CISE-MSI: DP: SaTC: MSI Research Capacity Building for Artificial Intelligence (AI)-enabled Vulnerability Assessment and Remediation in Cyberinfrastructure

该奖项全部或部分由《2021年美国救援计划法案》（公法117-2）资助。联邦机构和公共云提供商（如Amazon Web Services）已投入数百万美元开发高性能计算资源，以支持高影响力的计算工作流程。用户通过虚拟机（vm）和应用程序容器（例如Docker）等技术访问这些高级资源，这些开源软件通常包含不同严重级别的多个漏洞。尽管漏洞扫描器可以帮助发现这些威胁，但由于扫描器经常在多个虚拟环境中返回数十万个结果，安全从业者很难有效地评估、确定优先级并修复它们。该项目将创建一个新颖的人工智能（AI）支持框架，以(1)检测漏洞并在多个维度的漏洞数据中自动优先考虑易受攻击的vm，以及(2)将从扫描仪中发现的漏洞与已披露的漏洞联系起来，以确定合适的修复策略。这两个项目重点将实现云基础设施的自动化漏洞评估分析。项目成果将通过在学术和行业出版物场所传播知识，以及整合到网络安全和数据分析课程中，为网络安全、深度学习和文本分析的发展做出贡献。此外，这项工作将为两个西班牙裔服务机构中代表性不足的人口提供网络安全和人工智能分析教育和研究经验的独特结合，从而促进下一代网络安全专业人员的发展。这项研究利用了从该团队的nsf资助的合作伙伴组织收集的全面的开源软件和漏洞数据集，以促进设计一个新的支持ai的漏洞评估和补救（AI-VAR）框架，该框架由两个新颖且相互关联的研究重点组成。第一个要点结合并扩展了网络科学、多视图表示学习、自动编码器和注意力机制的原理，以在云基础设施中创建易受攻击的vm组，以进行优先级排序。而第二个要点则利用自监督对比表示学习和转换中的最先进方法，从漏洞扫描描述中捕获文本特征，并将发现的漏洞与包含修复策略的公开漏洞联系起来。该奖项反映了美国国家科学基金会的法定使命，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。

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).Federal agencies and public cloud providers such as Amazon Web Services have invested millions of dollars developing high-performance computing resources to support high-impact computational workflows. Users access these advanced resources through technologies such as virtual machines (VMs) and application containers (e.g., Docker) with open-source software that often contains multiple vulnerabilities at different severity levels. Although vulnerability scanners can help aid in the discovery of these threats, security practitioners struggle to effectively assess, prioritize, and remediate them since scanners often return hundreds of thousands of results across multiple virtual environments. This project will create a novel artificial intelligence (AI) enabled framework to (1) detect vulnerabilities and automatically prioritize vulnerable VMs across multiple dimensions of vulnerability data, and (2) link the discovered vulnerabilities from scanners with disclosed vulnerabilities to identify suitable remediation strategies. The two project thrusts will enable automated vulnerability assessment analytics for cloud infrastructures. Project outcomes will contribute to developments in cybersecurity, deep learning, and text analytics through the dissemination of knowledge at academic and industry publication venues as well as integration into cybersecurity and data analytics curricula. Moreover, this work will enable a unique combination of cybersecurity and AI analytics education and research experience for underrepresented demographics across two Hispanic Serving Institutions, and thereby foster the development of the next generation of cybersecurity professionals.This research leverages comprehensive open-source software and vulnerability datasets collected from the team’s NSF-funded partner organizations to facilitate the design of a novel AI-enabled Vulnerability Assessment and Remediation (AI-VAR) framework that consists of two novel and interconnected research thrusts. The first thrust combines and extends principles from network science, multi-view representation learning, autoencoders, and attention mechanisms to create groups of vulnerable VMs in cloud infrastructures for prioritization, while the second thrust draws upon state-of-the-art methods in self-supervised contrastive representation learning and transformers to capture textual features from vulnerability scan descriptions and link discovered vulnerabilities with disclosed vulnerabilities that contain remediation strategies.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.