Optimal attention deep learning based in-vehicle intrusion detection and classification model on CAN messages

Intrusion detection systems (IDS) have enormous significance to ensure the security of high-tech automobiles, especially those using the controller area network (CAN) bus for transmission between different electronic control units (ECUs). The CAN is a popular transmission protocol in automobiles, but it is vulnerable to a variety of attacks. To overcome these issues, several studies have investigated the use of IDS for the CAN bus. Researchers have been exploring safety issues of inter- and intra-vehicular transmission. In recent times, intrusion detection sensors are gained popularity despite how easily and effectively they can identify intrusion. Deep learning (DL) and machine learning (ML) algorithms have demonstrated their effectiveness for accurately and quickly identifying intrusions. However, DL approaches need vast quantities of data to accomplish superior outcomes which may be difficult in the case of CAN-based IDS. This manuscript presents an Optimal Attention Deep Learning based In-vehicle Intrusion Detection and Classification (OADL-IVIDC) model to secure CAN messages in model vehicles. The model begins with data preprocessing phase to effectively transform the input data into a more suitable format. For in-vehicle IDS, the OADL-IVIDC framework employs an attention-based augmented long short-term memory (A-LSTM) model. To further optimize the performance of the OADL-IVIDC system, hyperparameters are adjusted using the root mean square propagation (RMSProp) algorithm. The performance of the OADL-IVIDC system is assessed using a standardized car hacking dataset. Experimental results demonstrate that the OADL-IVIDC approach outperforms other techniques in various performance measures.