Use the Jupyter Notebook to integrate adversarial attacks into a model training pipeline to detect vulnerabilities

Description

With great power comes great responsibility. Adversarial examples in AI pose an asymmetrical challenge with respect to attackers and defenders. AI developers must be empowered to defend deep neural networks against adversarial attacks and allow rapid crafting and analysis of attack and defense methods for machine learning models.

Animesh Singh and Tommy Li explain how to implement state-of-the-art methods for attacking and defending classifiers using the open source Adversarial Robustness Toolbox. The library provides AI developers with interfaces that support the composition of comprehensive defense systems using individual methods as building blocks. Animesh and Tommy then demonstrate how to use a Jupyter notebook to leverage attack methods from the Adversarial Robustness Toolbox (ART) into a model training pipeline on Fabric for Deep Learning (FfDL). This notebook trains a CNN model on the Fashion MNIST dataset, and the generated adversarial samples are used to evaluate the robustness of the trained model.