Scholarlink

<p>In CRYPTO 2019, Gohr's work introduced a transformative application of deep learning methodologies in differential cryptanalysis of SPECK32/64. This breakthrough significantly enhanced the accuracy of differential cryptanalysis beyond the capabilities of conventional differential distinguishers, thereby marking a pivotal shift in the field of cryptography. In recent years, neural differential cryptanalysis has become the mainstream research hot spot of neural-assisted cryptanalysis. The Cambridge Laboratory initially proposed the Tiny Encryption Algorithm (TEA) family of block ciphers. They adopt a block size of 64 bits and a key size of 128 bits, which have been extensively studied due to their relevance in various cryptographic applications. This paper constructs the current optimal neural distinguisher of the TEA family block cipher algorithms. First, by traversing all differences with Hamming weights of 1 as input difference, find the input difference of the optimal neural distinguisher for TEA, Extended Tiny Encryption Algorithm (XTEA), and Corrected Block Tiny Encryption Algorithm (XXTEA). Based on them, construct 9 rounds (4.5 encryption cycles) of the effective neural distinguishers for TEA and XTEA, and 4-cycle distinguishers of XXTEA with an accuracy of 58.37, 52.28, and 95.77%, respectively. Furthermore, by utilizing SENet and multiple ciphertext input data formats, we further increase the above neural distinguishers so that accuracies achieve 77.22, 76.48, and 98.36%, respectively. Finally, the accuracy of the neural distinguisher is significantly higher than that of traditional differential distinguishers, according to the experiments. In conclusion, these results represent a significant contribution to the field of cryptography by demonstrating the effectiveness of advanced machine learning techniques in enhancing the performance of cryptanalysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Construction of neural distinguishers of the tiny encryption algorithm family block ciphers

  • 期刊论文

  • 发表日期：

    2025年7月31日

查看全文
  • Li Liu,
  • Weiping Zhou,
  • Zhen Shen,
  • Yong Zhu

摘要

In CRYPTO 2019, Gohr's work introduced a transformative application of deep learning methodologies in differential cryptanalysis of SPECK32/64. This breakthrough significantly enhanced the accuracy of differential cryptanalysis beyond the capabilities of conventional differential distinguishers, thereby marking a pivotal shift in the field of cryptography. In recent years, neural differential cryptanalysis has become the mainstream research hot spot of neural-assisted cryptanalysis. The Cambridge Laboratory initially proposed the Tiny Encryption Algorithm (TEA) family of block ciphers. They adopt a block size of 64 bits and a key size of 128 bits, which have been extensively studied due to their relevance in various cryptographic applications. This paper constructs the current optimal neural distinguisher of the TEA family block cipher algorithms. First, by traversing all differences with Hamming weights of 1 as input difference, find the input difference of the optimal neural distinguisher for TEA, Extended Tiny Encryption Algorithm (XTEA), and Corrected Block Tiny Encryption Algorithm (XXTEA). Based on them, construct 9 rounds (4.5 encryption cycles) of the effective neural distinguishers for TEA and XTEA, and 4-cycle distinguishers of XXTEA with an accuracy of 58.37, 52.28, and 95.77%, respectively. Furthermore, by utilizing SENet and multiple ciphertext input data formats, we further increase the above neural distinguishers so that accuracies achieve 77.22, 76.48, and 98.36%, respectively. Finally, the accuracy of the neural distinguisher is significantly higher than that of traditional differential distinguishers, according to the experiments. In conclusion, these results represent a significant contribution to the field of cryptography by demonstrating the effectiveness of advanced machine learning techniques in enhancing the performance of cryptanalysis.