DL-Based Transformer Architecture for Real-Time Cyberattack Prediction and Anomaly Detection (DLTA-RTCP)
Abstract
The increasing dependence on web interdependency and digital systems has resulted in a huge growth and complexity of attacks, thus demanding for smart intrusion detection system which can detect known and unknown attacks. This study introduces a Deep Learning Transformer Architecture for Real-Time Cyberattack Prediction and Anomaly Detection (DLTA-RTCP) to overcome this challenge. To address the challenges of capturing complex relationships in network traffic data and accurately identifying legitimate and malicious actions, the proposed framework employs transformer-based learning. The five traffic categories include Distributed Denial of Service (DDoS), Denial of Service (DoS), Reconnaissance, Theft and Normal traffic for multi-class classification. The performance is evaluated by utilizing Bot-IoT benchmark dataset which consists of 2,056 traffic records, and 25 informative features are selected. The data is divided into a training set and test set in the ratio of 70:30. The experimental results show that the presented framework is able to test with an accuracy of 93.97% and precision, recall, and F1 score of 83.98%, 73.64% and 75.02% respectively. A comparative analysis has also shown the superiority of DLTA-RTCP over some existing machine learning and deep learning methods such as KNN, XGBoost, GraphSAGE, GLIN, Bi-LSTM, and SVM, which validates its effectiveness in real-time detection of cyberattacks and identification of network anomalies.