Cleverhans

When it fits

Use CleverHans when you have a machine-learning model and want reference attack or defense implementations for adversarial-robustness experiments.

Who it is for

• Machine-learning researchers evaluating model robustness
• Developers constructing adversarial examples or prototyping defenses
• Educators and learners following the project’s MNIST and CIFAR-10 tutorials

Documented uses

• Benchmark a model’s vulnerability to adversarial examples.
• Construct adversarial examples using attack implementations.
• Explore or evaluate defenses against adversarial attacks.
• Study FGSM and PGD examples in the supplied tutorials.