# Multi-way Encoding for Robustness

## TL;DR

This paper introduces multi-way encoding as a novel output representation to enhance the robustness of deep models against adversarial attacks in computer vision tasks.

## Contribution

It proposes replacing one-hot encoding with multi-way encoding to decorrelate models and improve security against adversarial examples.

## Key findings

- Improved robustness against black-box and white-box attacks.
- Effective on multiple benchmark datasets including MNIST and CIFAR.
- Raises challenges in model watermarking detection.

## Abstract

Deep models are state-of-the-art for many computer vision tasks including image classification and object detection. However, it has been shown that deep models are vulnerable to adversarial examples. We highlight how one-hot encoding directly contributes to this vulnerability and propose breaking away from this widely-used, but highly-vulnerable mapping. We demonstrate that by leveraging a different output encoding, multi-way encoding, we decorrelate source and target models, making target models more secure. Our approach makes it more difficult for adversaries to find useful gradients for generating adversarial attacks. We present robustness for black-box and white-box attacks on four benchmark datasets: MNIST, CIFAR-10, CIFAR-100, and SVHN. The strength of our approach is also presented in the form of an attack for model watermarking, raising challenges in detecting stolen models.