Deep Neural Network-Based Trust Models for Securing Interconnected Transportation Systems in Sustainable Cities

Abstract

This study elaborates on a hybrid model of deep learning for detection and classification of malicious behaviour in a transportation system in order to ensure security in smart cities. The model consists of Neural Networks (NN) for spatial feature extraction; XGBoost for temporal analysis; and an Attention mechanism to prioritize data points that are important, improving the model's focus on critical features such as trust degree and vehicle location. By utilizing SMOTE, the model addresses the class imbalance so that under-learning of even the poorly represented attack types like DoS, Whitewash, and Brute-force be alleviated. It effectively learns the spatial-temporal patterns of the vehicle behaviour, thus providing a broader perspective of attacks evolving over time. Performance evaluation is conducted using standard classification metrics such as Accuracy, Precision, Recall, F1-Score, and ROC-AUC, which yield satisfactory results demonstrating that the model is highly accurate in classification as well as detecting malicious behaviour with low false positives. Further, the attention mechanism refines predictions through focus on the most important features. Thus, it is promising, scalable, and dependable in real-time attack detection and continuous monitoring in transportation systems, towards developing secure, resilient smart city infrastructure with a much greater emphasis on security and sustainability.