What if Adversarial Samples were Digital Images?

Benoît Bonnet (1) , Teddy Furon (1) , Patrick Bas (2)

1 LinkMedia - Creating and exploiting explicit links between multimedia fragments
2 CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189

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.

Mots clés: Image classification; adversarial samples; neural networks

Domaines: Traitement du signal et de l'image [eess.SP]; Cryptographie et sécurité [cs.CR]

Benoît Bonnet, Teddy Furon, Patrick Bas. What if Adversarial Samples were Digital Images?. IH&MMSEC 2020 - 8th ACM Workshop on Information Hiding and Multimedia Security, Jun 2020, Denver, France. pp.1-11, ⟨10.1145/3369412.3395062⟩. ⟨hal-02553006v2⟩