基于seq2seq模型的深度学习密码破译方法

Deep Learning Password Deciphering Method based on Seq2seq Model

随着信息安全受到人们越来越多的关注,信息的加解密问题成为当今研究的热点。通过将密码破译问题转化为机器翻译问题,设计了一种基于深度学习的密码破译方法。首先,将明文密文对看作为翻译对,使其更合适地用于本文所用的翻译模型。其次,采用词嵌入编码方式对输入数据进行编码,在确定数据输入形式的同时,保留序列之间的相关信息。最后,使用3种典型的多表置换密码验证模型的性能。仿真结果表明,无论哪一种加密算法,所提模型都具有良好的破译效果,破译准确率可以达到99%。

With more and more attention paid to information security, the problem of encryption and decryption of information has become a hot topic in current research. By transforming the password deciphering problem into machine translation problem, a deep learning based cryptographic decoding method is designed. Firstly, the plaintext ciphertext pairs are regarded as translation pairs to make them more suitable for the translation model used in this paper. Then, the input data is encoded by the word embedding coding method, and the relevant information between the sequences is retained while the data input form is determined. Finally, three typical multi-table permutation cryptosystems are used to verify the performance of the model. The simulation results indicate that no matter which encryption algorithm, the proposed model has good deciphering effect and the decoding accuracy can reach 99%.