Role of Machine Learning and Deep Learning Approaches in Designing Network Intrusion Detection System

There has been increasing number of attacks in computer communication networks including IoT and cloud infrastructure in recent years. Finding out malicious attacks from unknown sources is biggest challenge in network intrusion detection system. Identifying normal network traffic from malicious traffic is also a complex task in IDS. Good IDS requires highest detection rate and lowest false alarm. IDS acts as a second level of defence in addition to network firewall. Network IDS consists of three modules: data collection, feature selection and decision engine. IDS can be classified into two types namely misuse detection and anomaly detection commonly used in practise. To improve the IDS, detection efficiency researches are focusing on many deep learning techniques. Deep learning is an advanced subset of machine learning. These techniques are closer to artificial intelligence domain. Deep learning can be applied to many challenging learning problems and has generated good results. This paper investigates the appropriateness and analysis of application of deep learning and machine learning methods in implementation of network intrusion detection system.