MediaEval 2019: Concealed FGSM Perturbations for Privacy Preservation

TL;DR

This paper presents methods to generate concealed FGSM perturbations that hide images from scene classifiers while maintaining image quality, advancing privacy-preserving techniques in image processing.

Contribution

It introduces two novel approaches to minimize aesthetic damage caused by FGSM perturbations, enhancing privacy preservation without compromising image appeal.

Findings

Effective concealment of images from classifiers

Preservation of image aesthetic quality

Open-source code availability

Abstract

This work tackles the Pixel Privacy task put forth by MediaEval 2019. Our goal is to manipulate images in a way that conceals them from automatic scene classifiers while preserving the original image quality. We use the fast gradient sign method, which normally has a corrupting influence on image appeal, and devise two methods to minimize the damage. The first approach uses a map of pixel locations that are either salient or flat, and directs perturbations away from them. The second approach subtracts the gradient of an aesthetics evaluation model from the gradient of the attack model to guide the perturbations towards a direction that preserves appeal. We make our code available at: https://git.io/JesXr.