Explainable AI-driven IoT botnet detection for 6G networks
TL;DR
This study proposes an explainable AI-driven IoT botnet detection framework utilizing the N-BaIoT dataset to bridge the gap between high detection precision and human-interpretable reasoning, fulfilling essential trust and reliability requirements for next-generation 6G IoT security architectures.
Abstract
The rapid proliferation of Internet of Things (IoT) devices in smart homes, healthcare, and industrial automation introduces severe security challenges, especially with the transition toward 6G intelligent networks. Traditional machine learning models achieve high detection accuracy for IoT botnets but often lack interpretability, limiting their trustworthiness in mission-critical 6G environments. This study proposes an explainable AI-driven IoT botnet detection framework utilizing the N-BaIoT dataset to bridge this gap. The framework integrates correlation-based feature selection, mutual information analysis, ANOVA filtering, and low-variance elimination to construct a lightweight yet discriminative feature space. Several ensemble learning algorithms were evaluated, with LightGBM achieving the highest accuracy (99.82%) and exceptional stability across cross-validation folds. To enhance transparency, model decisions were explained using SHAP and LIME, revealing key traffic features that differentiate benign and malicious flows. The proposed system demonstrates not only high detection precision but also human-interpretable reasoning, fulfilling essential trust and reliability requirements for next-generation 6G IoT security architectures.