数据与计算发展前沿 ›› 2025, Vol. 7 ›› Issue (6): 23-34.

CSTR: 32002.14.jfdc.CN10-1649/TP.2025.06.003

doi: 10.11871/jfdc.issn.2096-742X.2025.06.003

• 专刊：第40次全国计算机安全学术交流会征文 • 上一篇    下一篇

基于威胁环境感知与大模型特征增强的区块链异常交易检测模型

梁飞1(),张世星2,*(),陈子睿3   

  1. 1.北京市公安局经济犯罪侦查总队，北京 100061
    2.中国人民警察大学智慧警务学院，河北 廊坊 065000
    3.中国人民公安大学信息网络安全学院，北京 100038
  • 收稿日期:2025-07-24 出版日期:2025-12-20 发布日期:2025-12-17
  • 通讯作者: 张世星
  • 作者简介:梁飞，北京市公安局经济犯罪侦查总队，中队长，副高级工程师，主要研究方向为图神经网络、区块链数据分析。
    本文承担工作为：模型原理设计、模型算法代码实现。
    LIANG Fei is a squad leader and associate senior engineer at the Economic Crime Investigation Brigade of Beijing Municipal Public Security Bureau. His main research interests include graph neural networks and blockchain data analysis.
    In this paper, he is responsible for the design of the model principles and the implementation of the model algorithm code.
    E-mail: 475662476@qq.com|张世星，中国人民警察大学智慧警务学院，副教授，主要研究方向为网络安全应用技术。
    本文承担工作为：指导优化模型。
    ZHANG Shixing is an associate professor at School of Information and CyberSecurity, People's Public Security University of China. His main research area is network security application technology.
    In this paper, he is responsible for guiding and optimizing the model.
    E-mail: 382927221@qq.com

A Blockchain Anomaly Transaction Detection Model Based on Threat Environment Perception and Large Language Model Feature Enhancement

LIANG Fei1(),ZHANG Shixing2,*(),CHENG Zirui3   

  1. 1. Economic Crime Investigation Brigade of Beijing Municipal Public Security Bureau, Beijing 100061, China
    2. School of Intelligent Policing, China People’s Police University, Langfang, Hebei 065000, China
    3. School of Information and CyberSecurity, People’s Public Security University of China, Beijing 100038, China
  • Received:2025-07-24 Online:2025-12-20 Published:2025-12-17
  • Contact: ZHANG Shixing

RichHTML

4

PDF

0

摘要：

【目的】目前主流的基于图神经网络检测区块链上恶意行为的方法多以特征聚合为主，本文改进传统方法，挖掘区块链交易中的威胁环境并利用大模型进行特征增强处理。【方法】提出的社团挖掘算法建立威胁环境结合大模型特征增强的检测模型（Graph Community-Aware Augmentation with Large Language Models）先通过社团挖掘算法在交易子图中进行划分，并作为节点地址所处的环境，将节点特征和所处社团的特征形成文本描述，使用大模型进行特征增强形成节点地址的大模型特征，最后将节点地址的原始特征和大模型特征进行融合作为模型输出的节点地址特征。【结果】模型在少量标签的样本中能够更好地捕捉和学习到恶意节点地址的行为特征。【结论】实验结果表明模型在两个公开的数据测试集中的指标上均超越了传统算法。

关键词: 社团挖掘, 大模型, 区块链

Abstract:

[Objective] Current mainstream methods for detecting malicious behavior on blockchains rely on graph neural networks and primarily focus on feature aggregation. This paper improves traditional approaches by exploring the threat environment perception in blockchain transactions and leveraging large language models for feature enhancement. [Methods] The proposed Graph Community-Aware Augmentation with Large Language Models model first divides transactions into subgraphs using a community detection algorithm, treating the resulting communities as the environmental context of node addresses. The node features and their associated communities are then converted into textual descriptions, which are processed by a large language model to generate enhanced features. Finally, the original node features and the LLM-enhanced features are fused to form the final node representation. [Results] The model demonstrates an improved ability to capture and learn behavioral characteristics of malicious node addresses, even in scenarios with limited labeled samples. [Conclusions] Experimental results show that the proposed model outperforms traditional algorithms across multiple metrics on two public benchmark datasets.

Key words: Community Detection, Large Language Models, Blockchain