Hybrid CNN-LSTM for Cyberattack Classification in Sudanese E-Government SDN

Abstract Sudan is going through a fast digital transformation that is promoting the widespread adoption of digital governmental services (e-government) using Advance Primary network architectures, including software-defined networks (SDN).However this change reveals the main network, network infrastructures and large national web portals, portals to sophisticated cyber attacks. the purpose of this study is to develop an intelligent deep learning methodology for identifying cyber attacks towards e-government services in Sudan and reduce these vulnerabilities. We propose a hybrid deep learning architecture that combines long short-term memory (LSTM) networks, and convolutional neural networks (CNNs). The model was trained and tested on a custom hybrid dataset. This dataset is a mixture of web datasets (CSIC HTTP 2010, ECML/PKDD 2007), global network benchmarks (CIC-IDS2017) and virtual SDN traffic generated by NS-3 simulator to address the absence of domain specific data. Results from the experiment indicate that the proposed CNN-LSTM model performs significantly better than the collective baselines and traditional machine learning, via an overall accuracy of 97.8%. Moreover, the model achieves optimal averaging prediction speeds and a False Positive Rate (FPR) of less than 1.5% presenting a scalable AI-driven baseline to enhance the cybersecurity and privacy position of the Sudanese government