Collaborative Research: SaTC: CORE: Medium: Audacity of Exploration: Toward Automated Discovery of Security Flaws in Networked Systems through Intelligent Documentation Analysis

网络系统的规范、开发人员指南和其他文档（例如，互联网应用、运营商网络）描述了这些系统是如何设计、使用和操作的。 这些文档是了解这些系统中安全漏洞的重要来源，由于难以分析其不精确，复杂和模糊的内容，因此尚未充分利用。Project Audacity（Automated Documentation Analysis for seCurity）旨在解决安全漏洞发现和修复的挑战。其创新之处在于开发创新技术，以实现自动化文档分析，从而实现安全保护。 该项目更广泛的意义和重要性包括将技术转让给工业界，使项目中代表性不足的群体的成员参与，并通过K9-12外联和社区服务传播成果。该项目的重点是通过自动恢复与安全相关的信息（例如，模型、安全属性）和令人困惑的描述（例如，不一致的语句）以评估它们的安全含义（例如，验证系统设计，验证系统实施的预测弱点）。该目的通过基于机器学习和自然语言处理的新颖技术来服务，用于分析不同类型的文档，诸如用于支付、单点登录和用于第三代合作伙伴计划或3GPP的文档。这种技术的示例包括用于找到与安全要求相关的语句的情感分析，以及比较关于类似的安全关键操作的不同语句以捕获不一致性的相似性和差异性分析。 此外，该项目还通过比较不同服务的文档和3GPP生态系统来研究新兴技术，例如服务联合，并分析其公共文本数据以进行风险测量，识别和缓解。 这项工作是对项目分析的补充，有助于提高网络系统的安全质量，有助于建立更好的程序和生态系统，使安全关键文档更精确、更一致、更不易出错。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Specifications, developer guides and other documentations of networked systems (e.g., Internet applications, carrier networks) describe how these systems are designed, used and operate. These documentations are important sources for understanding security weaknesses in these systems and have not been fully leveraged due to the difficulty in analyzing their imprecise, convoluted and ambiguous content. Project Audacity (AUtomated Documentation Analysis for seCurITY) aims at addressing the challenge for security weakness discovery and remedy. Its novelties are the development of innovative technologies to enable automated document analysis for security protection. The project’s broader significance and importance include transferring the technologies to industry, involving members from under-represented groups in the project and disseminating outcomes through K9-12 outreach and community services. The project focuses on mitigating security risks of both design flaws and implementation vulnerabilities in networked systems, through automatically recovering security-related information (e.g., models, security properties) and confusing descriptions (e.g., inconsistent statements) from documentations to evaluate their security implications (e.g., verification of system designs, validation of predicted weaknesses on system implementations). This purpose is served by novel techniques based upon machine learning and natural language processing for analyzing different types of documentations, such as those for payment, single-sign-on, and for the 3rd Generation Partnership Project or 3GPP. Examples of such techniques include sentiment analysis for finding the statements related to security requirements and a similarity and differential analysis that compares different statements about similar security-critical operations to capture inconsistency. Furthermore, the project studies emerging techniques such as service syndication through comparing the documentations of different services and the 3GPP ecosystem from analyzing its public text data for risk measurement, identification and mitigation. This work complements program analysis to help enhance the security quality of networked systems, contributing to a better procedure and ecosystem that make security-critical documentations more precise, more consistent and less error-prone.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.