Generative Adversarial Network (GAN) for Attack Simulation
Developed a GAN framework to simulate realistic cybersecurity attack vectors using the DNN-EdgeIIoT dataset. Engineered an adversarial training pipeline with Generator and Discriminator models to augment the dataset with synthetic samples and address data imbalance. Evaluated performance using a Random Forest Classifier, achieving 97.19% accuracy and an 8.7% improvement over baseline.