A Hybrid Anomaly Detection System for Electronic Control Units Featuring Replicator Neural Networks

摘要

Due to the steadily increasing connectivity combined with the trend towards autonomous driving, cyber security is essential for future vehicles. The implementation of an intrusion detection system (IDS) can be one building block in a security architecture. Since the electric and electronic (E/E) subsystem of a vehicle is fairly static, the usage of anomaly detection mechanisms within an IDS is promising. This paper introduces a hybrid anomaly detection system for embedded electronic control units (ECU), which combines the advantages of an efficient specification-based system with the advanced detection measures provided by machine learning. The system is presented for - but not limited to - the detection of anomalies in automotive Controller Area Network (CAN) communication. The second part of this paper focuses on the machine learning aspect of the proposed system. The usage of Replicator Neural Networks (RNN) to detect anomalies in the time series of CAN signals is investigated in more detail. After introducing the working principle of RNNs, the application of this algorithm on time series data is presented. Finally, first evaluation results of a prototypical implementation are discussed.