Token fraud identification and implications for post-crowdfunding performance

Abstract

Tokens issued by emerging Web3 applications serve multiple roles, including crowdfunding, payment, and governance, during the development of these applications. However, Web3 token fraud damages the trust of stakeholders, potentially contributing to the failure of these applications. We present an end-to-end mechanism for identifying wallet accounts suspected of Web3 token fraud and analyze the impact of such fraud on the performance of Web3 applications post-crowdfunding. First, we develop novel graph neural network models to identify fraudulent wallet accounts within evolving on-chain transaction networks using a crowd-reported fraud dataset. Next, we construct a dynamic ex ante fraud risk profile for each Web3 application by aggregating account-level fraud predictions. Finally, we evaluate the impact of risk profiles on Web3 application performance. Our results indicate a nuanced effect of Web3 token fraud. Web3 token fraud influences both application usage and user base expansion negatively. A prior surge in application usage may intensify the risk of token fraud, while earlier user base expansion can potentially alleviate this risk.