Securing Heterogeneous Computing Architectures in Self-Driving Cars: Isolation, Inter-Processor Communication, and Threat Mitigation

Considering we are handling identical systems of homogeneous computer architecture, the security of such computer structures also takes priority, and thus, we are here to help pay attention to them so that everything runs smoothly and in proper order. These systems leverage Field-programmable gate arrays (FPGA) alongside processing units such as Central Processing Units (CPU) and Graphics Processing Units (GPU) to perform functions such as navigation, decision-making, and real-time perception. Yet creating secure pathways for communications between processors, as well as potential security threats, makes it difficult enough to verify the separation of essential functions. The thesis will primarily focus on self-driving cars as a specific use case, along with three main topics: isolation, communication between processors, and threat mitigation. Furthermore, it will also explore how exchanging sensitive data and processes between different processors helps prevent unauthorised interventions. The security of data transmission guarantees that data can be kept private and authentic when passed between CPU, GPU, and FPGA components. The other half is what we call threat reduction. More secure systems that offer complete immunity against attacks, including unauthorised access, side-channel attacks, and error injects that can exploit shortcomings in any system.