Current - Issue

Modern network environments generate massive volumes of heterogeneous traffic, which creates higher risks of advanced cyber threats that develop unknown attacks. Digital Infrastructure organizations require dependable real-time intrusion detection systems to protect their data and maintain system security and service operations. Existing intrusion detection techniques, especially rule-based systems and conventional machine learning (ML) algorithms, exhibit restricted flexibility and insufficient ability to identify unknown threats. Therefore, a novel Hybrid Machine Learning and deep Learning-based Multi-class Attack Detection (Hy-MAD) framework has been developed in this paper. The method uses traditional ML classifiers with Convolutional Neural Network (CNN) technology for spatial feature extraction and Long Short- Term Memory (LSTM) networks for temporal sequence analysis. The proposed system enables accurate multi-class attack detection by employing this hybrid technology and is further integrated with an Intrusion Detection and Prevention System (IDPS) for real-time response and secure logging. The experimental findings show that the developed Hy-MAD method achieves better detection accuracy while decreasing false alarms and increasing system adaptability. The results from the experiment show that the developed Hy-MAD method achieves an accuracy of 95.8% while the random forest and CNN-LSTM models achieve respective accuracies of 78.5% and 92.3%. The Hy-MAD model achieves lower false-alarm rates of 66.7% and 50% than the existing Random Forest and CNN–LSTM models, respectively.