Comparative Analysis of Deep Learning Models for IoT Botnet Detection: CNN, LSTM, RNN, and ACLR

The rapid growth of Internet of Things (IoT) devices has led to a proportional increase in the demand for the Internet and people's overall dependence on it. Moreover, the rise of IoT malware attacks in 2018, compared to 2017, is astonishingly huge at 215.7 % thus, going from 10.3 million to 32.7 million. One can say that this is the result of the increased number of attack points that the relentless installation of IoT devices has created. IoT devices and networks are now almost completely at the mercy of hackers, and detection and mitigation techniques must be proper, quick, and effective in such situations. One of the factors that have contributed to the leaning and eventual adoption of machine learning (ML) as a possible solution is the massive amount of data produced and made available by IoT devices and networks. Hence, it is possible that ML gets widely implemented for intrusion detection in such IoT scenarios. This paper puts forward a refined ML-based system that can effectively identify attacks on IoT devices. It unites the Bayesian Optimization with Gaussian Process (BO-GP) technique and a decision trees (DT) classification model. The experiments are conducted on the Bot-IoT-2018 dataset to evaluate the performance of the proposed system. The experimental results indicate that the proposed optimized framework is effective and robust for detecting botnet attacks in IoT environments, as evidenced by high precision, accuracy, recall, and F score metrics.