【作者】 孟琳；

【导师】 李琼；

【作者基本信息】 哈尔滨工业大学 ， 计算机技术， 2015， 硕士

【摘要】 随着量子计算的兴起,基于计算复杂度的现代密码学正面临严重威胁,这些加密方式在量子计算强大的并行计算能力面前不堪一击。唯一被证明绝对安全的一次加密算法又面临密钥的产生和分发问题,无法大规模应用,然而量子密钥分配,即QKD(Quantum Key Distribution)的出现使得一次加密算法有了用武之地。目前阶段大都采用FPGA(Field-Programmable Gate Array)进行QKD后处理系统的实现,但是FPGA实现QKD后处理系统存在资源有限、时钟频率有限以及数据源速率的限制,各个数据处理模块内部算法会受到硬件的制约以及信道的扰动,处理速率无法满足实际需求。本文主要研究内容是使用GPU(Graphics Processing Unit)对量子密钥分配(QKD)后处理系统的算法过程进行加速。QKD后处理系统包括五个模块,分别是筛选模块、误码估计模块、误码协商模块、保密增强模块以及信道认证模块。其中本文进行改进和加速的模块为误码协商模块和保密增强模块。误码协商模块负责纠正筛选码中的误码,输出一致的无误码,该过程可能复杂度较高,该过程的效率直接影响到整个QKD后处理系统的效率,因此,对该过程进行加速是非常有必要的。该加速过程分为三个阶段,首先本文中利用CPU对LDPC(Low Density Parity Check Code)误码协商过程进行了实现,采用了BP(Belief Propagation)算法,其次,对该算法进行改进,改进之后其效率有明显的提高,最后,本文利用GPU对该过程进行实现,效率比在CPU上算法改进后的效率有更大的提升。在量子密钥分发过程中,经过误码协商之后,由于窃听者Eve也能从量子信道以及经典信道中得到信息,所以会不安全。保密增强能够将传输中的信息泄露消除,能够提高最终生成的密钥的安全性。本文中对该部分优化时主要用到了Toeplitz矩阵和快速傅里叶变换(FFT),首先在CPU的基础上对该过程进行实现,然后利用FFT对该过程用到的算法进行改进,最后利用GPU对该过程进行加速,加速的效果非常的明显。最后,本文将上述的两个模块在GPU上进行了联调,并且与在CPU和FPGA上的速率进行了比较,分析了加速的结果。结果表明GPU加速的效果还是比较明显的,并且LDPC误码协商是整个QKD后处理系统提速的瓶颈。更多还原

【Abstract】 With the development of quantum computing, modern cryptography based on computational complexity is now seriously threatened by the strong parallel computing ability of quantum computation. The only encryption algorithm, which is proved absolutely safe, can’t be widely used because it still has some problem in key generation and distribution. However, QKD, which is short for quantum key distribution, solved this problem and made this encryption algorithm really useful. Presently, QKD processing system are mostly implemented on FPGA, but the source of FPGA is limited, clock frequency is limited and the rate of data source is technically limited, the kernel algorithm of each data processing module is limited by hardware and affected by channel disturbance, the processing speed is unable to satisfy the actual demand.The main content of this paper is to use GPU to implement QKD post-processing system and accelerate it. QKD post-processing system includes five modules, which are filter, error estimation, error reconciliation, security enhancement and channel authentication. In this paper, error reconciliation module and security enhancement module are implemented and improved.Error reconciliation module is used for correct the error in the code, and output the correct code, the process is quite complex, and it affects the efficiency of the whole system directly. Therefore, it is quite necessary to accelerate this part. The acceleration process includes three parts, first of all, the LDPC(Low Density Parity Check Code) error code reconciliation module is implemented with BP(Belief Propagation) algorithm on CPU, secondly, the process is improved, and the executing speed is highly improved, finally, the module is implemented on GPU, and the efficiency is improved again, which is better than that improved on CPU.In the process of quantum key distribution, the Eve can also obtain information from the channel, so it is not safe. Security enhancement can eliminate the information leakage in the transmission, and can improve the security of the final generated key. In this part, we mainly used the Toeplitz matrix and fast Fourier transform(FFT). Firstly, this process is implemented on CPU, then FFT is used to improve the algorithm, finally, GPU is used to speed up the process, and the effect is very significantly.更多还原