Resilient System Design against False Data Injection Attacks on Connected and Automated Vehicles

Author

Chunheng Zhao, Clemson UniversityFollow

Date of Award

12-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Automotive Engineering

Committee Chair/Advisor

Pierluigi Pisu

Committee Member

Jerome McClendon

Committee Member

Bing Li

Committee Member

Mert D. Pesé

Committee Member

Robert Prucka

Abstract

This dissertation develops resilient system designs to defend connected and automated vehicles (CAVs) against false data injection (FDI) attacks across communication and perception domains. As CAVs integrate into transportation systems, they face critical security vulnerabilities where attacks can result in physical consequences, including crashes, economic losses, and threats to human life. This research addresses two attack surfaces. For connected vehicles, FDI attacks target vehicle-to-vehicle communication channels where adversaries manipulate shared information such as position and velocity data. For automated vehicles, adversarial FDI attacks exploit deep neural network (DNN)-based perception modules, causing misclassification of critical environments through crafted perturbations. To secure connected vehicles, this dissertation proposes two novel particle filter-based data fusion and attack detection frameworks. The first addresses vehicle platooning through a two-stage architecture leveraging cloud-based sandboxing. The second extends this approach to vehicle rerouting that scales across varying numbers of vehicles. To secure automated vehicles, this dissertation introduces complementary approaches. First, a robustified DNN combining discriminative and generative models enhances classification robustness through causal modeling, extended to CAN network intrusion detection. Second, a resilient vision-language model (VLM)-based end-to-end driving framework protects multimodal systems through adaptive modality reweighting and early fusion. This research contributes fundamental advances in CAV security by showing that integrated approaches combining control theory, data fusion, and machine learning offer superior protection. We envision that our technique will play an important role in securing CAVs and thus accelerate their spread.

Recommended Citation

Zhao, Chunheng, "Resilient System Design against False Data Injection Attacks on Connected and Automated Vehicles" (2025). All Dissertations. 4113.
https://open.clemson.edu/all_dissertations/4113

Author ORCID Identifier

0000-0002-3121-4779