An Optimized Hybrid Deep Learning Framework for Monitoring Botnet Attacks in IoT Networks

Abstract

The serious threat of botnet attacks in the IOT world today can be effectively addressed with deep learning (DL). However, to train the model, large and complex data sets are required, which adds cost and necessitates expensive hardware to perform intricate mathematical operations. The IoT plays a significant challenge in memory constrained by implementing the DL. The hybrid Principal Component Analysis (PCA) and Autoencoder is proposed to measures the performance to lower the IoT network traffic data and consumes bulky information. Following that, reduced datasets are examined and trained using hybrid Long Short Term Memory (LSTM) and Convolutional Neural Network (CNN) algorithms. The hybrid deep learning algorithms attain the feature to enhance the precision, accuracy, recall, and F1 scores 96.51%, 100%, 96.51%, and 98.22%, which were higher than the conventional methods. The model's performance is analyzed using extensive experiments.