ADVANCING ANOMALY DETECTION IN IOT: A COMPARATIVE STUDY OF MACHINE LEARNING AND DEEP LEARNING APPROACHES ON THE IOT-23 DATASET

Abstract

With the rise of technology, the IoT got easily available to people in various form and various field. Thus, a rapid amount of data is processed in a limited amount of time. So, reliability and scalability become prime issues. But we cannot forget about the security of this data as it can lead to many problems. We have firewall, antivirus etc., for security but to detect it earlier and solve the problem before any problem can occur is the challenging task. Intrusion detection is one of the solutions for it. Anomaly based IDS is better than signature-based IDS as it finds any irregularity then it detects and take proper action. Here we reviewed about the IoT, its architecture, its applications and its challenges. We find the motivation when we find the importance of security challenges of IoT. It can be solved using Machine Learning and Deep Learning as finding the anomaly in crucial to solve this problem. We tried to implement some machine learning and Deep Learning algorithm on a most recent dataset called IoT-23 which was developed by created by Avast AIC laboratory and Stratosphere IPS. We implemented random forest, decision tree, convolution neural network CNN, stacked long short-term memory gated recurrent unit and extreme gradient boosting. In which we find that decision tree and random forest are most suitable for anomaly detection in IoT-23 dataset.