In-Vehicle Network Security: Securing the Internal Communication Network Within a Vehicle and Preventing Unauthorised Access to Critical Vehicle Systems

Abstract

Abstract Modern vehicles are software-rich cyber-physical systems in which ECUs, zonal and central computing systems, sensors, diagnostic systems, and external interfaces share safety-critical and privacy-sensitive information. The evolution has created a larger attack surface for in-vehicle networks (IVNs), and widely used communication technologies like Controller Area Network (CAN) were never intended to provide authentication, confidentiality or protection against malicious manipulation. This article critically reviews recent open-access research on IVN cybersecurity. It presents an integrated defence-in-depth framework for CAN/CAN-FD, LIN, FlexRay, Automotive Ethernet, diagnostics, vehicle gateway, ECU security, firmware updates, and over-the-air updates. This article critically collates recent open-access research on IVN security. It proposes an integrated defence-in-depth framework for CAN/CAN-FD, LIN, FlexRay, Automotive Ethernet, diagnostics, vehicle gateway, ECU security, firmware updates, and OAT. It specifically focuses on Artificial Intelligence (AI), Machine Learning (ML), and ML-based Intrusion Detection Systems (IDS), including Convolutional, Recurrent, Autoencoder, Federated, and Lightweight approaches. Accurate detection rates are not enough in scenarios where the dataset may leak, class imbalances exist, adversarial attacks can target the system, false alarms can occur, and computation time and complexity are non-negligible; the analysis argues that all of these can significantly impact the system's security. The proposed framework thus does not treat AI-enabled IDS as an alternative to authentication, cryptographic protection, secure gateways, hardware-bound key protection, or software integrity as monitoring layers, but as an additional security layer that must be applied alongside the others. The framework maps attack surfaces, including security properties and complementary controls, throughout the vehicle's lifecycle. It offers a technically coherent way to mitigate unauthorised access to security-critical vehicle systems while also accounting for real-time, computational, and operational constraints and lifecycle requirements.