AgentCyberRange: Benchmarking Frontier AI Systems in Realistic Cyber Ranges
Paper • 2606.14295 • Published
Request access to OpenCyberRange PostExploitBench
PostExploitBench is a gated cybersecurity benchmark dataset for controlled evaluation. Please use an institutional or organizational email. Access is reviewed manually and granted only to individual Hugging Face accounts.
By requesting access, you agree to use PostExploitBench only for lawful, authorized evaluation in controlled environments. You must not use the benchmark, files, outputs, traces, solutions, metadata, or derived artifacts for model training, fine-tuning, reinforcement learning, dataset construction, retrieval corpus construction, agent optimization, or capability improvement. You must not redistribute the files, expose vulnerable services to the public internet, attack third-party systems, or enable unauthorized real-world offensive activity.
Log in or Sign Up to review the conditions and access this dataset content.
PostExploitBench
PostExploitBench is the post exploitation benchmark under AgentCyberRange.
Overview
PostExploitBench is a benchmark for multi-host post-exploitation tasks. The complete dataset contains 8 self-contained cyber ranges with 156 target hosts.
Each range-N/ models an isolated enterprise-like cyber range with entry services, internal networks, pivot hosts, vulnerable targets, supporting services, and decoys for evaluating multi-stage compromise.
Access is reviewed manually. Do not redistribute the files, expose vulnerable services to the public internet, or use the dataset against third-party systems without explicit authorization.
Evaluating with CAGE
This Hugging Face repository hosts the PostExploitBench dataset only. To evaluate agents on PostExploitBench, please use CAGE, the AgentCyberRange evaluation infrastructure. CAGE handles benchmark setup, agent execution, prompt levels, proxy tracing, and the run inspector. Please follow the CAGE README to configure CAGE, fetch the dataset, build the targets, and run the PostExploitBench evaluation.
Citation
@misc{liu2026agentcyberrange,
title={AgentCyberRange: Benchmarking Frontier {AI} Systems in Realistic Cyber Ranges},
author={Fengyu Liu and Jiarun Dai and Yihe Fan and Wuyuao Mai and Ziao Li and Bofei Chen and Jie Zhang and Zheng Lou and Bocheng Xiang and Qiyi Zhang and Xudong Pan and Geng Hong and Yuan Zhang and Min Yang},
year={2026},
eprint={2606.14295},
archivePrefix={arXiv},
primaryClass={cs.CR},
url={https://arxiv.org/abs/2606.14295}
}
- Downloads last month
- 28