本项目针对目前网络虚假图像泛滥，给网络空间治理带来严峻挑战的现状，将图像溯源取证放到社交网络的大环境中，从图像篡改检测、图像溯源和推理取证三个层面进行研究。首先面向不同形式的篡改手段，挖掘图像与多源辅助信息之间的统计相关性，提出基于一致性度量的图像篡改检测机制；其次构建多维度全覆盖的图像溯源模型，包括深度局部特征提取和乘积量化的原图溯源，基于图卷积传播路径重构的用户溯源，基于编解码方式差异性的设备溯源；最后，提出构建跨模态、深语义的取证知识图谱和全网络关系计算与深度推理取证方法。本项目将在真实数据环境下验证系统性能，以期在数字图像溯源取证的理论和关键技术方面取得突破，为网络空间安全、智慧司法提供强有力的理论支撑和技术保障。

Facing to the problem of fake images spreading on the Internet, which brings severe challenges to cyberspace supervision, this project puts image tracing analysis and forensics into the environment of social networks and aims to solve this problem separately from image tamper detection, image tracing and reasoning forensics. Firstly, to deal with different kinds of tampering, we explore the statistical correlation between images and multi-source auxiliary information, and put forward image tamper detection mechanism based on consistency metric. Secondly, we construct a multi-dimensional full coverage image tracing model, including deep local feature extraction and product quantization-based original image tracing, graph convolutional networks-based image propagation path reconstruction and user tracing, and coding and decoding difference-based image acquisition device tracing. Finally, we propose building a cross-modal deep semantic forensics knowledge graph and a full network relational calculation and deep reasoning forensics method. This project will verify its performance with real large-scale network data. It can obtain a breakthrough in the key techniques of digital image tracing analysis and forensics, and provide strong theoretical and technical support for cyberspace security and intelligent judiciary.