Towards Smart Online Network Intrusion Detection Systems

This project aims to develop a machine learning algorithm to detect intrusions in network traffic.More specifically, our focus will be on conceiving an advanced theoretical and practical model for network intrusion detection and classification. The model will be tested on the existing benchmarks with the hope to apply it using SQU network traffic.First, conduct a comparative study of different Machine Learning techniques used to discriminate normal connection from abnormal connection (intrusion) in the traffic of a network. Second, develop novel supervised machine learning techniques of automatic classification of the traffic intrusions using the same dataset. The focus will be on the automatic representation learning of the known (normal) traffic patterns in such a way to discriminate them easily from any abnormal pattern. Automatic representation learning avoids resorting to any handcrafted design of pattern features, which makes the detection/classification system smarter and more accurate.