A Lightweight Hybrid CNN-Transformer Architecture for Interpretable Evil Twin Attack Detection in Wi-Fi Networks

The study’s key contribution is a lightweight, interpretable CNN–Transformer model for Evil Twin attack detection, developed after benchmarking 63 ML/DL models on 3.78 million Wi-Fi records. The 5.08 MB model achieves near-perfect accuracy while providing attention-based explanations and strong performance on unseen and cross-dataset traffic.