Neurosymbolic Characterization for Reliable Access Control Policy Analysis
探索大型语言模型用于访问控制策略合成与摘要
专题命中 推理评测 :reasoning(abstract);分类 cs.AI
AI总结 研究LLM在访问控制策略自动生成中的有效性,发现非推理LLM仅45.8%生成等价策略,推理LLM达93.7%,并提出基于语义的请求摘要方法辅助策略分析。
Comments Accepted to ISSRE 2026. Major revision and retitling of arXiv:2510.20692v1. Refocuses the paper on reliable neurosymbolic access-control policy analysis; updates the PolicySummarizer method, multi-cloud evaluation, and user-study results. 13 pages, 6 figures. Corrected arXiv title metadata to match the accepted ISSRE version. No substantive content changes from the previous version