Machine Learning based Anomaly and Threat Detection System in Real Time Social Internet of Things using Dimensionality Reduction Techniques

The Social Internet-of-Things (SIoT) is an innovative model developing in the struggle to safeguard privacy and increase trust inside the Internet-of-Things(IoT). The revolutionary notion of SIoT combines the IoT with social platforms, allowing inanimate items to develop relationships with one another. The fast growth of SIoT has transformed networked networks, enabling effortless interaction between devices. However, integrating these systems creates major security issues, including anomalies and attacks that may compromise system integrity. This study presents a machine learning(ML) methods like Isolation Forest (IF), One class support vector Machine(SVM) and Kernel density estimation(KDE) for in SIoT networks using the Principal component Analysis(PCA) and Correlation Analysis as dimensionality reduction techniques. The findings shows that the IF method provides an accuracy of 93% with the dimensionality reduction techniques applied on RT-IoT dataset. By utilizing powerful machine learning algorithms and contextual analytics, the framework improves detection accuracy and enables strong device collaboration in highly dynamic contexts..