Detecting Evolving Cyber Threats in IoT Environments Using Machine Learning

In recent years, the evolution of IoT devices and applications has been accompanied by a corresponding evolution in cyberattacks targeting these systems.Traditional approaches in cybersecurity often rely on models trained using historical data, which may accurately predict known attack patterns but struggle to detect emerging threats or evolving attack strategies.To address this challenge, this study introduces a novel model named Concept-Drift XGBoost (CD-XGB).Recognizing evolving attacks as instances of concept drift, this research proposes an additional algorithm termed Improved Concept-Drift Identification (ICDI) to identify and adapt to changing attack patterns.The performance of the CD-XGB model is evaluated using three diverse datasets: UNSW-NB15 2015, HIKARI 2021, and CIC IoV 2024.The results demonstrate impressive accuracy rates of 99.96%, 99.99%, and 99.96% for the respective datasets, underscoring the effectiveness of CD-XGB in addressing the challenge of evolving cyber threats in IoT environments.