PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
May 29, 2026AIP Advances1 citationsOpen Access

Eth-GBAV: Large-scale Ethereum phishing detection via graph attention variational inference and broad learning system

View Full Paper
DSDawei SongYZYuheng Zhang

Key Points

  • The research aims to develop a framework for effectively detecting phishing activities in Ethereum networks, addressing challenges in topological obscurity and label sparsity.
  • Utilized a self-supervised framework named Eth-GBAV integrating graph attention and adversarial variational inference.
  • Implemented a biased random walk strategy to capture initial behavioral semantics of transaction nodes.
  • Employed a broad learning system to enhance feature extraction from VAE-derived embeddings.
  • Achieved an F1-score of 0.9847 and recall of 0.9839 on the XBlock dataset, surpassing competitive models.
  • Maintained an accuracy of 0.9592 and F1-score of 0.9069 on the Kaggle dataset, demonstrating high robustness.

Abstract

To address the challenges of topological obscurity and extreme label sparsity in large-scale Ethereum transaction networks, a novel self-supervised phishing detection framework named Eth-GBAV is proposed, integrating graph attention, broad learning, and adversarial variational inference. The framework initiates with a biased random walk strategy guided by transaction intensity and temporal dynamics to capture the initial behavioral semantics of nodes. To distill discriminative features from noisy backgrounds, a “Generative-Attention” encoding architecture is constructed, where a graph attention network aggregates weighted structural neighborhoods and a Variational Autoencoder (VAE) characterizes the underlying probability distribution of legitimate transaction patterns. By maximizing the evidence lower bound, anomalous accounts are effectively isolated through reconstruction residuals. Furthermore, the broad learning system is introduced as an efficient analytical decision layer. By mapping VAE-derived latent embeddings and reconstruction errors into an expanded high-dimensional feature space, the framework captures intricate behavioral correlations via mapping and enhancement neurons. Extensive experimental verification on two large-scale datasets demonstrates the superior performance of Eth-GBAV. On the XBlock dataset, it achieves a leading F1-score of 0.9847 and a recall of 0.9839, outperforming the most competitive state-of-the-art model by significant margins. On the Kaggle dataset, the framework maintains high robustness with an accuracy of 0.9592 and an F1-score of 0.9069.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Song et al. (2026) studied this question.

synapsesocial.com/papers/6a192df7fab5b468c4417029https://doi.org/10.1063/5.0325166
Ask AI
Helpful
Bookmark
Share
View Full Paper