基于CUDA的并行数字图像加密算法研究

发布时间：2019-02-17 12:37

【摘要】：数字图像具有数据冗余大、表达信息丰富的特点,传统加密算法并没有针对这些特点进行设计,无法安全高效的完成图像加密任务,因此设计结合图像特点的加密算法是有必要的。另一方面,随着GPU通用计算技术兴起与发展,利用GPU资源对图像进行并行处理成为研究热点。基于此背景下,本文对数字图像加密技术进行研究,提出了两种并行图像加密算法。具体研究如下:(1)提出一种基于扩散码的图像加密算法。算法将扩散码构造分组密码的思想应用于图像加密中。首先引出扩散码基础结构单元概念,它采用扩散码和布尔函数相结合的方式实现信元扩散和混合过程,具有扩散速度快、迭代次数少等优点。然后利用扩散码基础结构单元作为Feistel密码结构中的F函数通过三轮迭代完成图像加密过程。实验结果表明,算法具有加密速度快、密钥空间大等优势,并能够有效抵抗统计分析攻击、穷举攻击等攻击手段。(2)提出一种基于双混沌系统的图像加密算法。算法设计了并行混沌密钥生成算法、并行置乱算法和并行扩散算法并用于图像加密中。首先在分析单混沌系统密钥空间小、安全性不足的问题后,提出了Logistic映射生成控制参数作为Skew Tent映射的初始值,混合迭代产生密钥序列的方法,并实验验证了双混沌系统良好的随机性。然后利用生成的密钥序列,采用分治策略对图像进行行(列)置乱和行(列)扩散,完成整幅图像的加密。实验表明双混沌映射产生的密钥序列能够有效提高算法的密钥敏感性,同时算法对穷举攻击、统计分析以及差分攻击等攻击手段有很强地抵抗能力。(3)上述加密算法在CUDA平台下的并行优化。针对算法和数据的并行特性,在实验基础上对数据组织、存储方案等影响因素进行优化,优化后的算法效率显著增加,第一个算法相比于CPU平台下有近180倍的加速,第二个算法也有40多倍的加速比。本文将图像数据特点、CUDA平台特点以及加密算法中速度和安全的要求综合考虑,设计了两个安全高效的图像加密算法,在并行设备下可以完成图像或视频信息的实时传输。同时,本文所做工作也推动了GPU通用计算在数字图像加密领域的研究和发展。
[Abstract]:Digital image has the characteristics of large data redundancy and rich information. The traditional encryption algorithm is not designed for these characteristics and can not safely and efficiently complete the task of image encryption. Therefore, it is necessary to design an encryption algorithm based on the features of the image. On the other hand, with the rise and development of GPU general computing technology, parallel processing of images using GPU resources has become a research hotspot. Based on this background, digital image encryption technology is studied in this paper, and two parallel image encryption algorithms are proposed. The details are as follows: (1) an image encryption algorithm based on diffusion code is proposed. The algorithm applies the idea of constructing block cipher by diffusion code to image encryption. Firstly, the concept of diffusion code infrastructure unit is introduced, which combines diffusion code and Boolean function to realize cell diffusion and mixing process, which has the advantages of fast diffusion speed and less iteration times. Then the diffusion code infrastructure unit is used as the F function in the Feistel cryptosystem to complete the image encryption process through three iterations. Experimental results show that the algorithm has the advantages of high encryption speed and large key space, and can effectively resist attacks such as statistical analysis attacks and exhaustive attacks. (2) an image encryption algorithm based on double chaotic systems is proposed. Parallel chaotic key generation algorithm, parallel scrambling algorithm and parallel diffusion algorithm are designed and used in image encryption. Firstly, after analyzing the problem that the key space of single chaotic system is small and the security is insufficient, the method of generating control parameters of Logistic map as the initial value of Skew Tent map and producing key sequence by mixed iteration is proposed. The good randomness of the double chaotic system is verified by experiments. Then, by using the generated key sequence, the row (column) scrambling and row (column) diffusion of the image are carried out by divide-and-conquer strategy to complete the encryption of the whole image. Experiments show that the key sequence generated by double chaotic mapping can effectively improve the key sensitivity of the algorithm, and the algorithm attacks exhaustive. Statistical analysis and differential attack have strong resistance to attack. (3) parallel optimization of the above encryption algorithm on CUDA platform. According to the parallelism of algorithm and data, the influence factors such as data organization and storage scheme are optimized on the basis of experiment. The efficiency of the optimized algorithm is increased significantly. The first algorithm has nearly 180 times acceleration compared with CPU platform. The second algorithm has a speedup of more than 40 times. Considering the characteristics of image data, the characteristics of CUDA platform and the requirements of speed and security in encryption algorithm, two secure and efficient image encryption algorithms are designed in this paper, which can realize the real-time transmission of image or video information in parallel devices. At the same time, the work of this paper also promotes the research and development of GPU general computing in the field of digital image encryption.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309.7;TP391.41

【参考文献】