Explainable AI-Driven Detection of Adversarial Attacks in Deep Wireless Signal Classification

Deep learning (DL) models have been very successful in wireless signal classification, including modulation recognition, channel estimation, and spectrum sensing. However, their vulnerability to adversarial attacks— imperceptible, hostile perturbations intended to trick the models raises serious security issues. This research provides an Explainable AI (XAI)-based methodology for detecting and mitigating adversarial assaults on deep wireless signal classifiers. The proposed method provides interpretable insights into the decision-making process of deep learning models by leveraging advanced XAI techniques such as Gradient-weighted Class Activation Mapping (Grad-CAM) and SHapley Additive Explanations (SHAP), allowing for the detection of irregularities that indicate adversarial interference. Experimental results show that the XAI-driven technique achieves a detection accuracy of more than 90%, which improves the security and transparency of AI-powered wireless communication systems.