2020/10/13 by Warit Sirichotedumrong, Sirichotedumrong, Warit, Hitoshi Kiya +1
Computer Science · #Advanced Steganography and Watermarking Techniques #Adversarial Robustness in Machine Learning #Chaos-based Image/Signal Encryption #Cryptography and Security (cs.CR) #FOS: Computer and information sciences
paper · pdf · doi:10.48550/arxiv.2010.06198
openalex publication_date 2020/10/13 · openalex created_date 2022/09/12 · openalex updated_date 2026/07/28
Various visual information protection methods have been proposed for\nprivacy-preserving deep neural networks (DNNs). In contrast, attack methods on\nsuch protection methods have been studied simultaneously. In this paper, we\nevaluate state-of-the-art visual protection methods for privacy-preserving DNNs\nin terms of visual security against ciphertext-only attacks (COAs). We focus on\nbrute-force attack, feature reconstruction attack (FR-Attack), inverse\ntransformation attack (ITN-Attack), and GAN-based attack (GAN-Attack), which\nhave been proposed to reconstruct visual information on plain images from the\nvisually-protected images. The detail of various attack is first summarized,\nand then visual security of the protection methods is evaluated. Experimental\nresults demonstrate that most of protection methods, including pixel-wise\nencryption, have not enough robustness against GAN-Attack, while a few\nprotection methods are robust enough against GAN-Attack.\n