A secure and efficient deep learning-based intrusion detection framework for the internet of vehicles.

This swift growth in Internet of Vehicle (IoV) networks has created serious security issues, primarily in intrusion detection due to the fact that these are complex, dynamic, and large-scale networks. AES-256 encryption for strong real-time security and access control, along with Secure Multi-Party Computation (SMPC) and Homomorphic Encryption (HE) for privacy-preserving collaborative data processing and encrypted computations, are some of the innovative contributions to IoV security that this work presents. Z-score normalization and median imputation are two excellent methods for prepping high-quality data for a deep learning-based intrusion detection system (IDS). Vision Transformer (ViT), wavelet transforms, and GAT ensure effective feature extraction, and a novel hybrid optimization known as Crayfish-Mother secure Optimization (CMSO) method is proposed to optimize feature selection to its maximum and reduce computational cost. DenseNet, GoogleNet, AlexNet, and SqueezeNet are also integrated in the newly proposed DAGSNet architecture to enhance feature detection and classification, enhancing the dependability and effectiveness of the IDS for IoV security. A highly secure, effective, and precise intrusion detection system in IoV environments is guaranteed by this holistic approach with the minimum time of encryption and decryption (0.02 s, 0.82 s) and maximum precision of two datasets (0.991, 0.984).