Investigating Vulnerabilities and Potential Security Threats in Current Synthetic Data Generation and Usage in the Automotive Domain

Synthetic data generation, utilizing Generative Adversarial Networks like Simple GAN and Wasserstein GAN with Gradient Penalty, is vital for advancing automotive cybersecurity in the era of autonomous vehicles. It offers realistic datasets for testing and validation, addressing the complexities of system testing without compromising privacy. This research emphasizes the role of synthetic data. It proposes robust mitigation strategies to secure the Controller Area Network protocol, essential for vehicle communication but vulnerable to cyber threats such as denial-of-service attacks. We evaluate synthetic data's utility in both normal and cyberattack scenarios, addressing risks like unintended disclosure and reverse engineering through differential privacy and feature selection. The study employs advanced validation and benchmarking metrics to ensure data fidelity and privacy. Ultimately, this work advocates for ongoing refinement of these methods and broadening synthetic data applications, aiming to bolster automotive system resilience against cyber threats and setting a foundation for future secure synthetic data generation.