A research project about exploring the effect of adding noise on the discriminator in Generative Adversarial Networks (GAN), conducted under the supervision of Prof. Xueru Zhang. I developed an automated fuzzing system to enhance output image quality by dynamically updating noise in the discriminator. The project involved creating train and test sets with defined epochs and learning rates for both the generator and discriminator using Python. Additionally, I implemented four different types of noise in the discriminator and established a feedback mechanism for automated noise updates and monitoring loss trends.