# Do not rug on me: Zero-dimensional Scam Detection

**Authors:** Bruno Mazorra, Victor Adan, Vanesa Daza  
**arXiv:** [2201.07220](https://arxiv.org/abs/2201.07220)  
**Date:** 2022-01-16  
**Summary page:** https://www.summarizepaper.com/en/paper/2201.07220v1/

## Key points

- Uniswap is a popular decentralized exchange (DEX) that allows users to trade digital assets without trusted third parties
- Previous studies have detected rug pulls in EVM compatible chains, but the first academic step towards detecting and characterizing scam tokens on Uniswap was made in a previous study
- The authors increased their data set by 20K tokens and proposed a new methodology to label tokens as scams before they occur
- They devised a theoretical classification of different malicious maneuvers in Uniswap protocol and proposed various machine-learning-based algorithms with new relevant features related to token propagation and smart contract heuristics that can detect potential rug pulls before they happen
- These models achieved similar results with the best model obtaining an accuracy of 0.9936, recall of 0.9540, and precision of 0.9838 in distinguishing non-malicious tokens from scams prior to any malicious maneuver
- The authors' contribution provides valuable insights into how scammers operate on Uniswap and proposes effective methods for detecting potential rug pulls before they occur
- This research could help prevent investors from falling victim to fraudulent schemes while promoting transparency and trust within DeFi ecosystems like Uniswap's DEX platform.

## Summary

Uniswap, a decentralized exchange (DEX), has gained significant popularity due to its non-custodial and publicly verifiable nature that allows users to trade digital assets without the need for trusted third parties. Various projects have contributed to detecting rug pulls in EVM compatible chains, but the first academic step towards detecting and characterizing scam tokens on Uniswap was made in a previous study. The authors increased their data set by 20K tokens and proposed a new methodology to label tokens as scams before they occur. After manually analyzing the data, they devised a theoretical classification of different malicious maneuvers in Uniswap protocol and proposed various machine-learning-based algorithms with new relevant features related to token propagation and smart contract heuristics that can detect potential rug pulls before they happen. In general, these models achieved similar results with the best model obtaining an accuracy of 0.9936, recall of 0.9540, and precision of 0.9838 in distinguishing non-malicious tokens from scams prior to any malicious maneuver. The authors' contribution provides valuable insights into how scammers operate on Uniswap and proposes effective methods for detecting potential rug pulls before they occur. This research could help prevent investors from falling victim to fraudulent schemes while promoting transparency and trust within DeFi ecosystems like Uniswap's DEX platform.