Intelligent driving network threat hunting model integrating data dimension reduction and CNN-LSTM

Intelligent driving networks face increasingly complex network attack threats, and traditional defense methods are difficult to effectively deal with high-dimensional attack behaviors. Therefore, a threat hunting model for intelligent driving networks that combines data dimensionality reduction with convolutional neural long short-term memory networks is built. The model uses principal component analysis and autoencoders for joint dimensionality reduction of data, uses CNNs to extract local spatial features of network traffic, and uses long short-term memory networks to capture long-range temporal dependencies of attack behaviors. Experiments show that the model has an average true positive rate of 98.5% for known attacks, a false positive rate of only 2.1%, and a detection rate of 89.2% for unknown attacks. The ablation experiment verifies the necessity of each module. The system's detection delay for different types of attacks is less than 160ms, and the false interception rate is less than 1.5%, meeting the real-time and reliability requirements of intelligent driving scenarios. From this, the proposed method can effectively enhance the efficiency of active threat hunting in intelligent driving networks.