Improving IoT Threat Detection through Machine Learning using PCA for Dimensionality Reduction

The Internet of Things (IoT) is a vast system of interconnected systems that is rapidly growing, but it is also increasingly vulnerable to cyberattacks. Detecting and mitigating IoT threats is essential to ensure the security and integrity of IoT networks. Traditional methods for detecting and mitigating IoT threats are often unable to keep up with the volume and sophistication of modern attacks. This paper proposes a detection system for IoT networks that uses four Machine Learning algorithms (DT, RF, MLP, and NB). Applying machine learning techniques to high-dimensional data suffers from problems such as overfitting, larger computation time, and decreased machine learning model accuracy. Hence, principal component analysis (PCA) has been used here to reduce the dimension of the dataset. The suggested ML algorithms are trained and evaluated on the NSL-KDD datasets. The outcome indicates that Naive Bayes and MLP performed well with PCA and achieved better accuracy to detect and mitigate IoT threats.