Adversarial Robustness for Machine Learning