Deep learning-based intrusion detection system for in-vehicle networks with knowledge graph and statistical methods

In-vehicle networks are increasingly vulnerable to sophisticated cyberattacks, posing significant threats to the safety and security of connected vehicles. Traditional intrusion detection systems (IDS) often fall short in effectively identifying both known and unknown attacks, especially within the complex and dynamic environment of in-vehicle communication systems like the Controller Area Network (CAN). Current IDS approaches struggle with detecting novel attacks, incorporating semantic features, and identifying anomalies in non-periodic messages, highlighting the need for more advanced and robust solutions. To address these challenges, this study introduces HDL-IDS, a hybrid deep learning-based IDS specifically designed for in-vehicle networks. HDL-IDS integrates knowledge graphs, statistical methods, and deep learning techniques—namely Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks—to extract relevant features from raw network data and accurately classify network traffic. By combining the strengths of CNNs and LSTMs, the system captures both spatial and temporal dependencies, while the integration of knowledge graphs enhances semantic analysis. Experimental evaluations demonstrate that HDL-IDS outperforms existing IDS solutions, achieving accuracy levels of up to 99% with a low false-positive rate. The system effectively addresses key research gaps by improving detection of both known and unknown attacks, seamlessly integrating semantic features, and accurately detecting anomalies in non-periodic messages. Although HDL-IDS shows significant promise in enhancing the security of in-vehicle networks, further research is needed to optimize its efficiency, explore transfer learning techniques for detecting unknown attacks, and evaluate its performance across a broader range of datasets. Despite these ongoing challenges, HDL-IDS represents a meaningful advancement in the pursuit of secure and safe connected vehicles.