Understanding IoT-23 Dataset: A Benchmark for IoT Security Analysis

As the number of Internet of Things (IoT) devices rises, robust cybersecurity solutions are desperately needed to shield these interconnected systems from ever-evolving threats. This paper presents a novel approach to identifying attacks using machine learning (ML) techniques, using the IoT-23 dataset as an evaluation standard. The IoT-23 dataset offers a diverse collection of network traffic data collected from a range of IoT devices, facilitating the development and validation of intrusion detection systems (IDS) specifically tailored for IoT scenarios. This method includes preprocessing the raw data, identifying pertinent features, and training ML models such as Random Forest (RF), Extreme Gradient Boosting (XGB), and K-Nearest Neighbours (KNN). Performance evaluation uses key performance indicators (KPIs) including accuracy, recall, precision, and F1-score to assess how effective the recommended approach is. The findings demonstrate how ML may enhance the security design of IoT systems by offering a proactive means of thwarting emerging cyber threats and a means of alerting users to them. The accuracy of the RF model may reach up to 89 %, whereas other models, such as KNN and XGB, yielded results of 85 % and 88 %, respectively.