This project implements two real world dataset namely Modbus and Bot-Iot datasets. I've implemented a combined approach that implements both OOD denoising VAE and adversarial training as a defense mechanism against attacks like FGSM, PGD, Blackbox(uniform noise corruption), GAN-based and a Gaussian noise.