Advanced Machine Learning for Robust Botnet Attack Detection in Evolving Threat Landscapes

Abstract

As technology advances and security issues and cyberattacks increase, extensively Internet of Things (IoT) devices are linked to networks, and botnets have been emerging and evolving very fast, and they pose a dangerous threat. As systems become more complex, scale and, therefore, more complex, cyberattacks mounted against their vulnerabilities also increase. IoT transition is disrupted using these attacks, disrupting the IoT devices' networks and services approaches for botnet attack detection and classification using Machine Learning (ML) and Deep Learning (DL) have been developed within the framework of the IoT. This study provides an intrusion detection system (IDS) based on the Bidirectional Gated Recurrent Unit (Bi-GRU) for detecting botnet attacks in IoT networks. We use the N-BaIoT dataset for this purpose. The study opted for a Bi-GRU model, which can detect contextual dependencies in the past and the future, to deal with the sequential IoT traffic data. The Bi-GRU model performance achieved exceptional results in classifying network traffic. The system's accuracy in identifying both malicious and benign traffic was 99.99%. Additionally, the accuracy of these models rapidly rises and eventually levels out at almost 100%, indicating strong model performance. The model's ability to recognise various botnet attack types even in cases of data imbalance was demonstrated by important performance metrics such as ROC-AUC, accuracy, precision, recall, and F1-score. The results show that the proposed Bi-GRU-based IDS is a robust and improved solution for detecting IoT botnet attacks on a real-time basis. While the model performs impressively, it has some problems, including the minor misclassification in complex attack cases and dependency on a single dataset, which restricts its generalisation. Future work will focus on improving model robustness.