A Comparison of AI-Enabled Techniques for the Detection of Attacks in IoT Devices

Abstract

The purpose of this research is to enhance the security of Internet of Things devices. The deployment of these gadgets has increased exponentially, nowadays they can be found in every sector from industrial environments to applications in residential homes. This technology is important for enterprises due to the ability to manage different kind of data and control critical operations in an efficiently way. Consequently, these devices have become frequent targets of cyberattacks, highlighting the necessity of robust detection methods. However it is a complicated task to find a uniform security solution due to the variety of devices, with different capabilities and requirements. This study evaluates some artificial intelligence supervised learning algorithms such as XGBoost and Random Forest using the TON_IoT dataset, which are new generations of Internet of Things and Industrial datasets for evaluating the fidelity and the efficiency of different cybersecurity applications based on Artificial Intelligence. The analysis shows both algorithms are effective in detecting cyberattacks, achieving accuracies close to 98%, with a minimal variation in terms of precision and recall. These algorithms outperform the results obtained in previous works with the same dataset, then they can provide an additional security layer, allowing accurate identification of potential attacks. This research shows the importance of artificial intelligence algorithms in cybersecurity and their potential to improve the protection of this kind of devices.