http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Enhanced Adversarial Attack for Avoidance of Fake Image Detection
Kutub Uddin,Byung Tae Oh(오병태) 한국방송·미디어공학회 2023 방송공학회논문지 Vol.28 No.7
Image forensics is one of the most emerging topics in multimedia forensics to ensure the integrity of image content. Anti-forensic (AF) attacks, particularly generative adversarial network (GAN)-based attacks on fake images, can make forensic methods vulnerable. However, the effectiveness of AF attacks is limited to certain training conditions such as datasets, forensic methods, and attack types. Even though an AF attack is applied to misguide the forensic methods, forensic methods can be again updated using the AF dataset, which continues an infinite loop. This paper proposes an improved AF attack that can misguide all forensic methods. We update the forensic methods multiple times with multiple AF datasets and build an AF model that learns different forensic methods updated at different times. The experiments show that the proposed AF attack successfully deceives all forensic methods.
김동신(Dongsin Kim),오병태(Byung Tae Oh) 한국방송·미디어공학회 2021 한국방송공학회 학술발표대회 논문집 Vol.2021 No.6
In this paper, we propose a new method for inactive region padding using reinforcement learning. Inactive region is an area that has no information, such as 360 or 3DOF+ vidoes. However, these inactive regions degrade the compression performance in general. To improve the compression performance, simple filtering is applied between active and inactive regions. But it does not fully consider the characteristics of the images. In the proposed method, inactive regions are padded through reinforcement learning that can consider the characteristics of images and the compression process. Experimental results show that the performance is better than the conventional padding method.