Deepfake-Image Anti-Forensics with Adversarial Examples Attacks

Many deepfake-image forensic detectors have been proposed and improved due to the development of synthetic techniques. However, recent studies show that most of these detectors are not immune to adversarial example attacks. Therefore, understanding the impact of adversarial examples on their perform...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Li Fan, Wei Li, Xiaohui Cui
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	adversarial examples deepfake general detectors Poisson noise Information technology T58.5-58.64
Acceso en línea:	https://doaj.org/article/1afc76354295434683cdaa5a75e68368
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:1afc76354295434683cdaa5a75e68368
record_format	dspace
spelling	oai:doaj.org-article:1afc76354295434683cdaa5a75e683682021-11-25T17:39:56ZDeepfake-Image Anti-Forensics with Adversarial Examples Attacks10.3390/fi131102881999-5903https://doaj.org/article/1afc76354295434683cdaa5a75e683682021-11-01T00:00:00Zhttps://www.mdpi.com/1999-5903/13/11/288https://doaj.org/toc/1999-5903Many deepfake-image forensic detectors have been proposed and improved due to the development of synthetic techniques. However, recent studies show that most of these detectors are not immune to adversarial example attacks. Therefore, understanding the impact of adversarial examples on their performance is an important step towards improving deepfake-image detectors. This study developed an anti-forensics case study of two popular general deepfake detectors based on their accuracy and generalization. Herein, we propose the Poisson noise DeepFool (PNDF), an improved iterative adversarial examples generation method. This method can simply and effectively attack forensics detectors by adding perturbations to images in different directions. Our attacks can reduce its AUC from 0.9999 to 0.0331, and the detection accuracy of deepfake images from 0.9997 to 0.0731. Compared with state-of-the-art studies, our work provides an important defense direction for future research on deepfake-image detectors, by focusing on the generalization performance of detectors and their resistance to adversarial example attacks.Li FanWei LiXiaohui CuiMDPI AGarticleadversarial examplesdeepfakegeneral detectorsPoisson noiseInformation technologyT58.5-58.64ENFuture Internet, Vol 13, Iss 288, p 288 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	adversarial examples deepfake general detectors Poisson noise Information technology T58.5-58.64
spellingShingle	adversarial examples deepfake general detectors Poisson noise Information technology T58.5-58.64 Li Fan Wei Li Xiaohui Cui Deepfake-Image Anti-Forensics with Adversarial Examples Attacks
description	Many deepfake-image forensic detectors have been proposed and improved due to the development of synthetic techniques. However, recent studies show that most of these detectors are not immune to adversarial example attacks. Therefore, understanding the impact of adversarial examples on their performance is an important step towards improving deepfake-image detectors. This study developed an anti-forensics case study of two popular general deepfake detectors based on their accuracy and generalization. Herein, we propose the Poisson noise DeepFool (PNDF), an improved iterative adversarial examples generation method. This method can simply and effectively attack forensics detectors by adding perturbations to images in different directions. Our attacks can reduce its AUC from 0.9999 to 0.0331, and the detection accuracy of deepfake images from 0.9997 to 0.0731. Compared with state-of-the-art studies, our work provides an important defense direction for future research on deepfake-image detectors, by focusing on the generalization performance of detectors and their resistance to adversarial example attacks.
format	article
author	Li Fan Wei Li Xiaohui Cui
author_facet	Li Fan Wei Li Xiaohui Cui
author_sort	Li Fan
title	Deepfake-Image Anti-Forensics with Adversarial Examples Attacks
title_short	Deepfake-Image Anti-Forensics with Adversarial Examples Attacks
title_full	Deepfake-Image Anti-Forensics with Adversarial Examples Attacks
title_fullStr	Deepfake-Image Anti-Forensics with Adversarial Examples Attacks
title_full_unstemmed	Deepfake-Image Anti-Forensics with Adversarial Examples Attacks
title_sort	deepfake-image anti-forensics with adversarial examples attacks
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/1afc76354295434683cdaa5a75e68368
work_keys_str_mv	AT lifan deepfakeimageantiforensicswithadversarialexamplesattacks AT weili deepfakeimageantiforensicswithadversarialexamplesattacks AT xiaohuicui deepfakeimageantiforensicswithadversarialexamplesattacks
_version_	1718412091446001664

Deepfake-Image Anti-Forensics with Adversarial Examples Attacks

Ejemplares similares