Hybrid Attention Convolutional Neural Network and Soft Sign Long Short-Term Memory Based Cyber-Attack Detection and Classification
TL;DR
This paper proposed a hybrid attention Convolutional Neural Network (CNN) and Soft sign Long Short-Term Memory (LSTM) for effective detection and classification which ensures accurate detection and classification compared to other existing methods.
Abstract
The extensive use of smart devices and several security weaknesses of networks has intensively enhanced the number of cyber-attacks in Internet of Things (IoT) networks. The detection and classification of malicious traffic is a key to ensure the security of those systems. It aims to identify the behaviors and patterns that deviate significantly from the norm, indicating potential cyberattacks. This paper proposed a hybrid attention Convolutional Neural Network (CNN) and Soft sign Long Short-Term Memory (LSTM) for effective detection and classification. The min-max normalization is utilized in this experiment for data pre-processing and fed into Multi-Objective Sea Lion Optimization (MOSLO) based feature selection technique. Then, the selected features are given as input to the attention CNN and soft sign LSTM model. This model is estimated on NSL-KDD and ToN-IoT dataset and attains better results using accuracy, precision, recall, specificity, and F1-score. The obtained result shows that the proposed model achieves better accuracy of 99.45% on NSL-KDD dataset and 98.73% on ToN-IoT dataset which ensures accurate detection and classification compared to other existing methods like Feed Forward Neural Network (FFNN) and Improved Binary Golden Jackal Optimization-LSTM.