A Comparison Study of Generative Adversarial Network Architectures for Malicious Cyber-Attack Data Generation

The digitization trend that prevails nowadays has led to increased vulnerabilities of tools and technologies everyday life. One the many different types software attacks is botnets. Botnets enable attackers gain remote control infected machines, often leading disastrous consequences. Cybersecurity experts engage machine learning (ML) deep (DL) for designing developing smart proactive cybersecurity systems in order tackle such infections. development is, often, hindered by lack data can be used train them. Aiming address this problem, study proposes describes a methodology generation botnet-type tabular format. This involves design two generative adversarial network (GAN) models, one with six layers other eight layers, identify most efficient reliable terms similarity generated real ones. GAN models produce loops 25, 50, 100, 250, 500 1000 epochs. results are quite encouraging as, both between synthetic around 80%. eight-layer solution slightly better after running epochs, it achieved degree 82%, outperforming six-layer one, which 77%. These indicate solutions augmentation domain feasible lead new standards training trustworthy ML DL detecting tackling botnet attacks.