Titre :

What if Adversarial Samples were Digital Images?

Auteur(s) :

Bonnet, Benoît [Auteur]
Creating and exploiting explicit links between multimedia fragments [LinkMedia]
Furon, Teddy [Auteur]
Creating and exploiting explicit links between multimedia fragments [LinkMedia]
Bas, Patrick [Auteur]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]

Titre de la manifestation scientifique :

IH&MMSEC 2020 - 8th ACM Workshop on Information Hiding and Multimedia Security

Ville :

Denver

Pays :

France

Date de début de la manifestation scientifique :

2020-06-22

Titre de la revue :

IH&MMSec '20: Proceedings of the 2020 ACM Workshop on Information Hiding and Multimedia Security

Éditeur :

ACM

Mot(s)-clé(s) en anglais :

Image classification
adversarial samples
neural networks

Discipline(s) HAL :

Sciences de l'ingénieur [physics]/Traitement du signal et de l'image [eess.SP]
Informatique [cs]/Cryptographie et sécurité [cs.CR]

Résumé en anglais : [en]

Although adversarial sampling is a trendy topic in computer vision , very few works consider the integral constraint: The result of the attack is a digital image whose pixel values are integers. This is not an issue at ...
Lire la suite >Although adversarial sampling is a trendy topic in computer vision , very few works consider the integral constraint: The result of the attack is a digital image whose pixel values are integers. This is not an issue at first sight since applying a rounding after forging an adversarial sample trivially does the job. Yet, this paper shows theoretically and experimentally that this operation has a big impact. The adversarial perturbations are fragile signals whose quantization destroys its ability to delude an image classifier. This paper presents a new quantization mechanism which preserves the adversariality of the perturbation. Its application outcomes to a new look at the lessons learnt in adversarial sampling.Lire moins >

Langue :

Anglais

Comité de lecture :

Oui

Audience :

Internationale

Vulgarisation :

Non

Projet ANR :

Sécurité de l'Intelligence Artificielle pour des Applications Défense

Collections :

• Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189

Source :

Harvested from HAL