Retraction Notice: An Evolutionary Deep Learning Anomaly Detection Framework for In-Vehicle Networks—CAN Bus

Modern vehicles are no longer considered as mere mechanical-based device, instead, they have been hugely replaced by sophisticated electric devices known as Electronic Control Unit (ECU). These ECUs are communicating with each other by publishing/receiving messages that complies with well-known protocol called Control Area Network (CAN). CAN bus is responsible to ensure all critical parts in vehicle e.g. engine, braking, airbag deployment, steering wheel, acceleration, etc. are functioning properly. This indicates that CAN bus is considered as the back-bone network protocol in modern vehicles. Unfortunately, the CAN bus protocol is vulnerable to various cyberattacks due to the lack of security mechanism in the protocol which has introduced several attack surfaces and allows attackers to have legitimate access to the bus and launch malicious activities. This paper proposes a new effective security solution that can detect three different types of message injected attacks namely Denial of Service (DoS), fuzzy, and impersonation attacks in the CAN traffic based on deep learning model. Moreover, the proposed method makes the use of evolutionary optimization algorithm to avoid premature convergence and manual selection of deep learning network architecture. To assess the practicality and effectiveness of the proposed method, CAN traffic is logged using unmodified license vehicle. Furthermore, the proposed method is evaluated using two other different CAN traffic dataset that proves the proposed method can be applied for different car make and models.