摘要
现有的密码算法识别方案基于信息熵和随机性检测方法设计密文特征,存在识别准确率较低的问题。按照熵值估计方法提取密文特征,采用包含逻辑回归、支持向量机和决策树在内的5种常用机器学习算法,对DES、AES、3DES、Blowfish和CAST共5种分组密码进行分类实验。实验结果表明,基于熵源验证的识别方案能够对分组密码的工作模式进行有效区分,分类准确率达99%。同时,在ECB模式下对DES和AES的二分类识别准确率达95%,五分类实验识别准确率达62.7%,高于基于随机性检测识别方案的75%和52%。研究表明,使用熵源验证方法可以丰富密文特征库,提高密码算法识别准确率。
The existing cryptographic algorithm identification schemes are mainly based on information entropy and randomness testing methods to design ciphertext features,which has the problem of low classification accuracy.In this paper,ciphertext features are extracted according to the entropy estimation method,and five common machine learning algorithms including logistic regression,support vector machine,and decision tree are used to conduct classification experiments on five block ciphers DES,AES,3DES,Blowfish,and CAST.The experimental results show that the recognition scheme based on entropy source validation can effectively distinguish the working modes of block ciphers,with a classification accuracy of up to 99%.Meanwhile,the binary classification recognition accuracy of DES and AES with ECB mode is as high as 95%,and the recognition accuracy of the five classification experiments reaches 62.7%,outperfoming the 75%and 52%achieved by schemes relying solely on randomness detection.This research shows that the use of entropy source verification method can enrich the ciphertext feature library and improve the recognition accuracy of cryptographic algorithms.
作者
张家渟
李莘玥
顾纯祥
ZHANG Jiating;LI Xinyue;GU Chunxiang(Information Engineering University,Zhengzhou 450001,China;Henan Key Laboratory of Network Cryptography Tech-nology,Zhengzhou 450001,China;Anhui Normal University,Wuhu 241000,China)
出处
《信息工程大学学报》
2024年第4期472-477,共6页
Journal of Information Engineering University
基金
国家自然科学基金(61772548,23456789)
河南省优秀青年基金(222300420099)。
关键词
密码算法识别
特征提取
熵源验证
机器学习
随机性检测
cryptographic algorithm identification
feature extraction
entropy source validation
machine learning
randomized detection
作者简介
张家渟(1995-),男,硕士生,主要研究方向为基于机器学习的密码算法识别。E-mail:1093702691@qq.com。
