Signature-Over-The-Air with Transfer Learning IDS for Intelligent Connected Vehicles (ICV)

Intelligent Connected Vehicles (ICV) are considered the promising technology that will replace traditional vehicles soon, especially with the improvement in automation and network technologies. This gives rise to the possibility of new cyberattacks, including intra- and inter-vehicle communication attacks. This paper proposed an infrastructure-independent, Intrusion Detection System (IDS) to secure the Intra-Vehicles and external networks. The anomaly-based IDS in the model is assisted by a blacklist attacks signatures that are placed in the connected vehicles, where the signatures of the new attacks can be generated in a cloud-based security management system and updated to the network of connected vehicles using the over-the-air-update concept, so each network-connected vehicle will work as a packet inspector that helps to find the attacks log and add it to a cloud signature database. In the detection engine, Self-taught Transfer Learning (STL) is used to transfer the knowledge of a pre-trained Deep Belief Network (DBN) model from the source domain and construct a feed-forward Deep Neural Network (DNN). The comparison with baseline machine learning (ML)/deep learning (DL) algorithms shows that the proposed model achieves better performance in terms of Accuracy, Precision, Detection Rate (DR), F1-score, and Receiver Operating Characteristic (ROC).