Scholarlink

Concerns over proof of work and proof of stake techniques that are fundamental to the blockchain yet also susceptible to certain security flaws have grown as a result of increased adoption of the blockchain. Notably, malevolent nodes might work together to create a fake consensus, which can cause the blockchain to be incorrectly validated. This vulnerability affects both PoW and PoS algorithms. Creating strong defenses against malevolent nodes for these systems is crucial to resolving this pressing problem. This work investigates the possibilities of using metaheuristic approaches to optimize artificial intelligence systems for the detection of rogue nodes. An altered form of a metaheuristic optimization technique was presented and used on a semi-synthetic dataset that was made available to the general public. With the help of the suggested method, the top-performing classification algorithms were optimized, and their accuracy surpassed 87% suggest suitable potential.

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

Optimizing XGBoost for Blockchain Malicious Node Detection by Metaheuristics

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Luka Jovanovic,
  • Mihajlo Protic,
  • Nebojsa Bacanin,
  • Milos Mravik,
  • Miodrag Zivkovic,
  • Milos Antonijevic,
  • Jelena Cadjenovic Milovanovic

摘要

Concerns over proof of work and proof of stake techniques that are fundamental to the blockchain yet also susceptible to certain security flaws have grown as a result of increased adoption of the blockchain. Notably, malevolent nodes might work together to create a fake consensus, which can cause the blockchain to be incorrectly validated. This vulnerability affects both PoW and PoS algorithms. Creating strong defenses against malevolent nodes for these systems is crucial to resolving this pressing problem. This work investigates the possibilities of using metaheuristic approaches to optimize artificial intelligence systems for the detection of rogue nodes. An altered form of a metaheuristic optimization technique was presented and used on a semi-synthetic dataset that was made available to the general public. With the help of the suggested method, the top-performing classification algorithms were optimized, and their accuracy surpassed 87% suggest suitable potential.