基于字符画的生成式伪装方法

A Generative Disguise Method Based on Character Painting

传统基于马赛克拼图的信息隐藏需借助修改式嵌入的方式嵌入额外参数,难以抵抗密写分析且鲁棒性较低。针对上述问题,提出基于字符画的生成式图像信息隐藏方法。首先,利用密钥生成随机坐标序列来决定秘密比特在掩体图像中的嵌密位置;之后,遍历掩体图像像素,对于嵌密和非嵌密位置,分别放置与密钥和位置信息相关联或与掩体放置位置像素点相近的字符对秘密比特进行表达和掩盖;最后将字符放置过程产生的偏差利用误差扩散传递给掩体周围未处理像素,从而产生含密字符画图像。该方法仅利用随机放置的字符对秘密信息进行表达而不涉及任何参数的额外嵌入,因而不会遗留修改痕迹;同时该方法提取策略对非显著比特位不敏感,可抵抗一定程度的噪声攻击和低质量的JPEG压缩。实验结果表明,该方法具备良好的鲁棒性,且完全依赖于密钥,可产生高质量的含密字符画,可抵御信道攻击并具备较高的安全性。

Traditional mosaic-based information hiding involves the embedding of additional information by modification-based methods,which makes it hard to resist steganography and also causes low robustness.To solve these problems,we propose a generative disguise method based on character painting.Firstly,random integer position sequence is generated by key to locate the hiding position of secret bit in carrier image.Secondly,the carrier image is traversed,and characters placed in hiding positions are related with key and locations to express secret information while characters placed in non-hiding positions are similar to the corresponding carrier pixels to cover secret information.Finally,the deviation caused by character placement process is scattered to the around unprocessed pixels by error diffusion method to form the character painting hiding secret.The proposed method only uses randomly placed character to express secret information without modification trace,which can resist steganography.Moreover,the extraction of this method is insensitive to LSB,so it has great ability against a certain strength of noise attacks and low-quality JPEG compression.Experiment shows that the proposed method has strong robustness and fully depends on key.Hence,it can hide the secret bits in high quality character painting image and resist channel attacks with high security.