图像源辨识取证研究综述  

A Survey of Forensics Research on Image Source Identification Forensics

作　　者：陈艺芳 何自强[2] 文冠臣 康显桂[2] CHEN Yifang;HE Ziqiang;WEN Guanchen;KANG Xiangui(College of Cyberspace Security,Guangdong Polytechnic Normal University,Guangzhou,Guangdong 510665,China;Guangdong Key Laboratory of Information Security,Sun Yat-sen University,Guangzhou,Guangdong 510006,China)

机构地区：[1]广东技术师范大学网络空间安全学院,广东广州510665 [2]中山大学广东省信息安全技术重点实验室,广东广州510006

出　　处：《信号处理》2021年第12期2302-2322,共21页Journal of Signal Processing

基　　金：国家自然科学基金项目(62102100,61772571,62072484);国家重点研发计划课题(2019QY2203)。

摘　　要：图像源辨识取证是数字图像被动取证的一个重要研究方向,旨在从图像本身的特性出发对图像的相关来源进行识别和检测。图像源辨识取证主要包括:相机源识别、计算机图形学方法生成图像取证、AI合成图像的取证以及重获取图像的取证。其中相机源识别主要包括对采集图像所用相机的品牌、型号或个体进行识别。计算机图形学方法生成图像取证、AI合成图像的取证以及重获取图像的取证都属于非自然图像(包括生成、合成、重获取图像)与自然图像间的溯源分析。近年来,在图像源辨识取证方面的研究已经取得了一些研究成果。现有的研究方法大致分为两类,包括传统的基于模型的方法和基于深度学习的方法。本文整理了图像源辨识取证领域的研究方法,主要对研究框架和基本思路以及常用的评价指标、数据集进行了介绍,最后总结了当前图像源辨识取证的研究现状。Image source identification forensics is an important research direction of passive forensics of digital image, which aims to identify and detect the relevant sources of images based on the characteristics of the images themselves. Image source identification forensics mainly includes camera source identification, forensics of computer-generated(CG) image, forensics of AI synthesized images and forensics of recaptured image. Camera source identification mainly includes the identification of the brand, model or individual of the camera used for capturing image. Forensics of computer-generated(CG) image, forensics of AI synthesized images and forensics of recaptured image are source analysis between unnatural images(including generated, synthesized and reacquired images) and natural images. In recent years, some achievements have been made in the field of image source identification forensics. Existing research methods can be roughly divided into two categories, including traditional model-based methods and deep learning-based methods. In this paper, the research methods in the field of image source identification forensics were summarized. The research framework, basic ideas, commonly used evaluation indicators and data sets were introduced. Finally, the current research status of image source identification forensics were summarized.

关 键 词：数字图像取证 图像源辨识取证 相机源识别 图像溯源分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]