Five examples of adversarial machine learning attacks

In this episode I go through five common/popular of the many adversarial machine learning attacks that exist: Universal Adversarial Perturbations (UAP), Fast Gradient Sign Method (FGSM), ,Projected Gradient Descent (PGD), Adversarial Patch and One Pixel Attack - plus as an added bonus, my own method - Distributed Adversarial Regions (DARs)! This is a brief summary into some of the first methods in the field, but there are over 100 different kinds of attacks now. It can feel overwhelming but I promise, if you want to learn about this field, you can! See here for more about the many other attacks that exist: https://github.com/Trusted-AI/adversarial-robustness-toolbox https://github.com/cleverhans-lab/cleverhans