A Machine Learning-Based Intrusion Detection System for Securing Internet of Things Networks
Yiqian Zhang · 2025
This paper explores the deployment of machine learning (ML)-based Intrusion Detection Systems (IDS) to enhance security in Internet of Things (IoT) networks, which are increasingly vulnerable due to the diversity and scale of connected devices. Traditional security solutions are often inadequate for the dynamic and distributed nature of IoT environments. We propose a ML-based approach, utilizing supervised learning techniques, to detect both known and novel intrusion behaviors effectively. By analyzing the network activities, our model identifies anomalies that signify potential threats, thereby securing the network against a wide range of cyber-attacks including DDoS, malware, and data theft. We perform a comparative analysis of several ML algorithms including decision trees, random forest and k-nearest neighbor and then conduct extensive experiments on the public dataset BoTNeTIoT-L01 dataset. The experimental results demonstrate the effectiveness of decision trees and random forest in achieving high accuracy. The results underscore the superiority of machine learning methods over traditional rule-based systems in adapting to new threats and maintaining robust network security. The paper concludes with a discussion on future trends in IDS, particularly the integration of AI to automate response actions and the potential for decentralized learning models to preserve data privacy in sensitive applications.