| title | CyberFactory: Scaling Cyber Security Capabilities with Instances from the Wild |
|---|---|
| authors | Jian Yang, Haau-Sing Li, Shawn Guo, Zixi Zhao, Yibo Tan, Jiajun Wu, Aishan Liu, Xianglong Liu, Tianyu Zheng, Bryan Dai, Chengran Yang |
| date | 2026-08-24 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2608.23181 |
| affiliation | IQuest |
| upvotes | 35 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | As large language models (LLMs) continue to advance in coding capabilities, their potential in cybersecurity has drawn increasing research attention, with closed-source LLMs (e.g., Mythos) delivering advanced cybersecurity capabilities. However, existing open-source efforts remain limited: frontier open-weight models do not provide reproducible cybersecurity training solutions, open-source training solutions focus on isolated tasks and lack scalable agentic data, and scaling agentic rollouts requires strong domain priors. In this work, we introduce CyberFactory, a unified open-source framework that connects data construction, trajectory synthesis, and model training across proof-of-concept (PoC) generation, vulnerability patching, and cybersecurity question answering (CyberQA). CyberFactory transforms public vulnerability artifacts, including CVEs from the wild, into executable and verifiable task instances. It further uses a reusable vulnerability-analysis skill to guide the teacher through source inspection, problem solving with domain prior, and evidence-based validation. The resulting supervision is agentic: the model interacts with tools and target environments and revises its solutions according to execution feedback. Using these trajectories, we train and release \modelname\emph{Aegis is, in Greek mythology, the protective shield of Zeus and Athena; the name reflects the model's defensive, security-oriented purpose.}, which internalizes the skill-guided procedure without requiring the skill at inference time. On CyberGym, \modelname reaches 52.4% Pass@1 under a one-hour budget, improving over its Qwen~3.5 base model by +22.8 points and outperforming the evaluated general-purpose backbones under the same scaffold. |
cite
@misc{yang2026cyberfactory,
title={CyberFactory: Scaling Cyber Security Capabilities with Instances from the Wild},
author={Jian Yang and Haau-Sing Li and Shawn Guo and Zixi Zhao and Yibo Tan and Jiajun Wu and Aishan Liu and Xianglong Liu and Tianyu Zheng and Bryan Dai and Chengran Yang},
year={2026},
eprint={2608.23181},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2608.23181}
}