Wu Tian, Hu Xuan, Liu Chunnian
Digital Literacy and Skills Enhancement Research Center, Jiangxi Province Philosophy and Social Science Key Research Base, School of Public Policy and Administration, Nanchang University, Nanchang, 330031, China.
Sci Rep. 2024 Feb 28;14(1):4826. doi: 10.1038/s41598-024-55474-y.
Multi-image steganography, a technique for concealing information within multiple carrier mediums, finds remote sensing images to be particularly apt carriers due to their complex structures and abundant texture data. These characteristics bolster the resilience against steganalysis and enhance steganographic capacity. The efficacy of multi-image steganography hinges on the diplomatic strategy of cover selection and the meticulous allocation of the payload. Nevertheless, the majority of current methods, which are empirically formulated, predominantly focus on the texture complexity of individual images, thereby potentially undermining overall security. This paper introduces a security-oriented approach to steganographic payload allocation for multiple remote sensing images aimed at fortifying the security of multi-image steganography. Our primary contributions include employing a steganalysis pre-trained network to quantify texture complexity in remote sensing cover images, directly correlating it with security. Additionally, we have developed an adaptive payload allocation strategy for multiple images, which embeds a payload proximate to each image's maximal steganographic capacity while concurrently ensuring the security of the embedding process. Experimental results corroborate that our methodology excels in cover selection and payload allocation and achieves better undetectability against modern steganalysis tools.
多图像隐写术是一种在多个载体介质中隐藏信息的技术，由于其复杂的结构和丰富的纹理数据，遥感图像被认为是特别合适的载体。这些特性增强了对隐写分析的抵抗力，并提高了隐写容量。多图像隐写术的有效性取决于封面选择的策略和有效载荷的精心分配。然而，目前大多数基于经验制定的方法主要关注单个图像的纹理复杂性，从而可能损害整体安全性。本文介绍了一种面向安全的多遥感图像隐写有效载荷分配方法，旨在加强多图像隐写术的安全性。我们的主要贡献包括使用一个经过隐写分析预训练的网络来量化遥感封面图像中的纹理复杂性，并将其与安全性直接关联。此外，我们还开发了一种针对多图像的自适应有效载荷分配策略，该策略在嵌入有效载荷时接近每个图像的最大隐写容量，同时确保嵌入过程的安全性。实验结果证实，我们的方法在封面选择和有效载荷分配方面表现出色，并且在面对现代隐写分析工具时具有更好的不可检测性。
