GAttack: Generative Attack on In-Vehicle Network

The rapid progression of Artificial Intelligence (AI) is propelling the connected and autonomous capabilities of modern vehicles to new heights, leading to more sophisticated levels of network integration. This rapid technological stride, however, has escalated the cybersecurity challenges by broadening the attack surface, thus increasing the risk of compromise to In-Vehicle Networks (IVN) and vehicular functionalities. This work investigates the potential threats posed by sophisticated Generative AI-based attacks. It demonstrates a Long ShortTerm Memory (LSTM) based Generative Adversarial Network (GAN) attack mechanism, capable of bypassing state of the art Intrusion Detection System (IDS). The GAN model is capable of learning temporal characteristics of network traffic without needing the system details which makes it applicable to wide range of network. It is able to produce malicious data that can bypass the IDS without triggering the thresholds and penetrate the network stealthily, thereby compromising the integrity of the system. GAttack is validated by testing it against a state-of-the-art IDS, demonstrating a low detection rate of 35.29%, compared to the 85% or higher detection rates achieved against other known attacks.