Comparative Analysis of Machine Learning Techniques for DDoS Intrusion Detection in IoT Environments

This study addresses the challenge of Distributed Denial of Service (DDoS) attacks in the Internet of Things (IoT) environment by evaluating the effectiveness of Intrusion Detection Systems (IDS) using machine learning techniques. Due to the lightweight computational configuration of IoT systems, there is a need for a classifier that can efficiently distinguish between legitimate and malicious network traffic without demanding substantial computational resources. This research presents a comparative analysis of four machine learning models: (i) k-Nearest Neighbour (k-NN), (ii) Support Vector Machine (SVM), (iii) Random Forest (RF), and (iv) Multilayer Perceptron (MLP), to propose a lightweight DDoS intrusion detection classifier. A novel classification model based on the MLP architecture is proposed, focusing on minimalistic design and feature reduction to achieve accurate and efficient classification. The model is tested using the CICIDS2017 dataset and demonstrates high accuracy and computational efficiency, making it a viable solution for IoT environments where computational resources are limited. The findings show that the proposed µML-IDS model achieves an accuracy of 99.8%, F-score of 96.5%, and precision of 99.96%, with minimal computational overhead, highlighting its potential for real-world application in protecting IoT networks against DDoS attacks.