Time Series Mean Normalization for Enhanced Feature Extraction in In-Vehicle Network Intrusion Detection System

The growth of electronic control units (ECUs) has led to modern automobiles being more convenient and technologically advanced. In order to communicate amongst these control units, the widely used Controller Area Network (CAN) communication protocol is used. CAN buses do not, however, come with built-in security features while being reliable and cost-effective. Because of this weakness, they are vulnerable to several attacks from prospective enemies. The installation of an Intrusion Detection System (IDS) is, nonetheless, a practical answer to this issue. A CAN bus system’s security can be considerably improved by the use of an IDS. We suggest a Long Short-Term Memory (LSTM) based intrusion detection system that is capable of effectively addressing the detection and security needs of CAN bus systems. We think that using Long Short-Term Memory (LSTM) models has a number of benefits, including highly accurate classification and effective anomaly detection. Our research findings have shown that our Long Short-Term Memory (LSTM) based IDS successfully identifies assaults on CAN systems with an accuracy rate of 99% and little loss. By providing this solution, we want to aid in the creation of reliable intrusion detection systems that can successfully protect CAN bus systems.