Improving IoT device’s security in constrained environment to replicate vulnerability using machine learning approaches

Through its inexplicable intelligent services, IoT platforms have grown into a worldwide powerhouse that has taken over every part of our everyday lives in the past ten years. It allows multiple smart devices to communicate with one another with the least amount of human intervention. The Internet of Things’ services are becoming an essential aspect of human existence, and in the future years, there is an anticipation that the figure of devices connected to IoT networks will rise at a rate that is exponential. With usage in a variety of industries, including wearable technology, critical infrastructure, transportation, cities, homes, and hospitals, it is currently one of the most intriguing areas of computing. As the volume of IoT devices is growing rapidly, security concerns are also at risk. For such devices, traditional authorization and verification techniques are insufficient. Additionally, we require techniques that can identify the attack early on by analyzing past trends and promptly notify the system. Machine Learning (ML) has seen significant technological improvement in recent years, opening up numerous research avenues to address current and upcoming IoT concerns. In relation to giving system intelligence, machine learning(ML) algorithms are found to be superior when compared with basic security techniques. This paper introduces a novel methodology that uses machine learning (ML) classifiers, specifically Random Forest, Logistic Regression, K-nearest neighbor, Support Vector machine, adaboost, Naive bayes, and XgBoost to improve the security of IoT devices. The results of the experiment demonstrate the importance of Xgboost with the highest accuracy of all the other approaches in enhancing IoT security.