A Large-Scale Performance Evaluation of ML-Based and Deep-Neural Models in Evil Twin Attack Identification Using a Hybrid IEEE 802.11 Dataset

The key contribution of this research is a large-scale benchmark of 80 machine learning and deep learning models for Evil Twin attack detection using a hybrid IEEE 802.11 dataset of over 3.8 million samples. It further strengthens the evaluation through independent unseen and public datasets and considers not only detection performance but also generalization, computational efficiency, and deployment suitability, helping identify models that are practical for real-world wireless intrusion detection.