Document Type


Date of Award


Degree Name

Master of Science (MS)


Computer Science

First Advisor

Dr. Kyong Dong Kang


Cyber-physical system (CPS) has become an integral part of human life, ranging from aircraft to health care systems. The security of these critical components ensures its wider acceptability. Traditionally, many works to secure cyber-physical system (CPS) has been done in the cyber domain, like securing inter/intra CPS communication, securing the exposed software, rebuilding control input derived from sensor data post-digitization, using sensor fusion. All of this security software suffers from a basic attack wherein an attacker compromises the physical/analog sensing system. Researchers have made some progress in mitigating such attacks on physical/analog signals of CPS, the current state of the art methodology proposed in PyCRA uses temporal random signals for physical challenge-response authentication. Though this approach immensely enhances the capability of identifying the sensor attacks, it fails to provide any recovery mechanism to the system. Recent work like Dutta et al., 2017 tries to address this by introducing recursive least squares (RLS) based recovery mechanisms over PyCRA. Although these systems provide some recovery in trivial scenarios, they fail during longer attacks and also result in loss of control because of longer/frequent random no-signal periods. Which could be catastrophic in real-time systems. This work presents Spatio-Temporal Challenge-Response (STCR), an authentication scheme designed to protect active sensing systems against physical attacks occurring in the analog domain. This system utilizes multiple beam-forming and provides physical challenge-response authentication (CRA) in both spatial and temporal domain. Thus providing a much more resilient authentication mechanism that not only detects malicious attacks, but also provides recovery from them. We demonstrate the resilience and effectiveness of STCR over the state of the art in detecting and mitigating attacks through several experiments using a car following (CF) model. This model deploys CPS in the follower car to sense the lead car’s relative position and maintain a safe distance by manipulating acceleration.