Detection of Cryptocurrency Market Manipulations: Pump & Dump Schemes

Abstract

Cryptocurrency exchange markets are subject to a variety of manipulation schemes due to the lack of supervision imposed, encouraging manipulators to exploit these markets to generate extra returns. One of the popular scams is the pump and dump scheme, where perpetrators gather to artificially inflate the price of a cryptocurrency. Detection of these scams is addressed in several studies which mostly apply anomaly detection techniques. This paper investigates the properties and studies addressing pump and dump schemes in cryptocurrencies and traditional market, then conducts a supervised learning approach to detect cryptocurrency pump and dump cases. This is motivated by the availability of reliable labelled dataset and the lack of studies utilizing supervised learning techniques in cryptocurrency market. Support vector machines (SVMs), random forest (RF), and artificial neural networks (ANNs) are employed in this study. They are trained using level 1 data, which refers to the opening price, high price, low price, close price, and volume (OHLCV) gathered from CCXT APIs for Binance Cryptocurrency Exchange. This type of data is easy to get unlike order book data which is not easily found. The models varied in their performances with ANN achieving the best results in all metrics calculated. It has proven its ability to classify manipulations with 86% in the F1 score.