Scholarlink

<p>Cellular Automaton (CA) is an intriguing primitive for constructing efficient substitution boxes (S-boxes) for cryptographic applications. This paper presents a novel approach by combining CA and Reinforcement Learning (RL). Semi-bent Boolean functions derived from CA rules are used for the construction of an efficient S-box with desirable cryptographic properties like high nonlinearity and low differential uniformity. The process of selecting optimal CA rules is modelled as a Markov Decision Process (MDP), where a reinforcement learning agent navigates the state space of rule combinations to maximize a reward signal based on cryptographic properties. Various configurations of one-dimensional CA with 3-neighborhood and 4-neighborhood rules are explored to generate S-boxes through the application of cryptographically suitable CA rules discovered by the reinforcement agent. The proposed approach offers several benefits such as reduced memory footprint, exploration of an extensive solution space and inherent parallelism suitable for hardware implementations. Experimental results indicate that the S-boxes generated with 4-neighborhood CA exhibit superior cryptographic strength compared to previously proposed CA based S-boxes.</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

Design of cryptographically suitable lightweight cellular automata based substitution-boxes using reinforcement learning

  • 期刊论文

  • 发表日期：

    2025年9月20日

查看全文
  • Ebey S Raj,
  • A. Aravind,
  • Anita John,
  • Jimmy Jose

摘要

Cellular Automaton (CA) is an intriguing primitive for constructing efficient substitution boxes (S-boxes) for cryptographic applications. This paper presents a novel approach by combining CA and Reinforcement Learning (RL). Semi-bent Boolean functions derived from CA rules are used for the construction of an efficient S-box with desirable cryptographic properties like high nonlinearity and low differential uniformity. The process of selecting optimal CA rules is modelled as a Markov Decision Process (MDP), where a reinforcement learning agent navigates the state space of rule combinations to maximize a reward signal based on cryptographic properties. Various configurations of one-dimensional CA with 3-neighborhood and 4-neighborhood rules are explored to generate S-boxes through the application of cryptographically suitable CA rules discovered by the reinforcement agent. The proposed approach offers several benefits such as reduced memory footprint, exploration of an extensive solution space and inherent parallelism suitable for hardware implementations. Experimental results indicate that the S-boxes generated with 4-neighborhood CA exhibit superior cryptographic strength compared to previously proposed CA based S-boxes.