【摘要】 针对通过滤波得到的加密图像不能有效抵抗剪切攻击的问题，提出了一种结合混沌系统、正交拉丁方和分块滤波的图像加密算法。本算法利用明文图像的SHA256哈希值和外部密钥生成内部密钥，并根据内部密钥得到混沌的初始参数值，利用混沌序列构造正交拉丁方，将正交拉丁方与图像异或，置乱拉丁方生成的矩阵对做像素扩散，再与相同的拉丁方异或，利用滤波器从图像四角分块卷积做二次扩散，最后与三个正交拉丁方异或，第一个加密图像作为主加密图像存储在第一通道中，另外两个加密图像经过旋转、分块和水平反转后作为备用图像存储在第二和第三通道中，当加密图像受到攻击，可使用滤波器加密产生的冗余像素块重构加密图像。本方法的核心优势在于构建了基于多通道冗余存储的抗剪切自修复机制。区别于传统算法，该方法创新性地将主加密图像与经几何变换的备用图像分存于三个通道，当图像遭受剪切攻击时，可利用滤波加密产生的冗余像素块进行有效重构。同时，结合明文哈希驱动的动态密钥与“拉丁方置乱-分块滤波扩散”的双重加密体系，在确保算法高敏感性和大密钥空间的基础上，实现了安全性与鲁棒性的双重提升。实验结果表明，该算法在抵抗剪切攻击、直方图分布、相关性分析、信息熵和差分攻击等评估指标上均具有良好表现，适用于高安全需求的图像保护场景。更多还原

【Abstract】 To address the vulnerability of filtered encrypted images to shear attacks, this paper proposes an image encryption algorithm that combines chaotic systems, orthogonal Latin squares, and block filtering. The algorithm generates an internal key from the plaintext image’s SHA256 hash value and an external key to determine the chaotic initial parameters. The orthogonal Latin square is constructed using chaotic sequences and XORed with the image to generate a scrambled Latin square matrix pair. Pixel diffusion is applied to this pair, followed by XORing with the original Latin square, through block convolution using corner-applied filters. Finally, the result is XORed with three orthogonal Latin squares. The primary encrypted image is stored in the first channel, while the second and third channels store backup images, generated through rotation, segmentation, and horizontal inversion. When the encrypted image is attacked, redundant pixel blocks from filter encryption can be used for reconstruction. The core advantage of this method lies in establishing a shear-resistant self-healing mechanism based on multi-channel redundant storage. Unlike traditional algorithms, this method innovatively stores the primary encrypted image and geometrically transformed backup images across three channels. When subjected to shear attacks, redundant pixel blocks generated by filter encryption enable effective reconstruction. Simultaneously, by integrating a plaintext hash-driven dynamic key with a dual-encryption system combining Latin square scrambling and block-based filtering diffusion, it achieves dual enhancements in security and robustness while ensuring high algorithm sensitivity and a large key space. Experimental results demonstrate that the algorithm exhibits robust performance across evaluation metrics including shear attacks, histogram distribution, correlation analysis, information entropy, and differential attacks, making it suitable for high-security image protection scenarios.更多还原