Robust Machine Learning In Adversarial Setting With Provable Guarantee