兼具鲁棒性和抗检测性的图像隐写关键技术研究

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.

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。

图像隐写通过将秘密信息嵌入到公开的数字图像中来实现隐蔽通信。现有大多数隐写算法过分的追求抗检测性，使得其实现的隐蔽通信过于脆弱。本项目分析隐写通信可能面临的各种主/被动攻击威胁，设计能力可控的隐写模型，研究多属性约束下隐写框架中的核心模块，形成一系列鲁棒性和抗检测性兼顾的隐写方案。经过三年实施，项目取得了一系列成果：构造了鲁棒性信息隐藏编码，增强了STC隐写框架的鲁棒性；构造了码率可变的自适应信息隐藏编码，可根据信道的局部特性选择嵌入率；构造了最小扰动的格QIM，减少了嵌入操作的载体扰动；构造了支持基于属性信息提取的信息隐藏编码，提供了分级、分层的隐藏保护措施。在IEEE Communication Letter、IEEE Internet of Things、西安电子科技大学学报等国内外期刊和会议上发表SCI/EI索引论文10篇；提交国内发明专利4项，澳大利亚发明专利1项，均已授权。这些成果为传统隐写术提供了各类额外的功能特性，扩宽了隐写术的使用范围。可应用在数字图像版权保护、图像数据的隐私保护等应用场所，此外，为特殊信道环境的隐蔽通信提供了可行的技术解决方案，具有一定的军事应用价值。