摘要
密码体制识别是现实中开展密码分析的前提,属于密码学区分分析范畴.随着应用情境和通信环境日趋复杂,密码体制的识别需考虑更广范围的密码体制集合,这使得当前主流的单层识别方案的识别能力受到考验.在分析现有识别方案演变趋势的基础上,文中研究了密码体制的分层识别问题.从密码体制识别问题基本含义出发,给出了一个密码体制识别的定义系统,并将密码体制的单层识别和分层识别统一于该系统下.进一步提出一种基于随机森林的密码体制分层识别方案,通过簇分和单分两阶段,首先对密文所属密码体制类别进行识别,继而识别其具体密码体制.实验及分析在由42种密码体制产生的共41 000个密文文件组成的数据集及其子集上展开.为保证方案整体识别效果,对定义的CM-簇分、CSN-簇分和CSBP-簇分可行性进行了探讨,通过实验验证了CM-簇分和CSN-簇分的可行性,同时发现CSBP-簇分可行的一种情形.对CSBP-簇分,在该情形的3种设置下,运用所提分层识别方案开展了密码体制识别,并与4种单层识别方案进行了比较.实验表明,分层识别方案准确率比对照组中最优单层识别方案的准确率分别提高19.55%、21.40%、22.99%.
As one kind of cryptographic distinguish analysis,cryptosystem recognition is a prerequisite for real-world cryptanalysis since the most of specific cryptanalysis techniques were developed under the assumption of awareness of the exact cryptosystem that produces cipher texts.Currently,there is a great necessity for designers of recognition schemes to consider bigger sets of cryptosystems due to complex application situations and communication environments,which has influenced the recognition capability of prevalent single-stage schemes.With our analysis on the trend of the evolution of existing cryptosystem recognition schemes,we studied the twostage cryptosystem recognition problem in this paper.After revisiting the fundamental meaning of cryptosystem recognition,we proposed a definition system for cryptosystem recognition problem,which naturally unifies single-stage and two-stage recognition problems and schemes under its framework.To our best knowledge,the proposed definition system is the first speciallybuilt one for cryptosystem recognition problem that takes the recognition itself as central point,whereas the majority of previous works were conducted by simply treating cryptosystem recognition as a pattern recognition task.Further,we suggested a two-stage cryptosystem recognition scheme based on random forest.Our scheme is divided into 2 sequential procedures,‘cluster recognition'and‘single recognition'.In the stage of cluster recognition,the scheme first recognized the cluster of cryptosystems to which the cipher text belongs and actually fulfilled the task of identifying the type of cryptosystem.Then,the exact cryptosystem is recognized by single recognition within the cluster.In order to reach better recognition results,we conducted a specific research on feature extraction for the 2 recognition stages.Specifically,we selected 1 existing feature and designed 8 new features for cluster recognition,selected 5 existing features and designed 5 new features for single recognition.Experiments of this paper
出处
《计算机学报》
EI
CSCD
北大核心
2018年第2期382-399,共18页
Chinese Journal of Computers
基金
国家重点研发项目(2016YFE0100600)
信息保障技术国家重点实验室开放基金(KJ-15-008)资助~~
关键词
密码体制识别
簇分
单分
特征提取
分层识别
随机森林
cryptosystem recognition
cluster recognition
single recognition
feature extraction
2-stage recognition
random forest
