In findings that could prevent a repeat of the 2017 bitcoin price spike, researchers at Imperial College London claim to have developed a model for a crypto "pump-and-dump" scam, making it possible to proactively detect them.
Jiahua Xi and Benjamin Livshits write in their recently released paper, "The Anatomy of a Cryptocurrency Pump-and-Dump Scheme":
"The model exhibits high precision as well as robustness, and can be used to create a simple, yet very effective trading strategy, which we empirically demonstrate can generate a return as high as 80% within a span of only three weeks."
A pump-and-dump scheme is a type of market manipulation in which an often obscure cryptocurrency is hyped by those who buy low on an announced spree, only to sell it off quickly for a tidy profit.
This usually involves the spreading of rumors by the scam's planner from those "in the know." According to the US Commodities Futures Trading Commission (CFTC), such messages take the form of, for example: "5 minutes till pump, next message will be the coin! Tweet about us and send everyone the link to telegram [sic] for outsiders to see what we are pumping so they can get in on the action too!! lets [sic] take it to the MOON!!!!!"
Typically, because these scams are organized by planners that have amassed large sums of the cryptocurrency beforehand, pump-and-dumps usually primarily benefit the planner who "dumps" his/her coins first. This can all take just a few minutes.
An example of this is Russian crypto exchange YoBit, which tweeted that it would be buying a "random coin" every few minutes in October. The coin turned out not to be random and the stunt not only led to a dramatic increase of trading on the YoBit exchange, but also a significant profit for anyone that was in on the scam.
Building a Defense
Per the researchers, these scams have grown into a major nuisance. The researchers believe there are – on average – two pump-and-dump scams a day, accounting for $7 million in trading volume per month. There are more than 100 Telegram channels currently dedicated to the pumping-and-dumping of lesser coins.
By studying 237 pump-and-dump events between July 21 and November 18, the researchers have developed a model that could be used to train a machine-learning algorithm to quickly detect the telltale signs that a scam is about to occur.
The algorithm, after being taught the pattern, detected six instances of pump-and-dump from October 30 and November 6 – it was correct five times. It is unclear if or when this algorithm might be used to larger effect, such as monitoring entire exchanges.
This is not the first time that pump-and-dump has been researched. Researchers from the University of Florida and Princeton have concluded earlier that the schemes are damaging to the crypto market and "are detrimental to the liquidity and price of cryptocurrencies."
The CFTC recognizes pump-and-dump to be a major problem. The agency offers a bounty program for whistleblowers, paying from 10 to 30 percent of the monetary sanctions recuperated from perpetrators.