论文部分内容阅读
尽管加密算法已得到改进,加密系统的安全性仍然是密码系统设计者关注的重点。边信道攻击可利用加密系统的物理漏洞来获取秘密信息。目前提出的多种边信道信息分析方法中,机器学习被认为是一种有前景的方法。基于神经网络的机器学习可获得指令标志(功耗与电磁辐射),并自动识别。本文对椭圆曲线加密(Elliptic curve cryptography,ECC)的现场可编程门阵列(field-programmable gate array,FPGA)实现展开了新的实验研究,探讨了基于学习向量量化(Learning vector quantization,LVQ)神经网络的边信道信息表征的效率。LVQ作为多类分类器的主要特点是它具有学习复杂非线性输入-输出关系、使用顺序训练程序和适应数据的能力。实验结果表明基于LVQ的多类分类是边信道数据表征的强大且有前景的方法。
Although cryptographic algorithms have been improved, the security of cryptographic systems remains the focus of cryptographic system designers. Side channel attacks exploit the physical vulnerabilities of cryptographic systems to obtain confidential information. Currently, a variety of methods of edge channel information analysis, machine learning is considered as a promising method. Machine learning based on neural networks can obtain instruction signs (power consumption and electromagnetic radiation) and automatically identify them. In this paper, a new experimental study on field-programmable gate array (FPGA) implementation based on Elliptic curve cryptography (ECC) has been carried out. Based on learning vector quantization (LVQ) neural network The efficiency of edge channel information characterization. The main feature of LVQ as a multi-class classifier is its ability to learn complex non-linear input-output relationships, use sequential training programs, and adapt to data. Experimental results show that the multi-class classification based on LVQ is a powerful and promising method for edge channel data characterization.