Deep Learning-Based Detection of IoT Botnet Attacks: An Exploration of Residual Networks

Abstract

Modern enterprises increasingly employ Internet of Things (IoT) devices across various sectors to enhance service provision, with applications spanning from healthcare to academia. However, the widespread adoption of IoT technology introduces significant security vulnerabilities. Particularly, these devices are susceptible to cyber-attacks, notably those orchestrated by botnets. The challenge of addressing this security issue is further compounded by the devices' memory and energy constraints, which limit the implementation of robust security measures. The present study introduces a Deep Learning Techniques (DLT) based approach, termed Detection of Intrusions in IoT using Residual Networks (DIIOTRNs), to preemptively identify IoT botnet attacks. These attacks typically undergo several stages prior to execution, providing an opportunity for early detection. The proposed DIIOTRNs framework integrates Convolution Neural Networks (CNNs) and Long Short-Term Memories (LSTMs) to effectively detect potential threats. The framework was subjected to empirical testing and demonstrated promising results, achieving accuracy levels exceeding 90%. Thus, the DIIOTRNs approach offers a promising solution to the pressing issue of IoT security, particularly in the context of botnet attacks. Further research is warranted to refine and optimize this framework for broad adoption across the IoT landscape.