A Robust Intrusion Detection Mechanism Using Ensemble Approach for IoT Paradigm

Intrusion detection systems (IDSs) protect computer systems by analyzing internet traffic and activities to identify possible threats. This study aimed to create an advanced IDS using feature selection and ensemble learning techniques that achieve high accuracy. The study gains relevance by using the TON_IoT dataset for training and testing. The study was divided into two main phases, each adding their significance. This strategy maximizes the IDS's performance by choosing the most informative features and utilizing the advantages of various classifiers. The second stage involved applying the ensemble learning approach, which yielded a potent model combining the present algorithms' advantages. The study's findings show how it affects attack detection quality and false alarm rate reduction. According to experimental analysis and the TON_IoT benchmark dataset findings, the proposed approach outperformed several existing deep learning techniques. It displayed a maximum accuracy of 99.63% in binary classification and 99.78% in multiclass classification scenarios for network attack identification.