A Review of Intrusion Detection System in IoT with Machine Learning Approach: Current and Future Research
Eddy Prasetyo Nugroho, Taufik Djatna, Imas Sukaesih Sitanggang, Agus Buono, Irman Hermadi
Abstract
Internet of Things (IoT) devices with their network services are often vulnerable to attacks because they are not designed for security. Especially with the rapid technological advances that make data increase exponentially. This is targeted by malicious users to exploit vulnerabilities or interfere with many vulnerability attacks. Therefore, deal with this vulnerability, an intrusion detection system that involves machine learning techniques is needed. Intrusion Detection System (IDS) is targeted to get intrusion in a communication system by looking at the IDS types and methods. This is influenced by the characteristics of the IoT network involved and the reference dataset used in the detection system. This dataset determines the categories or classes of attacks upon which the IDS decides whether or not to intrusion. Reference databases that already exist and are often used, such as KDD Cup 99, NSL KDD, and attack datasets obtained from conditions. In developing IDS in IoT Device, the Machine Learning approach can be used to solve the type of algorithm used consisting of supervised learning, unsupervised learning, or Reinforcement learning. These algorithm methods can be used include SVM, Decision Tree, K-NN, ANN, RNN, and others. From the review analysis of dominant research conducted in 2015-2020, the largest percentage was obtained using the artificial neural network and deep learning algorithm for the intrusion classification process, with details of 16% ANN, 12% RNN, and DNN.