facebook pixel image Identifying Money Laundering with Link Analysis and Machine Learning

Identifying Money Laundering with Link Analysis and Machine Learning

One of the most common elements of money laundering is called wash trading, where an individual or group of investors places matching buy and sell orders to transfer assets between accounts, without changing the ultimate owner of the funds.

Where and Why Wash Trades Happen

Wash trades are an integral part of most money laundering schemes because they effectively disguise the original illegal source of funds by shuffling them between accounts, after the assets have been placed in the financial system. Wash trades are also used to commit investment fraud, because they artificially create volume in financial markets to drive up demand, and thus prices. Once the value of assets rises, insiders sell to unsuspecting investors and profit handsomely, after which liquidity dries up and prices collapse, leaving other traders without their shirts.

For this reason, wash trading on major US exchanges was made illegal in 1936, with the passage of the Commodity Exchange Act, the same law that brought all commodity trading into regulated exchanges. However, complex wash trading schemes still undoubtedly occur. Cryptocurrency platforms, in particular, are rife with wash trades, adding to fears that the new digital assets are a speculative bubble. According to estimates by the Blockchain Transparency Institute, as of April 2019, “Over 60% of all exchanges ranked on popular data sites have little to no volume and were found to be over 96% fake each.”


Anti-Money Laundering and Fraud Professionals Rely on Visallo. Contact Us to chat with the Visallo team or Schedule a Demo.


Examples and Defining Features

The simplest possible stylized example of a wash trade would be a buy order from account A for $X, and a subsequent sell order from account B for $X just moments later, or vice versa, where both accounts actually have the same owner.

These exact matched orders would be easily spotted by fraud investigators, however, so wash trades typically have a highly complex structure to disguise their nature. In particular, they almost always involve more than two orders, where those on opposite sides of trades do not match exactly, but are within some small margin of each other.

For example, the table below, from the IEEE article, Detecting Wash Trade in Financial Market Using Digraphs and Dynamic Programming, shows how four matched orders between A and B effectively create two transactions of substantial volume (490 shares), but with no significant net transfer of shares between the two parties.

Order # Trader Time Buy / Sell Price Volume
001 A 9:00:000 AM Buy 125 500
002 B 9:00:001 AM Sell 124.2 490
003 B 9:10:000 AM Buy 125.2 490
004 A 9:10:001 AM Sell 125 500


To identify likely wash trades, analysts typically look for groups of transactions that satisfy three criteria:

  • A short amount of time between buy and sell orders, so that the parties involved in the wash trade can be sure to transact with each other
  • An ask or bid price that will clear the market, so that the trades can be executed
  • Volumes that are nearly identical, so that orders involved in the trade are more likely to be matched with one another

Detecting Wash Trades with Algorithms

One way to identify potential wash trades is to apply algorithms from operations research to market data. For example, queuing algorithms examine moving windows of buy and sell orders in parallel to search for opposing orders that are close in price and volume. Other algorithms essentially recast the problem of detecting wash trades as a special case of an interesting mathematical problem called the knapsack problem, and use recursive dynamic programming to identify potential wash trades.

Identifying Wash Trades with Investigative Software

As complements to algorithmic detection methods, other ways to spot likely wash trades are more visual in nature, and lend themselves to link analysis tools like Visallo.

  • Look for cyclical graphs connecting groups of accounts with similarly sized transactions, since a closed loop between different accounts is a defining feature of all wash trades.
  • Work outwards from suspicious accounts by visualizing their transactions with other counterparties, enabling analysts to start their investigation from the most suspicious trades out of thousands in the broader market.
  • Examine temporal transaction patterns with tools such as interactive timelines, helping analysts more easily spot suspicious activities by eye

Visallo insider threat screenshot

Above is an example of how link analysis with a tool like Visallo can help you see connections in complex data.

Summary

Wash trades are illegal transactions commonly associated with both money laundering and investor fraud, but their complex structure and the number of counterparties involved make them difficult to detect.

To identify these trades in broader data sets of transactions, analysts can look for their defining features using either queuing algorithms or dynamic programming. In addition, software like Visallo enables investigators to visualize the network structures that make these types of trades distinctive.


Visallo is one of the best Link Analysis tools on the market for AML and Fraud Detection. Contact Us to chat with the Visallo team or schedule a demo.