Robustness to Adversarial Disruptions on Sensing and Communications Infrastructure

Lead PI: Zhiyun Qian 

Communication systems are subject to a variety of adversarial actions that can degrade both the reliability and security of a system. For example, communication links are open to jamming/eavesdropping. Software modules are vulnerable to exploits, which can lead to loss of integrity, disruption or confidentiality and will need to be protected. This thrust aims to develop a situational-aware, decentralized, and robust security framework to address the security challenges in communications. In short, we aim to intelligently differentiate the different types of threats and take the best course of actions accordingly. The work will be divided into three main components: (1) understanding and profiling our own systems, (2) gathering data continuously and inferring the current situation, (3) developing novel countermeasures against various adversarial actions. The solution will be developed taking into account both the physical layer technologies/characteristics and higher layers. Since the communication systems can be complex, dynamic, and distributed, we plan to investigate machine learning based techniques that can quickly and flexibly adjust to the situation. However, we also need to be aware of the adversarial attacks against these ML models. Finally, we also seek to develop advanced techniques that can allow the communication system to operate despite the ongoing threats and attacks.