Hybrid Deep Learning Framework for Intelligent Cybersecurity Threat Detection

Abstract

This paper proposes a hybrid deep learning framework designed to enhance cybersecurity threat detection in enterprise networks. As cyber threats become increasingly sophisticated, traditional signature-based detection methods often fail to identify novel, zero-day attacks. We introduce a model combining Convolutional Neural Networks (CNNs) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal dependency analysis. Experimental results demonstrate that our proposed architecture achieves a detection accuracy of 98.4%, significantly outperforming traditional ResNet-50 models. The framework effectively mitigates false positives, providing a robust solution for real-time network traffic monitoring. We conclude that integrating spatial and temporal learning features is essential for modern threat intelligence systems.