Enhanced Model For Intrusion Detection In A Cloud Environment

Abstract

The cyber security industry is rapidly growing due to the expanding size of networks, which increases the attack surface for hackers. As cyber threats become more sophisticated, defenses must also advance. This research employs the Zhang 2019 model, utilizing a Deep Generative Neural Network (DGNN) for adversarial learning. DGNN comprises two components: G-net, which generates malicious inputs (intrusion samples), and D-net, which identifies these from real inputs. Over time, G-net produces more sophisticated malicious inputs as D-net improves its detection. The developed model monitors and analyzes data in a cloud computing environment, detecting intrusions from both external and internal users using Support Vector Machine and Bayesian network techniques. Data is collected, pre-processed, normalized, and classified into normal (ND) or intrusion data (ID). The normal data is treated as regular network requests, while intrusion data undergoes further processing before entering the deep learning training model. 

In the testing phase, a test data was obtained with the use of data packet receiver and classification of the intrusions as True positive, false positive, true negative and false negative. Each frame was analyzed for true positive and true negative, the result gotten from the classification and analyses of the intrusions were used to compare and evaluate the performance of the enhanced intrusion detection model to ascertain if it was better than the Zhang model. The enhanced model effectively predicts and detects intrusion attacks and unauthorized users, providing valuable information about malicious network traffic and alerting security personnel to potential invasions.