一种抗几何攻击的无载体图像隐写方法  

A coverless image steganography method against geometric attacks

作　　者：向旭宇 谭洋 秦姣华[1] 谭云[1] XIANG Xuyu;TAN Yang;QIN Jiaohua(College of Computer and Mathematics,Central South University of Forestry and Technology,Changsha 410004,China)

机构地区：[1]中南林业科技大学计算机与数学学院,湖南长沙410004

出　　处：《华中科技大学学报（自然科学版）》2024年第11期31-36,共6页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(62372478);湖南省研究生科研创新资助项目(CX20230769);中南林业科技大学研究生科技创新基金资助项目(2023CX02019)。

摘　　要：针对现有算法在传输隐写图像过程中安全性差、难以抵抗几何攻击的问题,提出一种抗几何攻击的无载体图像隐写方法.首先,利用EfficientNet模型根据输入图像构建载体图像数据集,然后对载体图像进行同心圆分块;从最内层圆向外计算子块的颜色矩分量生成鲁棒特征序列,根据图像特征值和图像路径建立倒排索引;最后,根据图像特征序列的位数对秘密信息进行分割,并根据映射规则选择合适的载体图像进行传输.实验结果表明:该方法在提升容量的同时对几何和非几何攻击均具有鲁棒性.To address the issues of poor security and difficulty in resisting geometric attacks during the transmission of steganographic images in existing algorithms,a coverless image steganography method resistant to geometric attacks was proposed.First,an EfficientNet model was used to construct a carrier image dataset based on the input images,and then the carrier images were divided into concentric circular blocks.The robust feature sequence was generated by calculating the color moments of the sub-blocks from the innermost circle outward.An inverted index was established based on the image feature values and image paths.Finally,the secret information was segmented according to the bit length of the image feature sequence,and an appropriate carrier image was selected for transmission based on the mapping rules.Experimental results show that this method enhances capacity while demonstrating robustness against both geometric and non-geometric attacks.

关 键 词：信息隐藏 无载体隐写术 颜色矩 EfficientNet模型 鲁棒性 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]