Two stage malware detection model in internet of vehicles (IoV) using deep learning-based explainable artificial...

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Intelligent Transportation Systems

Background:

• The Internet of Vehicles (IoV) faces significant privacy and security challenges due to increasing malware threats.
• Malware can lead to data theft, corruption, and cybercrimes, impacting network operations and user safety.
• Existing malware detection solutions often require improvement in speed and accuracy.

Purpose of the Study:

• To propose a novel deep learning-based explainable artificial intelligence model (MDMIoV-DLXAI) for enhanced malware detection and classification in IoV environments.
• To improve the accuracy and efficiency of malware detection systems in connected vehicles.
• To leverage explainable AI (XAI) for better decision-making in AI-driven security.

Main Methods:

• Data normalization using min-max normalization.
• Feature selection employing the Reptile Search Algorithm (RSA).
• Malware classification using a hybrid Bidirectional Long Short-Term Memory with Multi-Head Self-Attention (BiLSTM-MHSA) model, optimized by the Pelican Optimization Algorithm (POA).
• Explainable AI (XAI) using SHAP for enhanced decision-making.

Main Results:

• The proposed MDMIoV-DLXAI model achieved a superior accuracy of 97.393% in malware detection.
• The combination of DL models, optimization algorithms, and XAI techniques demonstrated significant performance improvements.
• Experimental evaluation confirmed the effectiveness of the model against existing techniques.

Conclusions:

• The MDMIoV-DLXAI model offers a robust and accurate solution for malware detection in IoV networks.
• Explainable AI enhances the trustworthiness and interpretability of AI-driven security systems in vehicles.
• This research contributes to securing intelligent transportation systems against evolving cyber threats.