Secure Motion Claim Verification and Decision-Making for Vehicular Cyber-Physical Systems

Abstract

Vehicular cyber-physical systems (CPSes) that integrate sensing, computing, control, and communication, together into various connected physical objects or infrastructures, have received significant interest in recent years. In a vehicular CPS, participating autonomous vehicles (AVs), which can be ground, aerial, or even marine, depend on reliable and secure vehicle-to-everything (V2X) communications to exchange useful information such as motion states (including positions, velocity, and acceleration), traffic conditions and road emergencies, etc. Acknowledging the correct V2X information can help system participants be aware of the environment and take proper action such as accelerating, brakes, or changing lanes, in order to enhance system safety, efficiency, as well as reduce energy cost. On the other hand, attackers have put numerous attention to contaminating the V2X communication for malicious purposes. Active malicious attackers may delay the V2X messages, or even spread messages containing false motion state information via V2X communication. For example, if one malicious vehicle claims accelerating but in fact slowing down, and meanwhile its following vehicle believes this claim, a severe collision may occur. However, traditional crypto-primitives can only guarantee the authenticity and integrity of these messages, but not the information truthfulness since attackers can modify it at the source. Also, stealthy attackers can send silent aerial vehicles (i.e. drones do not transmit anything.) to break the drone geofencing, aiming at illegally recording sensitive information. Thus, being able to securely verify motion states as well as determine the optimal response is of crucial importance in vehicular CPSes. Previous works that focus on locating/tracking AVs, either require sophisticated onboard hardware such as radar, lidar, and camera, or rely on centralized infrastructure (e.g. roadside units (RSUs)) to collect trustworthy information from honest neighbors. However, onboard sensors are usually very expensive. Honest neighbors as well as deploying RSUs for every local CPS cannot always be assumed. In this dissertation, we focus on developing a series of cost-effective secure schemes that achieves motion state verification, vehicle localization, tracking, and on-road decision-making, without the help of RSUs, expensive sensors, and honest neighbors, in a timely manner. We first motivate our research by introducing a message delay attack on the V2X communication channel. Then we propose a secure motion verification scheme for ground vehicles by utilizing the opportunistic reflection of the actively transmitted RF signal from vehicles. After that, an RF-based passive localization approach is proposed for a silent drone by exploiting the OFDM symbol properties in the signal reflection process. Fourthly, we propose a secure vehicle tracking scheme by examining the measurement residuals in an extended Kalman filter with the help of only one honest neighboring participant. Finally, a novel trust-aware decision-making framework based on utility-maximization and reinforcement learning is proposed, to help vehicles find the optimal action given the correct traffic information.