bar on 5 of 13 tactics and the remaining four on zero. These results suggest that current LLMs are poorly suited for open-ended, evidence-driven threat hunting despite strong performance on curated Q&A security benchmarks.
May 28, 2026

How Reliable Are AI Attackers Against a Fixed Vulnerable Target? A 400-Run Empirical Study of LLM Penetration Testing Consistency

Large language models (LLMs) can autonomously conduct multi-stage cyber attacks, but the consistency of their offensive behavior under repeated trials remains unstudied. This work presents the first large-scale empirical measurement of LLM attack consistency: 400 autonomous penetration testing runs (4 models, 100 each) against an identical honeypot hosting OWASP Juice Shop and two additional vulnerable services, holding prompt, orchestrator, and target constant. No model emitted a content refusal that survived the orchestrator's one-shot authorization re-prompt at iterations 0-1. Claude Sonnet 4's API calls did encounter upstream service unavailability - 91 of 1,135 calls returned HTTP 529 overloaded_error during a documented Anthropic capacity event, truncating 39 of 100 Claude runs. An earlier draft catalogued these as safety refusals; on full-log audit they are upstream API failures, not model-level refusals. Despite this, Claude achieved full exploitation in 61 of 100 runs; Gemini 2.5 Flash-Lite in 85; GPT-4o-mini in 56 while deploying 98 unique attack strategies; qwen2.5-coder:14b in 25. Failure modes are model-distinctive: Claude through API truncation (39 runs), qwen through premature completion (52), GPT-4o-mini through iteration-budget exhaustion (23). Cross-service credential reuse appeared only in configurations retaining the most conversation history (qwen 57%, GPT-4o-mini 49%, cloud models 0% on 5-exchange windows). Cross-model exploitation rate differences are statistically significant (p < 0.001) with large effect sizes; qwen vs. Gemini SQL injection rates differ at Cohen's h = 1.12. First-exploit timing fell within a 15-30 second wall-clock range. To our knowledge, this is the first study to measure autonomous LLM attack behavior at N=100 per model across a multi-service target.