Abstract
Uniswap, like other DEXs, has gained much attention this year because it is anon-custodial and publicly verifiable exchange that allows users to tradedigital assets without trusted third parties. However, its simplicity and lackof regulation also makes it easy to execute initial coin offering scams bylisting non-valuable tokens. This method of performing scams is known as rugpull, a phenomenon that already existed in traditional finance but has becomemore relevant in DeFi. Various projects such as [34,37] have contributed todetecting rug pulls in EVM compatible chains. However, the first longitudinaland academic step to detecting and characterizing scam tokens on Uniswap wasmade in [44]. The authors collected all the transactions related to the UniswapV2 exchange and proposed a machine learning algorithm to label tokens as scams.However, the algorithm is only valuable for detecting scams accurately afterthey have been executed. This paper increases their data set by 20K tokens andproposes a new methodology to label tokens as scams. After manually analyzingthe data, we devised a theoretical classification of different maliciousmaneuvers in Uniswap protocol. We propose various machine-learning-basedalgorithms with new relevant features related to the token propagation andsmart contract heuristics to detect potential rug pulls before they occur. Ingeneral, the models proposed achieved similar results. The best model obtainedan accuracy of 0.9936, recall of 0.9540, and precision of 0.9838 indistinguishing non-malicious tokens from scams prior to the malicious maneuver.