NFT图像隐写

NFT(non-fungibletoken)图像为元宇宙中进行创作、交易、分享和收藏的数字艺术作品。不同于自然图像，NFT图像的内容为用户自主定义，在数据空间分布较广，这为秘密信息的隐藏提供了极大便利，因此借助NFT图像进行隐蔽通信成为图像隐写的一个新分支。提出了一种NFT图像的隐写方法。对一幅给定的NFT图像，对高频与边缘轮廓部分进行增强，以丰富图像中有利于掩盖隐写修改痕迹的细节部分，从而使增强后的图像更适合隐写，将其作为载体。根据增强前后图像像素的差异确定载体图像各像素加1或减1的倾向修改方向，并根据此差异调整载体各像素的修改代价以满足确定的倾向修改方向，进一步提升隐写抗检测性。利用主流隐写编码框架在载体图像中进行信息嵌入。实验结果表明，所提方法应用于NFT图像时的抗检测性优于现有的数字图像隐写方案，对于HILL、Mi POD、DFEI隐写方案，所提方法可分别将隐写分析错误率（PE值）平均提升8.7%、9.2%、6.2%（所有嵌入率与隐写分析特征情况平均值）。所提方法适用于NFT图像，为除自然图像与生成图像以外的第3类载体（即NFT图像）提供了针对性的隐写方法。待NFT图像数量较为...

Steganography in NFT images

The images with non-fungible token (NFT) are employed as the digital artistic works in metaverse for creation, transaction, sharing, and collection.Being different from natural images, the content of NFT images is defined by user and distributed in the digital space widely.It is convenient for the hidden of secret data.In this case, covert communication with NFT images is a new branch of image steganography.Then, a steganographic method for NFT images was proposed accordingly.Given a NFT image, the regions of its profile and the components with high frequency were enhanced firstly to enrich the details which were beneficial to hide the modification trace of steganography.In this way, the enhanced image was used as cover since it is more suitable for steganography.Then, the tendency modification direction of each pixel was determined by the differences between the enhanced image and the given image.The differences were also used to determine the cost value of modification amplitude.Thus, the undetectability of steganography can be increased further.Secret data was embedded into the cover image using the popular steganographic coding schemes.Experimental results showed that the proposed method had imporoved undetectability on NFT images compared with existing digital steganographic schemes.Compared with HILL, MiPOD, and DEFI, the proposed method can increase the detection error P_E of steganalysis by 8.7%, 9.2% and 6.2%, respectively (the average value for the cases of different payload and steganalytic features).Therefore, the proposed method is suitable for NFT images and it provides targeted steganographic method for the third kind of images, i.e., NFT images, except of natural images and generated images.For further study, the deep learning-based steganographic method can be designed for NFT images using the strong fitting and learning ability of neural networks.