Optimizing IoT Threat Detection with Kolmogorov-Arnold Networks (KANs)

TL;DR

This paper explores the use of Kolmogorov-Arnold Networks (KANs) for IoT intrusion detection, showing they outperform traditional neural networks and are more interpretable, providing a promising alternative for IoT security.

Contribution

It introduces KANs with learnable activation functions as a novel approach for IoT intrusion detection, demonstrating improved performance and interpretability over conventional models.

Findings

KANs outperform traditional MLPs in accuracy.

KANs achieve competitive results with state-of-the-art models.

KANs offer superior interpretability for IoT intrusion detection.

Abstract

The exponential growth of the Internet of Things (IoT) has led to the emergence of substantial security concerns, with IoT networks becoming the primary target for cyberattacks. This study examines the potential of Kolmogorov-Arnold Networks (KANs) as an alternative to conventional machine learning models for intrusion detection in IoT networks. The study demonstrates that KANs, which employ learnable activation functions, outperform traditional MLPs and achieve competitive accuracy compared to state-of-the-art models such as Random Forest and XGBoost, while offering superior interpretability for intrusion detection in IoT networks.