A Synoptic Review on Feature Selection and Machine Learning models used for Detecting Cyber Attacks in IoT

Abstract

There is a colossal increase in the cyberattack on the Internet of Things due to the rapid increase in its adoption rate worldwide. For ease of use, these devices are accessed by the end-user using an open network which increases the surface area of the attack which puts the user's privacy at stake. Any adversary can exploit the vulnerability in the IoT devices from anywhere which prioritises the privacy and security of computing resources. Machine learning models with an optimal feature selection method have been considered as a viable solution for mitigating various cyber-attacks and detecting malicious network traffic. This study intends to review, various machine learning algorithms and feature selection methods used for various cyber-attacks detection. Different machine learning techniques including Convolutional Neural Net-work, Random forest, Logistic regression, Random Forest used by researchers that are suitable for mitigating various attacks like Denial of service attacks, BotNet attacks etc have been discussed. This study provides a comprehensive comparison of different ML models and the feature selection methods used to train the models.