Journal of Computer Applications ›› 2024, Vol. 44 ›› Issue (9): 2770-2776.DOI: 10.11772/j.issn.1001-9081.2023091254

• Cyber security • Previous Articles     Next Articles

Random validation blockchain construction for federated learning

Tingwei CHEN, Jiacheng ZHANG, Junlu WANG()   

  1. Faculty of Information,Liaoning University,Shenyang Liaoning 110031,China
  • Received:2023-09-13 Revised:2023-12-20 Accepted:2023-12-22 Online:2024-03-15 Published:2024-09-10
  • Contact: Junlu WANG
  • About author:CHEN Tingwei, born in 1974, Ph.D., professor. His research interests include intelligent transportation, machine learning.
    ZHANG Jiacheng, born in 1998, M. S. candidate. His research interests include blockchain, machine learning.
  • Supported by:
    National Key Research and Development Program of China(2021YFF0901004);Applied Basic Research Program of Liaoning Province(2022JH2/101300250);Digital Liaoning Intelligent Manufacturing to Strengthen Province Fund (Direction of Digital Economy)(13031307053000568)

面向联邦学习的随机验证区块链构建

陈廷伟, 张嘉诚, 王俊陆()   

  1. 辽宁大学 信息学部,沈阳 110031
  • 通讯作者: 王俊陆
  • 作者简介:陈廷伟(1974—),男,内蒙古赤峰人,教授,博士,CCF会员,主要研究方向:智能交通、机器学习
    张嘉诚(1998—),男,湖北黄冈人,硕士研究生,主要研究方向:区块链、机器学习
    王俊陆(1988—),男,辽宁丹东人,讲师,博士,CCF会员,主要研究方向:大规模图处理、大数据、流数据。
  • 基金资助:
    国家重点研发计划项目(2021YFF0901004);辽宁省应用基础研究计划项目(2022JH2/101300250);数字辽宁智造强省专项资金(数字经济方向)资助项目(13031307053000568)

Abstract:

A random verification blockchain construction and privacy protection method for federated learning was proposed to address the issues such as local device model gradient leakage, the ability of centralized server devices to exit at will, and the inability of global models to resist malicious user attacks in existing federated learning models. Firstly, blockchain leadership nodes were elected randomly by introducing verifiable hash functions, thereby ensuring the fairness of voting a node to create block. Secondly, a verification node cross detection mechanism was designed to defend against malicious node attacks. Finally, based on differential privacy technology, blockchain nodes were trained, and incentive rules were constructed on the basis of the contribution of nodes to the model to enhance the training accuracy of the federated learning model. Experimental results show that the proposed method achieves 80% accuracy for malicious node poisoning attacks with 20% malicious nodes, which is 61 percentage points higher than that of Google FL, and the gradient matching loss of the proposed method is 14 percentage points higher than that of Google FL when the noise variance is 10-3. It can be seen that compared to the federated learning methods such as Google FL, the proposed method can ensure good accuracy while improving the security of the model, and has better security and robustness.

Key words: federated learning, blockchain, differential privacy, incentive mechanism, anomaly detection

摘要:

针对现有联邦学习模型中存在的本地设备模型梯度泄露、中心化服务器设备可随意退出、全局模型无法抵御恶意用户攻击等问题,提出面向联邦学习的随机验证区块链构建及隐私保护方法。首先,引入可验证哈希函数以随机选举区块链的领导节点,确保节点出块的公平性;其次,设计了验证节点的交叉检测机制防御恶意节点的攻击;最后,基于差分隐私技术训练区块链节点,根据节点对模型的贡献程度构建激励规则进行节点激励,提高联邦学习模型的训练准确率。实验结果表明,所提方法在20%恶意节点的情况下,对于恶意节点的投毒攻击能够达到80%的准确率,相较于Google FL提升了61个百分点,而所提方法在噪声方差为10-3时梯度匹配损失比Google FL提升了14个百分点。可见,相较于Google FL等联邦学习方法,所提方法在提升模型的安全性前提下能够保证良好的精确度,具有更好的安全性和鲁棒性。

关键词: 联邦学习, 区块链, 差分隐私, 激励机制, 异常检测

CLC Number: