Resource Efficient Intrusion Detection Systems for Internet of Things Using Online Machine-Learning Models

The proliferation of Internet of Things (IoT) technology across diverse domains has led to a surge in both the frequency and diversity of cyber-attacks targeting IoT infrastructures. In response to this escalating security challenge, several machine learning-based Intrusion Detection Systems (IDSs) have been developed. However, the machine-learning models used in IDSs are generally not designed with IoT infrastructure's resource constraints in mind. In response to that issue, we propose the utilization of online machine learning models to build efficient IDS for IoT. In this study, we present an extensive exploration of the implementation of online machine-learning algorithms to develop efficient IDSs for IoT. To evaluate the performance of the online machine-learning models, we tested several online machine-learning models using the TON-IoT dataset which was designed specifically for evaluating Artificial Intelligence (AI)-based security applications in IoT. The experimental results showed that the online machine-learning models exhibit performance on par with batch machine-learning models while saving significant computational resources. This notable benefit highlights the potential of online machine-learning models as promising candidates for developing machine learning-based IDSs for IoT infrastructure.