Extending the Scope of Provable Adversarial Robustness in Machine Learning