Enhancing IoT Device Security: A Comparative Analysis of Machine Learning Algorithms for Attack Detection

This study sought to compare the effectiveness, efficiency, and scalability of supervised learning algorithms; logistic regression, decision tree, and random forest in IoT networks' attack detection and evaluate the effectiveness of these algorithms in adapting to evolving attack techniques in IoT networks. The study deployed data from a Telecom company encompassing a dataset with a total of 10,000 records and 8 attributes. Furthermore, the dataset comprised both normal and malicious traffic, with 3,000 records classified as attacks and 6,000 records classified as normal traffic. To ensure the creation of reliable and predictive models, a statistical sampling technique called Synthetic Minority OverSampling Technique (SMOTE) was employed. Based on the experiments, the logistic regression algorithm proved to be the most accurate, followed by random forest, and lastly the decision tree algorithm. In the context of IoT device security, the research contributed to an understanding of data preprocessing techniques, feature engineering, and model evaluation. The correlation analysis and heatmap visualization provide valuable insights into the relationships between various variables and highlight potential patterns and trends in the data. This study provides significant knowledge on the improvement of IoT devices' security via machine learning algorithms.