Efficient Detection of DDoS Attacks in E-Government Clouds Using Sparse Neural Networks

E-government (Electronic Government) refers to the use of information and communication technology (ICT) by governmental bodies to provide public services, manage government operations, and facilitate communication with citizens, businesses, and other government entities. E-government platforms have expanded significantly in recent years, offering a range of services such as online tax filing, license renewals, healthcare services, and more. These services, often hosted on cloud-based infrastructures, benefit from improved scalability, efficiency, and accessibility through the use of cloud computing technologies. However, the growing volume of legitimate traffic from a diverse range of users on e-government platforms makes distinguishing between normal and malicious traffic, such as DDoS (Distributed Denial of Service) attacks, increasingly difficult. To address this challenge, this manuscript proposes the use of Sparse Neural Networks (SNN) for DDoS attack detection in cloud-based e-government systems. Initially, data from the CSE-CIC-IDS2018 dataset is collected and processed through a feature selection phase. The selected features are selected using the Energy Valley Optimizer. These optimized features are fed into the Sparse Neural Network, which categorizes traffic as benign, DDoS, or DoS. To further enhance the accuracy of DDoS detection, the Ladybug Beetle Optimization (LBOA) is employed to optimize the Sparse Neural Network's parameters, ensuring optimal performance in classifying attacks. This approach aims to safeguard cloud-based e-government systems from cyber threats while maintaining reliable public service delivery. In comparison to current methods such as data clustering for the detection of DDoS attacks in the cloud for e-government (DBSCAN), e-government impact factor analysis cleverly using enhanced machine learning (ML-EGOV), and Problems with Machine Learning in Cloud Computing E-Government Models in Developing Nations, the proposed DCNN-QSA-PDOA method achieves 17.43%, 8.44%, and 11.64% higher accuracy, respectively.