Adversarial Robustness Toolbox (ART) provides tools that enable developers and researchers to evaluate, defend, certify and verify Machine Learning models and applications against the adversarial threats. 

GitHub: https://github.com/Trusted-AI/adversarial-robustness-toolbox