Ethereum phishing detection based on graph neural networks

With the development of blockchain, cryptocurrencies are also showing a boom. However, due to decentralized and anonymous nature have inevitably become hotbed for fraudulent crimes. For example, phishing scams frequent, which not only jeopardize financial security but hinder promotion blockchain technology. To solve this problem, paper proposes graph neural network-based detection method Ethereum, validates it using Ethereum datasets. Specifically, feature learning algorithm named TransWalk, consists random walk strategy transaction networks multi-scale extraction Ethereum. Then, an fraud framework is built based on conduct extensive experiments dataset verify effectiveness scheme in identifying detection.