Mitigating IoT Botnet Attacks: Machine Learning Techniques for Securing Connected Devices
Govindarajan Lakshmikanthan, Sreejith Sreekandan Nair, J. Partha Sarathy, Sachin Singh, S. Santiago, B. Jegajothi · 2024
The increasing proliferation of IoT devices has led to an escalated threat of botnet attacks that jeopardise the security and trustworthiness of the interconnected ecosystems. This paper proposes using advanced machine learning models to prevent IoT botnet attacks by instead focusing on the intricate splintering of the malicious bots’ network traffic that characterises bot network behaviour. The research applies a model fusion approach that combines various models, including Deep Neural Network (DNN), Long Short-Term Memory (LSTM) networks, and XG Boost, to improve detection performance and reduce false alarms at optimal levels. The data set includes a variety of benign and malicious IoT traffic, which aids the models in recognising numerous attack strategies and vectors. The architecture and models can accurately identify the target of the attack with a low chance of false positives resulting from overfitting. By utilising advanced feature engineering and dimension reduction techniques, enhance data processing in the context of real-time, scalable botnet detection. Our proposed model offers protection of the IoT infrastructure from all the known behaviours of botnets and the evolution of new ones for a robust defence mechanism. The results show enhanced IoT ecosystem security regarding detection metrics, highlighting greater precision and resilience in the attacks.