A SMART ADVERSARIAL THREATS DETECTION SYSTEM USING HYBRID OF BIGAN-BIGRU TECHNIQUE

Subject Area: Cybersecurity

The study presents modelling a smart adversarial threats detection system using hybrid of Bi-Directional Generative Adversarial Network (Bi-GAN) and Bi-Directional Gated Recurrent Unit (Bi-GRU). The aim is to integrate Bi-GAN model in the network to capture complex and dynamic threat patterns. The primary objective is to develop a data model that effectively processes network threat features and generates a high-performing detection rate for adversarial threats. The methodology used is dynamic system development method. The collected threat data were pre-processed using analysis of variance technique for feature selection and then Convolutional Neural Network (CNN) for feature transformation. The features were then used to train the Bi-GAN, which was designed to learn temporal and sequential characteristics of adversarial attack patterns, making it ideal for real-time network monitoring. The model was implemented using python programming language and simulated in Kali-linux and Virtual lab, environment. The result of the study suggests that the Bi-GAN model is highly effective at detecting adversarial attacks. The results demonstrated that the Bi-GAN model considered adversarial attack vector such as denial of service, ip-sweep, flood, sql-injection attack and legitimate packets. The experimental findings showed that the proposed system could reliably identify all the attack vectors IP addresses and classifying them as threat, while allowing throughput for only the legitimate packets. In conclusion, this research contributes to the field of cybersecurity by providing a novel framework for addressing the limitations of existing intrusion detection systems.