UPSET and ANGRI : Breaking High Performance Image Classifiers

TL;DR

This paper introduces two novel attack methods, UPSET and ANGRI, for fooling high-performance image classifiers by generating universal and image-specific perturbations, demonstrating their effectiveness on MNIST and CIFAR10 datasets.

Contribution

The paper presents two new targeted attack algorithms, UPSET and ANGRI, that effectively fool image classifiers with universal and image-specific perturbations.

Findings

Both methods successfully fool classifiers on MNIST and CIFAR10.

Universal perturbations can target specific classes effectively.

Image-specific perturbations achieve high fooling rates.

Abstract

In this paper, targeted fooling of high performance image classifiers is achieved by developing two novel attack methods. The first method generates universal perturbations for target classes and the second generates image specific perturbations. Extensive experiments are conducted on MNIST and CIFAR10 datasets to provide insights about the proposed algorithms and show their effectiveness.