图像隐写通过将秘密信息嵌入到公开的图像数据中来实现隐蔽通信，其能提供传输行为的隐私保护，因此备受各国重视。然而传统的图像隐写难以抵御网络空间中多样化的攻击手段。因此，本项目将深入分析隐蔽通信可能面临的各种主/被动攻击威胁，扩展原有隐写理论模型，研究多属性约束下隐写嵌入框架中的各核心模块，形成一系列鲁棒性和抗检测性兼顾的图像隐写方案。项目具体研究：1）能力可控隐写理论模型，设计场景关联的隐蔽通信模型，并估计理论嵌入容量上界；2）载体鲁棒性嵌入域选择，通过载体图像、嵌入位置、变换域的选择增强对各类攻击的鲁棒性；3）容错性自适应隐写编码，实现隐写编码的鲁棒性增强，及信息隐藏编码的抗检测性增强；4）多维度嵌入扰动度量，设计出描述准确、形式精简的扰动度量函数。其研究能够解决鲁棒性图像隐写实际应用中的某些技术难题，有助于构建新时期网络环境下安全、可靠的隐蔽通信。

Image steganography is a technique for covert communication. It aims to conceal secret information into published digital images without causing suspiciousness. Image steganography has received wide attention due to its ability of preserving the privacy of communication behavior. However, the diverse attacks in the fast-growing cyberspace make the typical image steganography seem inadequate. In view of this, this project tries to develop a series of image steganographic schemes with high robustness and high undetectability. The project analyzes the potential active/passive attacks, extends the typical steganographic model, and studies the core modules in the image steganographic framework constrained by robustness, undetectability, imperceptibility, etc. The objective of this project is to: 1) design controllable steganographic model, which models the covert communication depending on the real scenario, and estimates its maximum capacity, 2) extract robust embedding regions by selecting suitable cover images, embedding locations, and transformed domains, 3) construct fault-tolerant steganographic codes by increasing the robustness of steganographic codes or increasing the undetectability of data hiding codes, 4) develop accurate distortion metrics with concise forms to measure multiple attributes. The project can solve technical challenges in the applications of robust image steganography and give essential foundation of methods for the secure and reliable covert communication.