Resilient Anomaly and Intrusion Detection in Critical Cyber-physical Systems

关键网络物理系统在可靠性、性能和安全性方面具有独特的操作要求。随着这些系统变得更加互连，并且暴露在气隙之外，它们引入了比传统的关键系统安全方法更大的攻击面。为了确保操作的连续性和正确性，必须保护关键系统免受这些新出现的威胁;此外，网络物理威胁形势正在迅速发展，这已成为一个紧迫的问题。异常检测方法为当前的挑战提供了有前途的解决方案，但已发布的端到端方法很少，而且往往已经过时，而现有的构建模块在一个或多个关键的实用性措施方面存在缺陷。该项目旨在开发强大的异常检测原语，并将这些原语组合到更高级别的架构中，这些架构在现实世界的任务中表现良好，并且可以很容易地应用于现实世界的用例，从而弥合理论和实践。

Critical cyber-physical systems exhibit unique operational requirements in terms of reliability, performance, and security. As these systems become more interconnected and exposed beyond the airgap, they introduce a much larger attack surface than traditional approaches to critical system security are designed to accommodate. To ensure continuity and correctness of operation, critical systems must be protected against these emerging threats; moreover, the cyber-physical threat landscape is evolving at such a rapid pace that this has become a pressing issue. Anomaly detection methods present promising solutions to the challenge at hand, but published end-to-end approaches are sparse and often outdated, while existing building blocks are deficient in one or more key measures of practical applicability. This project aims to develop robust anomaly detection primitives and compose these primitives into higher-level architectures that measurably perform well on real-world tasks and can be readily applied to real-world use cases, thereby bridging theory and practice.