Welcome to Foolbox¶
Foolbox is a Python toolbox to create adversarial examples that fool neural networks.
It comes with support for many frameworks to build models including
and it is easy to extend to other frameworks.
In addition, it comes with a large collection of adversarial attacks, both gradient-based attacks as well as black-box attacks. See foolbox.attacks for details.
Foolbox is work in progress and any input is welcome. The source code can be found on GitHub.