Vulnerability Detection for Token Trading System in Decentralized Finance

The thesis focuses on identifying and mitigating security vulnerabilities within the Decentralized Finance (DeFi) ecosystem, particularly in token trading systems. It explores various DeFi mechanisms, such as Flash Loans, the unlimited approval of ERC20 tokens, and deflationary token designs, uncovering critical risks and potential exploitation methods. Additionally, a key contribution is the development of DeFiGuard, a novel detection framework utilizing Graph Neural Networks to identify price manipulation attacks effectively. The research provides a foundation for enhancing the security of token trading systems in the DeFi ecosystem.