The Domain Name System (DNS) is widely used in network communication but is vulnerable to DNS tunneling attacks that hide malicious data within normal DNS traffic. This project proposes a hybrid neural network detection system using a Voting Classifier that combines Random Forest, XGBoost, and Multi-Layer Perceptron. The model is trained on the FGSM_combined dataset containing both normal and adversarial DNS traffic. Results show high accuracy and improved detection performance, and a Flask-based web application allows users to upload traffic data for instant attack detection.

Keywords: DNS Tunneling, Adversarial Attacks, Voting Classifier, Ensemble Learning, Hybrid Neural Network, FGSM, Random Forest, XGBoost, Multi-Layer Perceptron, and Intrusion Detection.