AN INTELLIGENT INTRUSION DETECTION SYSTEM FOR CLOUD COMPUTING ENVIRONMENTS USING SYMMETRY-PRESERVING DUAL-STREAM GRAPH NEURAL NETWORKS

Abstract

With the development of cloud computing, data transmission security continues to grow daily. The widespread use of cloud computing has made the necessity of strong data transmission security more apparent. Although these benefits, the risks related to cloud services have also increased as their usage has accelerated. Consequently, one of the most important defenses for identifying attacks in the cloud computing platform is an intrusion detection system (IDS). The huge volume of traffic present in the cloud environment presents certain difficulties for current IDSs to handle and analyze at the same time, which reduces the accuracy of detecting attacks. Therefore, a smart Intrusion Detection System for Cloud Computing Environments Using Symmetry-Preserving Dual-Stream Graph Neural Networks (IDS-CC-SPDGNN) is developed in this study. Firstly, the input data are sourced from two well-established datasets such as UNSW-NB15 and CIC-IDS2017. To ensure data quality, a Non-uniform Weighted Guided Filtering (NUWGF) technique is employed for cleaning and normalization. Following pre-processing, the Hiking Optimization Algorithm (HOA) selects the most relevant features. The proposed Symmetry-Preserving Dual-Stream Graph Neural Networks (SPDGNN) model is then utilized to distinguish network traffic as benign or malicious behavior by exploiting graph-based structural representations. To further enhance detection accuracy, the Starfish Optimization Algorithm (SOA) is applied to optimize the weight parameters of the SPDGNN. The proposed IDS-CC-SPDGNN technique attains 7.12%, 3.84% and 5.69% higher accuracy, 4.89%, 6.45% and 3.92% higher precision when compared with existing techniques. Analyses have shown that the proposed approach has proven successful in securing cloud servers from a range of possible threats.