Identity Inference on Blockchain Using Graph Neural Network

The anonymity of blockchain has accelerated the growth illegal activities and criminal behaviors on cryptocurrency platforms. Although decentralization is one typical characteristics blockchain, we urgently call for effective regulation to detect these ensure safety stability user transactions. Identity inference, which aims make a preliminary inference about account identity, plays significant role in security. As common tool, graph mining technique can effectively represent interactive information between accounts be used identity inference. However, existing methods cannot balance scalability end-to-end architecture, resulting high computational consumption weak feature representation. In this paper, present novel approach analyze user's behavior from perspective transaction subgraph, naturally transforms task into classification pattern avoids computation large-scale graph. Furthermore, propose generic neural network model, named $\text{I}^2 \text{BGNN}$, accept subgraph as input learn function mapping achieving de-anonymization. Extensive experiments EOSG ETHG datasets demonstrate that proposed method achieve state-of-the-art performance