Detection and Classification of Network Attacks in Encrypted Wireless Traffic using Machine Learning Techniques

As encryption becomes the standard, traditional deep packet inspection is losing its edge. Security systems that once relied on reading data payloads are now struggling to keep up with threats they can no longer see because encryption masks the data payload, conventional security methods are now often blind to malicious activity. This paper introduces a machine learning framework designed to detect and classify network attacks within encrypted wire-less traffic using the AWID dataset. By focusing on flow-based and statistical features rather than the hidden content itself, our approach maintains high-level security without compromising user privacy and evaluated three distinct models using MATLAB: a Linear Support Vec-tor Machine (SVM), a Coarse Decision Tree, and a tri-layer Artificial Neural Network (ANN). Our findings reveal that while the Coarse Tree struggled with underfitting (41.67% accuracy) and the SVM reached 87.89%, the tri-layer neural network achieved a near-perfect accuracy of 99%. These results, validated through precision, recall, and F1-score metrics, demonstrate that deep learning architectures are uniquely equipped to overcome the challenges of modern encrypted environments and provide robust network protection.