Repro - On the Existence of Consistent Adversarial Attacks in High-Dimensional Models