Problem
Cryptocurrency pump-and-dump schemes are a form of market manipulation where organizers artificially inflate the price of a low-liquidity cryptocurrency, then sell their holdings at a profit while leaving late entrants with significant losses. We report on the first large-scale measurement study of cryptocurrency pump-and-dump schemes, collecting data from over 400 pump events across Telegram and Discord channels between June 2018 and February 2019.
Approach
We show that these schemes are rampant, with at least one pump occurring every day and an average volume increase of 13% per event. By analyzing the temporal and economic patterns of these pumps, we demonstrate that they exhibit a characteristic lifecycle that can be detected algorithmically.
Results
We propose a filter to flag coins that are likely pump targets based on social media signals, and propose a classifier that can flag pumps in progress within 3 seconds of the initial announcement. This approach can serve as a foundation for building warning systems that protect investors.