Intrusion Detection System as Audit in IoT Infrastructure using Ensemble Learning and SMOTE Method

With the industrial revolution 4.0, the use of IoT-based systems is increasing, both in the field of health manufacturing, urban planning, housing, and even automotive. Therefore, the security of the IoT system needs to be considered, this is related to data integrity, privacy, service stability. Through intrusion detection, activities on the IoT system will be able to be analyzed whether there are suspicious activities that can interfere with or threaten IoT services. In several previous studies in the literature, the approach used to detect intrusions in the IoT system has a high false alarm rate. This research proposes an approach through machine learning, specifically the ensemble learning approach and the synthetic minority over-sampling technique (SMOTE) method as a method of detecting intrusions in the IoT system which is expected to produce better performance. The results of this study indicate that the proposed approach is able to detect intrusion and classify into five types of intrusion including normal intrusion, probe, dos, r2l, u2r. Based on the evaluation results, the proposed approach can improve the performance of intrusion detection in terms of accuracy to 97.02%, detection rate of 97%, false alarm rate 0.16%, compared to base learning and approaches in previous studies used as intrusion detection methods, but in processing time performance have not shown satisfying results.