Ethernet network phishing node detection method and system based on graph contrast learning

By employing a graph-based contrastive learning approach, combining graph convolutional networks and time series models, we extract and enhance the features of phishing nodes in the Ethereum network. This solves the data imbalance problem, achieves more efficient phishing node detection, and improves detection accuracy and recall.

CN117955718BActive Publication Date: 2026-06-12INNER MONGOLIA UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
INNER MONGOLIA UNIVERSITY
Filing Date
2024-01-29
Publication Date
2026-06-12

AI Technical Summary

⚠Technical Problem

Existing technologies for phishing node detection in the Ethereum network suffer from poor detection performance due to insufficient learning of phishing node features caused by data imbalance, and existing methods may also disrupt the original data distribution.

⚗Method used

A graph-based contrastive learning approach is adopted. By statistically analyzing each node in the target transaction graph, the statistical, structural, and transaction features of the nodes are extracted. Feature enhancement and classification are performed using graph convolutional networks and time series models. By combining feature discarding and edge perturbation techniques, more meaningful features are generated for classification.

🎯Benefits of technology

Without disrupting the original data distribution, it improves the performance and recall of phishing node detection, enhances detection accuracy and model robustness, and is able to better identify phishing nodes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117955718B_ABST
    Figure CN117955718B_ABST
Patent Text Reader

Abstract

The application discloses an Ethereum phishing node detection method and system based on graph contrast learning, relates to the technical field of phishing node detection, and comprises the following steps: first, counting nodes in a target transaction graph to obtain statistical features of each node; then, performing feature extraction on the nodes by using a structure feature extraction model to obtain structure features of each node; then, performing feature extraction on the nodes by using a transaction feature extraction model to obtain transaction features of each node; finally, splicing the statistical features, the structure features and the transaction features, and classifying by using a trained classifier to determine whether each node is a normal node or a phishing node; by comprehensively considering the statistical features, the structure features and the transaction features of the nodes, the problem of insufficient feature learning of a small amount of phishing nodes caused by data imbalance can be effectively solved, so that phishing node detection can be completed without destroying the original data distribution of the target transaction graph, and the detection performance of Ethereum phishing nodes is improved.
Need to check novelty before this filing date? Find Prior Art