Machine Learning Algorimths Identify Cryptocurrency Pump-and-Dump Schemes Before they Happen
According to MIT Technology Review’s on December 4, 2018, security researchers have discovered a machine learning algorithm that can spot cryptocurrency pump-and-dump schemes in advance, which can help prevent and stop any pump-and-dump projects from developing further.
The Anatomy Of a Pump-And-Dump Scheme
While pump-and-dump schemes are common in conventional commodity trading markets, they have become trendy in the cryptocurrency market. A pump-and-dump scheme often begins when an organizer selects a fairly unknown obscure token and starts accumulating these tokens.
The organizer announces that a pump operation will begin and will select a random cryptocurrency at a particular time on select anonymous channels such as Telegram. The organizer then reveals at a specific time what the cryptocurrency token chosen is. Other interested parties will begin purchasing the token which leads to a sudden sharp increase in the price of the cryptocurrency token.
When the price reaches its highest point, the token organizers begin a sell-off. The participants make a quick profit by selling their tokens and profiting from anybody who is unlucky enough to have joined in the process later. The pump-and-dump schemes take place in just a few minutes. Here the organizer makes the most significant amount of profit but many others who have also participated in early also make a nice profit as well.
Developing the Machine Learning Algorithm
Jiahua Xu and Benjamin Livshits from the Imperial College of London are the security researchers responsible for developing the machine learning algorithm. Xu and Livshits have been studying how cryptocurrency pump-and-dump scheme operates and have recently published the first in-depth account of these scams. They’ve managed to construct an algorithm that can detect these pump-and-dump scams early.
Xu and Livshits used historical data from over 236 pump-and-dump events that occurred between July 21 and November 18, 2018, to train a machine-learning algorithm to spot the signs of a pump-and-dump scam. To understand the scheme in detail, the researchers went into an in-depth study of a single pump-and-dump scam and recorded the change in price and trading volume of the specific cryptocurrency.
While many cues reveal a pump-and-dump scheme, the study suggested that the greatest tip-off generally occurs when there is an unexpected volume of trades with an obscure token.
Although the machine-learning algorithm can help undermine or prevent future cryptocurrency pump-and-dump schemes, MIT Technology Review believes that, in the event where a machine learning algorithm can discover these scams, the organizers will quickly change their activities so it will be harder for the algorithm to detect.
Unfortunately, cryptocurrency scams will ultimately stick around for a while. An in-depth, detailed understanding of how they operate is, however, essential to prevent pump-and-dump schemes from developing further and spreading more widely in the cryptocurrency community.