Scholarlink

As blockchain technology rapidly advances, it faces significant security threats from attacks like selfish mining, which exploit consensus algorithm vulnerabilities, undermining system security. Traditional analysis methods, primarily reliant on Markov models, are often utilized to address these complex threats. However, these methods frequently fall short in accurately simulating the multifaceted nature of blockchain attacks, leading to gaps in security measures. To bridge this gap, we introduce AutoMiner, an innovative reinforcement learning-based framework that integrates Miner Monte Carlo Tree Search (MMCTS) with Long Short-Term Memory (LSTM) networks for simulating and detecting potential mining attacks within the Proof of Work (PoW) framework. Our experimental results demonstrate AutoMiner’s capability to outperform traditional selfish and honest mining strategies, achieving up to \(20\%\) higher profits under specific scenarios. This highlights AutoMiner’s potential in enhancing blockchain security by providing a more comprehensive and effective approach to analyzing and mitigating mining attacks.

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

AutoMiner: Reinforcement Learning-Based Mining Attack Simulator

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Wei Li,
  • Lide Xue,
  • Ziyang Han,
  • Bingren Chen,
  • Xishan Zhang,
  • Xuehai Zhou

摘要

As blockchain technology rapidly advances, it faces significant security threats from attacks like selfish mining, which exploit consensus algorithm vulnerabilities, undermining system security. Traditional analysis methods, primarily reliant on Markov models, are often utilized to address these complex threats. However, these methods frequently fall short in accurately simulating the multifaceted nature of blockchain attacks, leading to gaps in security measures. To bridge this gap, we introduce AutoMiner, an innovative reinforcement learning-based framework that integrates Miner Monte Carlo Tree Search (MMCTS) with Long Short-Term Memory (LSTM) networks for simulating and detecting potential mining attacks within the Proof of Work (PoW) framework. Our experimental results demonstrate AutoMiner’s capability to outperform traditional selfish and honest mining strategies, achieving up to \(20\%\) higher profits under specific scenarios. This highlights AutoMiner’s potential in enhancing blockchain security by providing a more comprehensive and effective approach to analyzing and mitigating mining attacks.