Generative AI Threats to Maritime Navigation Using Deceptive ISAR Images

The state of the art in deepfake technology demonstrates rapid technological evolution. Machine learning algorithms are becoming increasingly talented at generating synthetic yet realistic media, raising ethical, social security, and political concerns. Concurrently, the increasing frequency of supply chain attacks and Advanced Persistent Threats (APTs) poses an evolving threat to the security of critical infrastructure, drawing the attention of state-level actors. Our research on AI-generated threats explores the fabrication of images that depict non-existent events, demonstrating the ability to create false radar images. Specifically, counterfeit Inverse Synthetic Aperture Radar (ISAR) images, created using Generative Adversarial Networks (GANs), closely resemble real targets and present a significant threat to maritime operations when exploited in supply chain attacks or by APTs. Radar systems are one of the main elements of ship navigational chains that provide vital information on the surrounding area in terms of distance and speed. Real data analysis in this paper has been conducted on an ISAR database extracted from the NATO SET-196 trials, demonstrating the capability of GANs to create convincing fake ISAR images. Such experiments raise awareness of the vulnerabilities of imaging radar systems to novel, AI-generated cyberattacks.