A Review Paper on Network Intrusion Detection System with Machine Learning Approach

Abstract

Computer networks are prone to cyber as a consequence of global internet use; as
a consequence, academics have developed several IDSs. Identifying intrusions is one of the
main significant study topics in data security. It aids in the detection of misuse & attacks as a
safeguard for the network's integrity. ML, Bayesian-based method, nature-inspired metaheuristic methods, swarm intelligent approach, and Markov neural net is some ways to find
the most effective characteristics and thus improve the effectiveness of IDS. Over the years,
numerous databases have been used to evaluate various projects. This publication provides a
comprehensive assessment of an IDS with Machine learning approaches. Rapid
advancements in the internet and communication fields have resulted in a massive expansion
of network size and data. As a result, many new attacks are being developed, making it
difficult for network security to detect intrusions accurately. Furthermore, intruders with the
intent of launching various attacks within the network cannot be overlooked. An intrusion
detection system (IDS) is a tool that inspects network traffic to ensure its confidentiality,
integrity, and availability and thus protects the network from possible intrusions. Despite the
researchers' best efforts, IDS continues to face difficulties in improving detection accuracy
while lowering false alarm rates and detecting novel intrusions. Machine learning (ML) and
deep learning (DL)-based IDS systems have recently been deployed as potential solutions for
quickly detecting intrusions across the network. The taxonomy in this article is based on the
notable ML and DL techniques used in designing network-based IDS (NIDS) systems, and it
first clarifies the concept of IDS. The strengths and limitations of the proposed solutions are
discussed in depth in this comprehensive review of recent NIDS-based articles. The proposed
methodology, evaluation metrics, and dataset selection are then discussed, as well as recent
trends and advancements in ML and DL-based NIDS. We highlighted various research
challenges and provided future scope for research in improving ML and DL-based NIDS by
utilizing the shortcomings of the proposed methods