Diversity Measure to Tackle the Multiclass Problem in IoT Intrusion Detection Systems

Abstract

The advent of the Internet of Things (IoT) has instigated transformations in various domains, such as healthcare, smart homes, agriculture, transportation, and manufacturing. With the swift proliferation of IoT networks, new security challenges have surfaced, exposing them to a plethora of attacks. To counter these, machine learning-driven intrusion detection strategies have been introduced, which scrutinize the behavior and communication patterns of IoT devices to identify and nullify any suspicious undertakings. While these methodologies demonstrate high accuracy and minimal false alarm rates in static scenarios, their performance stability in dynamic, evolving environments remains undetermined. One critical issue pertains to multiclass problems, wherein the complexity of diverse attack types can significantly affect the efficacy of machine learning-based intrusion detection systems, if not promptly recognized and addressed. This paper introduces an innovative IoT Intrusion Detection approach that incorporates the Diversity measure as a model drift detection method to tackle the multiclass problem in IoT networks. Our proposed approach can detect previously unknown attacks in IoT networks through an advanced drift detection technique.