AutoGuard:基于强化学习的DevSecOps流水线自愈主动安全层
密码学与安全
2025-12-05 v1 人工智能
机器学习
性能
摘要
当代DevSecOps流水线需要应对在持续集成与持续部署环境中不断演进的安全威胁。现有方法,如基于规则的入侵检测和静态漏洞扫描,无法适应系统的变化,导致响应时间更长且组织面临新兴攻击向量的暴露。针对上述限制,我们将AutoGuard引入DevSecOps生态系统,这是一个由强化学习(RL)驱动的自愈安全框架,旨在主动保护DevSecOps环境。AutoGuard是一个自安全环境,持续观察流水线活动以发现潜在异常,同时主动修复环境。该模型基于随时间动态持续学习的策略进行观察和响应。RL智能体通过基于奖励的学习逐步改进每个动作,旨在提升智能体实时预防、检测和响应安全事件的能力。使用模拟持续集成/持续部署(CI/CD)环境进行的测试表明,AutoGuard成功将威胁检测准确率提升了22%,将事件平均恢复时间(MTTR)降低了38%,并相较于传统方法提升了整体事件韧性。关键词:DevSecOps,强化学习,自愈安全,持续集成,自动威胁缓解。
引用
@article{arxiv.2512.04368,
title = {AutoGuard: A Self-Healing Proactive Security Layer for DevSecOps Pipelines Using Reinforcement Learning},
author = {Praveen Anugula and Avdhesh Kumar Bhardwaj and Navin Chhibber and Rohit Tewari and Sunil Khemka and Piyush Ranjan},
journal= {arXiv preprint arXiv:2512.04368},
year = {2025}
}
备注
Accepted and Presented at 1st IEEE Uttar Pradesh Section Women in Engineering International Conference on Electrical Electronics and Computer Engineering (UPWIECON 2025) organized by NIELIT Dehradun held during 30th 31st October 2025