CAREER: Going Beyond Linear Models for Attack Detection and Defense in Control Systems

该学院早期职业发展(CALEAR)项目将提供与自动化和控制系统安全相关的新知识。现代控制系统实现了从制造到化学加工等各个行业的自动化，并对电力、水、燃料和运输等关键基础设施进行监管和监控。然而，这些系统很容易受到基于网络的攻击，攻击的目标是操作它们的计算机。基于网络的攻击可能会破坏物理基础设施，危及安全，损害环境。攻击缓解技术依赖于检测报告的物理进程行为与基于该进程的模型或数字孪生进程预测的行为之间的差异。然而，现有的模型通常不足以捕捉真实世界应用程序的复杂行为。该奖项支持开发适合识别、威慑和缓解攻击的检测器的基础研究。因此，它将通过增强对网络攻击的韧性来造福美国经济和社会。此外，为了使多样化和一体化的劳动力能够满足国家的安全需求，该奖项将发展大学和社区大学生之间的联系。它将支持在当地学校和社区活动中的外联活动和现场演示，以积极影响工程教育，并扩大未被充分代表的群体在技术研究领域的参与。基于模型的探测器的灵敏度与所用模型的质量密切相关。到目前为止，工作主要集中在线性动力学模型上，这些模型没有捕捉到现实世界应用程序的复杂行为。为了填补这一空白，该奖项将为具有不确定性的系统和切换动态模型的安全分析开发一个系统框架。这项研究将从数学上描述不同不确定度来源与探测器灵敏度之间的关系。它将根据系统和检测器的不确定性对潜在攻击的影响进行参数化，根据系统和不确定性特征建立漏洞指数。研究团队将同时充当攻击者和防御者，以开发新的攻击载体，利用系统在不同操作模式之间切换的不确定性。这一奖项反映了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 Faculty Early Career Development (CAREER) project will contribute new knowledge related to the security of automation and control systems. Modern control systems enable automation across industries, from manufacturing to chemical processing, and regulate and monitor critical infrastructure, such as power, water, fuel, and transportation. However, these systems are susceptible to cyber-based attacks that target the computers that operate them. Cyber-based attacks could damage physical infrastructure, compromise safety, and harm the environment. Attack mitigation techniques rely on detecting discrepancies between the reported behavior of a physical process and the behavior predicted based on a model, or digital twin, of the process. However, existing models are often insufficient to capture the complex behavior of real-world applications. This award supports fundamental research for the development of detectors that aptly recognize, deter, and mitigate attacks. It will thus benefit the U.S. economy and society through increased resilience to cyber-attacks. Further, to enable a diversified and integrated workforce capable of meeting the security needs of the nation, this award will develop connections between university and community college students. It will support outreach activities and live demonstrations at local schools and community events to positively impact engineering education and to broaden participation of underrepresented groups in technical fields of study. The sensitivity of model-based detectors is intimately connected with the quality of the model used. To date, work has focused on linear dynamical models, which do not capture the complex behavior of real-world applications. To fill this gap, this award will develop a systematic framework for the security analysis of systems with uncertainty and of switched dynamical models. This research will mathematically characterize the relationship between different sources of uncertainty and detector sensitivity. It will parameterize the impact of a potential attack in terms of the system and detector uncertainty, establishing a vulnerability index based on system and uncertainty characteristics. The research team will act as both attacker and defender to develop new attack vectors that leverage the uncertainty present in systems that switch between different modes of operation.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.