Machine Learning-Based Detection of Pump-and-Dump Schemes in Real-Time

2024-12-25

AI Summary

This work addresses the prevalent cryptocurrency “pump-and-dump” (P&D) manipulation in Telegram groups by proposing the first end-to-end real-time detection framework. Methodologically, it integrates heterogeneous multi-source data: natural language processing (NLP) for intent classification of group messages; time-series anomaly detection on on-chain price trajectories; and cross-platform correlation analysis across Telegram, exchanges, and on-chain data. Innovatively, it systematically evaluates the susceptibility to manipulation across diverse token standards—including ERC-20, ERC-721, BRC-20, Inscriptions, and Runes. Empirically validated on Poloniex, the framework achieves a 55.81% top-5 hit rate (24 out of 43) for target tokens among 50 randomly sampled cryptocurrencies and retrospectively identifies 2,079 historical P&D events. These results significantly mitigate information asymmetry and reduce investor risk in decentralized markets.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Intelligent Robots: Manipulation
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Economics, Online Markets and Human Computation: Economic aspects of blockchain and cryptocurrencies
• Web Mining and Content Analysis: Content-based information diffusion
• Security and Privacy: Cryptocurrency and smart contracts

Abstract

Cryptocurrency markets often face manipulation through prevalent pump-and-dump (P&D) schemes, where self-organized Telegram groups, some exceeding two million members, artificially inflate target cryptocurrency prices. These groups sell premium access to inside information, worsening information asymmetry and financial risks for subscribers and all investors. This paper presents a real-time prediction pipeline to forecast target coins and alert investors to possible P&D schemes. In a Poloniex case study, the model accurately identified the target coin among the top five from 50 random coins in 24 out of 43 (55.81%) P&D events. The pipeline uses advanced natural language processing (NLP) to classify Telegram messages, identifying 2,079 past pump events and detecting new ones in real-time. Our analysis also evaluates the susceptibility of token standards - ERC-20, ERC-721, BRC-20, Inscriptions, and Runes - to manipulation and identifies exchanges commonly involved in P&D schemes.

Problem

• Cryptocurrency
• Pump-and-Dump Detection
• Financial Risk Reduction

Innovation

• Real-time Prediction
• Natural Language Processing
• Pump-and-Dump Detection