- Conference Article
- 10.1109/icdiss68238.2025.11320686
Generative Adversarial Defense Models for Simulated Cyberattack Scenarios in Virtual Networks
- Nov 14, 2025
- Karthick Cherladine + 5 more +5
This research discussed an idea for a defensive system using a GAN-based model to protect against cyber-attacks on virtual network systems. The plan was to train the GAN-based model using simulated attacks created by the attacker generator portion of the model, so when it is exposed to a new real cyber-attack, it will be able to recognize and block it. Both portions of the model (the attacker and defender) were designed to learn from each other during the training process. As demonstrated via simulation within MATLAB, the overall accuracy of the model increased as training progressed and it began to identify increasingly complex attacks. Additionally, it was found that the model trained at a faster pace than previously experienced and produced fewer errors than before training. Ultimately, this research demonstrated that <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$A I$</tex> is capable of adapting into a defensive thinking mode and not simply passively detecting attacks. It is somewhat similar to developing a “smart wall” where after each successful cyber-attack, the wall becomes stronger due to continuous learning.
Read more