Detection of Cyber-Attack Using Deep Learning
Author : Dr. S Krishnaveni and V Elakkiya
Abstract :
Deep learning-based framework for cyber-attack detection that integrates Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The proposed hybrid model leverages the CNN’s ability to extract spatial features and the LSTM’s strength in learning temporal dependencies from sequential network traffic data. The dataset, consisting of features such as protocol type, service, flag, and byte-level attributes, is preprocessed through label encoding, normalization, and reshaping to suit the CNN-LSTM architecture. Experimental evaluation demonstrates that the model achieves high classification accuracy while minimizing false positives, outperforming conventional intrusion detection methods. Performance is validated using confusion matrices, classification reports, ROC curves, and violin plots, confirming the robustness of the approach. This study highlights the potential of CNN-LSTM models in providing intelligent, scalable, and reliable solutions for strengthening cybersecurity in modern digital infrastructures.
Keywords :
Cyber Attack Detection, Intrusion Detection System (IDS), CNN-LSTM, Deep Learning, Anomaly Detection, Network Security, Machine Learning, ROC-AUC, Classification Accuracy.
