Explainable Machine Leaning-based False Data Injection Classification Framework for AVs
Vrutik Soni, Dev Mehta, Ansh Shah, Sriya Dhingani, Rajesh Gupta, Sudeep Tanwar, Aparna Kumari · 2025
With the increasing adoption of Autonomous Vehicles(AV) in modern transportation systems, there are risks associated with False Data injection (FDI) attacks, which target AV sensors.There is a need for an efficient system that can detect and identify this attack, ensuring the correct operations of AVs. We conducted experiments using a dataset containing FDI attack data for Industrial IoT (IIoT) environments integrated with AVs. This research proposes an effective approach for FDI attack detection and classification. Feature selection using SHapley Additive exPlanations (SHAP) was performed to increase the efficiency of ML models. To enhance the interpretability of the framework, we integrated Local Interpretable Model-Agnostic Explanations (LIME) based Explainable AI (XAI) techniques, providing transparency in attack detection. Our approach achieved a significant accuracy of 99% using the Random Forest classifier. The integration of XAI with ML model prediction helps in attaining high accuracy, explainability and efficient identification of FDI attacks. This study bridges the research gap by providing a solution that aids in detecting and neutralizing FDI attacks in AV systems and contributing to the security and reliability of autonomous operations.