Devising classification algorithms that are robust to worst-case perturbations has emerged as a challenging problem in theoretical machine learning. In this work, we study whether there is any learning task for which it is possible to design classifiers that are only robust against polynomial-time adversaries; just like how it is done cryptography. Indeed, numerous cryptographic tasks (e.g. encryption of long messages) are only secure against computationally bounded adversaries. We show that computational limitation of attackers can indeed be useful in robust learning by demonstrating a classifier for a learning task in which computational and information theoretic adversaries of bounded perturbations have very different power. Namely, while computationally unbounded adversaries can attack successfully and find adversarial examples with small perturbation, polynomial time adversaries are unable to do so unless they can break standard cryptographic hardness assumptions (in particular digital signatures). No background on machine learning or cryptography will be assumed. Joint work (to appear in ALT 2020) with Sanjam Garg, Somesh Jha, and Saeed Mahloujifar.