Exploring Lightweight Deep Learning Techniques for Intrusion Detection Systems in IoT Networks: A Survey

Abstract

The proliferation of Internet users has coincided with a commensurate increase in the amount of very important, sensitive, and private information being transferred across the Internet. Malicious actors are increasingly targeting networks to breach them and obtain illegal access to critical information since this trend has revealed holes in security systems. In addition to endangering the privacy of the data concerned, these breaches disrupt the smooth functioning of systems. Therefore, in light of these dangers, intrusion detection systems (IDSs) are now an essential part of any cybersecurity program. The goal of these systems is to detect and report any suspicious activity by constantly monitoring and analyzing network traffic. Numerous review articles have investigated various methods for network intrusion detection. To improve detection accuracy while keeping computing efficiency high, this survey study investigates lightweight deep learning techniques for intrusion detection systems. These techniques include pruning, quantization, clustering, and collaborative optimization. This study analyzes five different types of new real-world traffic datasets (i.e., CSE-CIC- IDS2018, NSL-KDD, Bot-IoT, ToN IoT Network, and UNSW-NB15) and evaluates the performance of several machine learning and deep learning techniques. This survey provides metrics for measuring the accuracy of intrusion detection across various systems, which may be used to assess performance.