Malicious Bot Detection in Large Scale IoT Network Using Unsupervised Machine Learning Technique

The extensive research of Internet of Things (IoT) apps and connected digital gadgets has been heavily targeted by intruders launching spread attacks due to lossy wireless networks. Attackers employ botnets, which are attack vectors made up of captured bots created for a specific purpose, to gain control of systems and components by acting maliciously. In order to minimize those problems, a distributed machine learning model was employed to extract the proper feature and pick features that would protect the application or network from hostile attacker behavior. To develop an effective and efficient secure identification of IoT-based risks in the heterogeneous network, a well-structured model must be built for training and testing along some distribution of the dataset toward verifying the recommended system. In order to build the best botnet attack detection model based on the numerous attack characteristics of the botnet, attack component analysis has been proposed in the study to classify the best attack feature subsets on various attack features acquired from the benchmark dataset. Genetic algorithms have been used to extract discriminating characteristics from network log data in order to provide the best feature subsets.