A Hybrid Machine Learning Approach for the Efficient Detection of Botnet Attacks in IoT Environments

Abstract

The rapid advancement of internet technologies has led to a significant rise in
cyber-attacks, with botnet attacks emerging as one of the most serious threats. IoT devices, due
to their limited built-in security mechanisms, have become easy targets for botnet infections. As
botnets continue to evolve with complex and dynamic attack behaviors, traditional identification
methods struggle to detect malicious activities effectively. The objective of this research is to
develop an efficient and intelligent botnet detection model capable of accurately identifying
multiple attack types in IoT environments. Manual systems rely heavily on predefined patterns
and human intervention. Manual systems are time-consuming, prone to errors, and inefficient for
large-scale IoT environments. To address these challenges, the proposed system introduces a
hybrid Machine Learning model named ACLR, which combines ANN, CNN, LSTM, and RNN
into a unified stacked architecture. Using the UNSW-NB15 dataset containing nine different
attack categories, the proposed model demonstrates strong detection capability and effectively
identifies complex botnet behaviors. These outcomes highlight the model’s ability to enhance the
security of IoT environments by providing reliable and automated detection of malicious
network activities