A group of researchers from the UK has decided to make use of the Machine Learning and Artificial Intelligence to recognize one of the biggest nightmares of the crypto world, Pump & Dump Scams. The choice of tool and reliability on machine learning has come into the picture because, currently more than 2000 crypto tokens are trading on the market, so it’s not easy to keep a track of mischievous criminals.
Pump & Dump is a coined term for insider trading in the market. This form of scam has been in the market for quite some time. Cryptocurrencies in the absence of liquidity factor and less-knowledge in the community, make it a perfect target for insider trading kind of scams.
Jiahua Xu and Benjamin Livshits from the Imperial College London wrote a paper called, “The Anatomy of a Cryptocurrency Pump-and-Dump Scheme.” In this paper, they have used machine learning techniques to detect Pump & Dump scams before they happen.
The study paper released by Jiahua Xu has made some fascinating revelations, on an average two Pump & dump happens on a daily basis which generates $7 million worth of trade volume in a month. The other key observation was the great buying spree that goes in the market right before a significant dump. So, the team behind the implementations of Machine learning programmed it to detect any such unusual buying spree.
After putting in all these observations onto the machine and training it to detect the scam, the team tasted quite a success. The project predicted six such upcoming dumps, out of these 6 cases, 5 showed more than 100% rise in the buying value and the sixth one did not show much growth. So, 5/6 is a high success rate, making the experiment a success.
How Do Pump & Dump Scams work?
The insider trading or Pump & Dump scams have been troubling the authorities for quite some time now. Usually, the organizers behind the pump select a currency for the dump, the lesser known and cheaper the coin, the easier it is to manipulate it. The choice of the coin is cheaper ones because moving a heavy coin would require a ton of capital and a lot of attention from the regulators.
Once the choice of the coin is made, the organizers start buying in vast quantities to have the upper hand on the market. A date and time are decided, and all other followers are informed of the same. On the due date, the organizer sets a Pump target and start pumping the bought coins on to the market. Others after looking at the high volume trade, join in the bandwagon, only to realize it later that the whole process was pre-orchestrated. The Pump & Dump hardly takes more than a couple of minutes.
The market returns to a usual point, and the hoarded cryptos turn out to be the dump since others were not aware of the elaborate plan. Thus, the name insider trading
The use of machine learning to detect such fragile aspect and manipulation of the trade market shows us how two relatively new technologies can bind together to weed out some grass root level problems. Both Machine Learning & Blockchain has come a long way from its initial days and had to overcome a few more issues to become mainstream. But, one thing is quite evident that the technology needs more refining and real-world use case to stake its claims on being a mainstream form of technology.