Abstract: For most of chaotic image encryption methods, pixel is often seen as a whole unit to process in pixel value scrambling. It is very weak to resist known cleartext attack. Aiming at the problem, a high dimension chaotic encryption method based on discrete Hopfield feedback neural network is proposed. In the process of pixel value scrambling, each bit of pixel is the handle object. So the scrambling of each pixel value is a composite process. The validity of this method is verified through experimental results.

Key words: feedback neural network, chaos, Lorenz system, image encryption, scrambling

摘要： 大部分混沌图像加密方法在进行图像像素值置乱时采用像素值整体处理方式（如异或运算等），这在抵抗已知明文攻击时是非常脆弱的。针对这个问题，提出了一种基于离散Hopfield反馈神经网络的高维混沌图像加密方法。在进行图像像素值置乱时，操作对象是每个像素的每个比特位，每个像素单元的置乱是一个复合处理过程。通过实验验证了该方法的有效性。

关键词: 反馈神经网络, 混沌, Lorenz系统, 图像加密, 置乱