An Anomaly Detection Method Based on Multiple LSTM-Autoencoder Models for In-Vehicle Network

The CAN (Controller Area Network) protocol is widely adopted for in-vehicle networks due to its cost efficiency and reliable transmission. However, despite popularity, the lacks built-in security mechanisms, making it vulnerable attacks such as flooding, fuzzing, DoS. These can exploit vulnerabilities disrupt expected behavior of network. One main reasons these concerns that relies on broadcast frames communication between ECUs (Electronic Control Units) within To tackle this issue, we present an intrusion detection system leverages multiple LSTM-Autoencoders. proposed utilizes diverse features, including transmission interval payload value changes, capture various characteristics normal network behavior. effectively detects anomalies by analyzing different types features separately using LSTM-Autoencoder model. In our evaluation, conducted experiments real vehicle traffic, results demonstrated system’s high precision with a 99% rate in identifying anomalies.