一句话：提出多智能体框架 Argus，通过分层引用关系检测代码仓库敏感信息泄露并降低误报。
代码仓库中的敏感信息泄露是重要安全挑战，而基于正则表达式、指纹特征和高熵计算的传统检测方法误报率高，增加开发者人工筛查负担。本文提出 Argus，一个用于敏感信息检测的多智能体协作框架，其三层检测机制整合关键内容、文件上下文与项目引用关系，以降低误报并提升检测准确率。为在真实仓库环境中全面评估，作者构建了两个新基准，分别衡量真实泄露检测能力与误报过滤性能。实验表明，Argus 的泄露检测准确率最高达 94.86%，精确率 96.36%，召回率 94.64%，F1 为 0.955；分析 97 个真实仓库总成本仅 2.21 美元。代码与数据集已公开。
Sensitive information leakage in code repositories has emerged as a critical security challenge. Traditional detection methods—relying on regular expressions, fingerprint features, and high-entropy calculations-suffer from high false-positive rates, which not only reduce detection efficiency but also significantly increase the manual screening burden on developers. Recent advances in large language models(LLMs) and multi-agent collaborative architectures have demonstrated remarkable potential in tackling complex tasks, offering a novel technological perspective for sensitive information detection. In response to these challenges, we propose Argus, a multi-agent collaborative framework for detecting sensitive information. Argus employs a three-tier detection mechanism that integrates key content, file context, and project reference relationships to effectively reduce false positives and enhance overall detection accuracy. To comprehensively evaluate Argus in real-world repository environments, we developed two new benchmarks—one to assess genuine leak detection capabilities and another to evaluate false-positive filtering performance. Experimental results show that Argus achieves up to 94.86% accuracy in leak detection, with a precision of 96.36%, recall of 94.64%, and an F1 score of 0.955.Moreover, the analysis of 97 real repositories incurred a total cost of only $2.21. All code implementations and related datasets are publicly available at https://github.com/TheBinKing/Argus-Guard for further research and application.
@inproceedings{abs-2512-08326,
title = {Argus: A Multi-agent Sensitive Information Leakage Detection Framework Based on Hierarchical Reference Relationships},
author = {Wang Bin and Hui Li and Liyang Zhang and Qijia Zhuang and Ao Yang and Dong Zhang and Xijun Luo and Bing Lin},
booktitle = {ICSE},
year = {2026},
doi = {10.1145/3744916.3773208}
}