In the recent past, machine learning has been proven  to be susceptible to carefully crafted adversarial examples. Here is an overview of the most common adversarial attacks in white and black box setting.