Deep-pwning is a framework for evaluating the robustness of machine learning tools against adversarial attacks. It has become widely known in the data science community that naive machine learning models, such as deep neural networks trained with the sole aim of classifying images, are very easily fooled.
The following diagram shows Explaining and Harnessing Adversarial Examples, I. J. Goodfellow et al:
Cybersecurity being an adversarial field of battle, a machine learning model used to secure from attackers ought to be robust against adversaries. As a consequence, it is important to not only report the usual performance metrics, such as accuracy, precision, and recall, but also to have some measure of the adversarial robustness of the model. The deep-pwning framework is a simple toolkit for doing so.
