adversarial

One Neuron to Fool Them All

We propose a notion of neuron sensitivity in terms of adversarial robustness, along with an attack that works as well as PGD. The notion can be extended as a regularization term, providing adversarial robustness without adversarial training.

Anshuman Suri, David Evans

A2-LINK: Recognizing Disguised Faces via Active Learning and Adversarial Noise Based Inter-Domain Knowledge

We propose an active-learning framework that utilizes studen-teacher learning and adversarial noise to achieve SOTA on inter-domain face-identification.

Anshuman Suri, Mayank Vatsa, Richa Singh

Hardening Deep Neural Networks via Adversarial Model Cascades

We propose an adversarial defense that is simultaneously robust to multiple attacks, and evaluate its efficacy over benchmark datasets.

Deepak Vijaykeerthy, Anshuman Suri, Sameep Mehta, Ponnurangam Kumaraguru