Collaborative Research: CISE-MSI: RPEP: SaTC: HBCU Artificial Intelligence and Cybersecurity (AI-CyS) Research Partnership

该奖项的全部或部分资金来自《2021年美国救援计划法案》(公法117-2)。该项目汇集了来自七所历史上的黑人学院和大学(HBCU)和三个国家研究实验室(NRL)的研究人员，以在人工智能(AI)和网络安全的交叉点上发展AI-CyS研究伙伴关系。网络安全漏洞的规模和速度正在增长，使人类主动应对威胁的能力变得紧张；因此，开发人工智能和机器学习(ML)技术以支持预测和检测能力是一个重要的研究领域。参与AI-CyS的HBCU已经拥有或正在迅速发展网络安全研究能力，这种能力将通过增加学生和教职员工的参与和培训来进一步发展。为了开发这一潜力，项目组将利用其现有的研究活动和合作，深化与其他HBCU和NRL的关系。NRL将提供额外的研究资源和指导，包括定期远程会议和实地访问NRL，围绕共同感兴趣的项目，以及在NRL进行学生实习和教师访问的机会。该团队还将努力扩大伙伴关系的影响，方法是：(A)主办年度研究会议，将合作伙伴中的研究人员聚集在一起；(B)在现有伙伴关系的基础上增加额外的研究项目、HBCU和具有互补研究和教育兴趣的国家实验室合作伙伴；以及(C)开发跨大学课程和指导计划，以培训HBCU的学生成为未来网络安全领域的研究和劳动力领导者。这些努力将共同推动网络安全研究、伙伴机构的研究能力以及HBCU学生的研究和教育机会，这些学生往往是计算机领域代表性不足的群体的成员。该伙伴关系将围绕五个种子研究项目组织，以最大限度地提高现有能力和合作。第一个将使用强化学习来改进当前自主网络地图软件代理的目标选择。第二种方法涉及分析网络跟踪路由数据，以便更好地了解和解决其映射路径能力的限制，以便更好地检测和分类网络异常。第三个重点是开发然后防御针对计算机视觉算法的对抗性攻击，在这些攻击中，攻击者向对象添加视觉补丁，以愚弄对象检测和分类工具。第四部分将分析现有的生成“深度假”视频的工具和算法，以开发近实时检测伪造视频的方法，并将其应用于监控和身份验证任务。第五个将扩展概率序列模型，以在物联网设备的背景下在网络堆栈的多个级别上开发威胁检测器。除了这些具体的种子研究项目外，项目组还将分析其活动，为研究能力建设工作制定基于证据的最佳做法。这些努力将包括开发机制以扩大当前的合作和培育资源，在HBCU成员之间分享专业知识，并主办年度研究会议，允许来自现有和潜在合作伙伴HBCU的教职员工和学生展示他们的网络安全研究，并建立与其他研究人员和资源的联系。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势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).This project brings together researchers from seven Historically Black Colleges and Universities (HBCUs) and three National Research Laboratories (NRLs) to develop the AI-CyS research partnership at the intersection of artificial intelligence (AI) and cybersecurity. Cybersecurity vulnerabilities are growing at a scale and speed that strains human capacity to proactively address threats; developing AI and machine learning (ML) techniques to support prediction and detection capabilities is thus an important area of research. The HBCUs involved in AI-CyS either already have or are rapidly developing cybersecurity research capacity that would be further developed by increasing student and faculty involvement and training. To develop that potential, the project team will leverage its existing research activities and collaborations to deepen relationships with both other HBCUs and with the NRLs. The NRLs will provide additional research resources and mentoring through both regular remote meetings and on-site visits to the NRLs around projects of mutual interest, and opportunities for student internships and faculty visits at the NRLs. The team will also work to expand the impact of the partnership by (a) hosting an annual research conference to bring researchers together across the partners, (b) adding additional research projects, HBCUs and national lab partners with complementary research and educational interests to the existing partnership, and (c) developing cross-university curricula and mentoring programs to train HBCU students to be future research and workforce leaders in cybersecurity. Together, these efforts will advance research in cybersecurity, research capacity at the partner institutions, and research and educational opportunities for students at HBCUs, who are often members of underrepresented groups in computing. The partnership will be organized around five seed research projects chosen to maximize existing capacity and collaborations. The first will use reinforcement learning to improve target selection by current autonomous network mapping software agents. The second involves analyzing network traceroute data to better understand and work around limitations of its ability to map paths in order to better detect and classify network anomalies. The third focuses on developing, then defending against, adversarial attacks on computer vision algorithms in which attackers add visual patches to objects to fool object detection and classification tools. The fourth will analyze existing tools and algorithms for generating “deepfake” videos to develop methods to detect forged video in near real-time, with applications to surveillance and authentication tasks. The fifth will extend probabilistic sequential models to develop threat detectors at multiple levels of the network stack in the context of Internet of Things devices. Beyond these specific seed research projects, the project team will also analyze its activities to develop evidence-based best practices for research capacity-building efforts. These efforts will include developing mechanisms to expand current collaborations and foster resources, sharing expertise between HBCU members, and hosting an annual research conference that allows faculty and students from both current and potential partner HBCUs to showcase their cybersecurity research and create connections to other researchers and resources.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.