INTELLIGENT NETWORK TRAFFIC ANOMALY DETECTION USING ML ALGORITHMS

With the increased internet services and digital communication, network security has emerged as a significant issue. Conventional rule-based systems of intrusion detection find it difficult to detect emerging cyber threats and thus machine learning is a more viable approach towards detecting anomalies. The system is an innovative network traffic anomaly detection system based on the latest machine learning algorithms. This benchmark dataset is the KDD Cup 1999 dataset, which includes both malicious and normal network traffic records. Preprocessing of data, feature selection, and training of models is done to enhance the accuracy and efficiency of detection. Several algorithms such as Decision Tree, Random Forest, Gradient Boosting, and CatBoost are compared in terms of performance. CatBoost has the highest accuracy of more than 99% on test data, performing better than the techniques in the base paper. The system is also suitable in the real world by developing a Flask-based web application to perform user authentication, prediction of anomalies, and visualization of performance.