A lightweight energy consumption ensemble-based botnet detection model for IoT/6G networks

Abstract

The potential for significant damage to an enterprise network by an intruder or cybercriminal wielding a botnet is substantial. Such malicious actors actively scan vulnerable connected devices, aiming to incorporate them into their botnet network for exploitation. Previous attempts to mitigate this issue have been met with varying success levels, often exhibiting inaccuracies and consuming excessive energy. The proposed model introduces a streamlined ensemble-based detection framework tailored for identifying botnets within IoT networks. Leveraging Machine Learning (ML) techniques, the framework effectively detects and safeguards the network's infrastructure. The proposed approach identifies crucial features by employing a method for univariate feature selection, coupled with an ensemble-based framework. Botnet attacks consume a significant amount of energy in IoT devices. The proposed model detects and avoids botnet attacks, which can save energy and make IoT networks more sustainable. The suggested model synergizes the capabilities of two hyper-tuned ML algorithms, namely XGBoost and LightGBM. Experimental findings underscore the effectiveness of the proposed model, demonstrating a remarkable 100% accuracy rate in detecting malicious botnets within the network, surpassing other models which ranged between 97% and 99% accuracy.