Implementation of Intrusion Detection System for Internet of Things Using Machine Learning Techniques

Abstract

In recent times, Internet brought revolution by connecting the whole world to share the information at one platform. Since data is the most valuable asset, every organization is putting its best effort and spending a lot of money on various security solutions like firewall, antiviruses, etc. to prevent its data and resources from unauthorised access and cyber-attacks like phishing, hacking, eavesdropping, etc. In spite of bulk of these security mechanisms, hackers are still able to exploit the vulnerabilities in the web applications to steal user’s credentials. Intrusion detection system (IDS) is proposed by researchers to detect malicious activity in the network to mitigate the cyber-attacks. In this paper, different techniques of machine learning namely K-nearest neighbor, multilayer perceptron, decision tree, Naïve Bayes and support vector machine have been evaluated for implementation of IDS to classify network connections as normal or malicious. Four measures, i.e., accuracy, sensitivity, precision and F-score, have been taken to assess ability of machine learning techniques under study. Experimental results have shown that decision tree is best classifier for IDS.