MIT AI Lab: After analyzing 200,000 bitcoin transactions, only 2% were "illegal"

According to Coindesk's August 2 report, blockchain analysis firm Elliptic partnered with the Massachusetts Institute of Technology (MIT) to publish a public dataset on bitcoin transactions related to illegal activities.

(Source: Pixabay )

The team's research details how MIT-Watson Watson's Artificial Intelligence Lab researchers used machine learning software to classify 203,769 bitcoin node transactions worth about $6 billion. The study explores whether artificial intelligence can help the current anti-money laundering process.

• The adjustment posture of BTC has intensified, and the mainstream currency differentiation still needs to be improved.
• Heavy! Central Bank's deployment in the second half of the year: Accelerating the development of legal digital currency (full text)
• Bitfinex said NYAG's investigation was too demanding, and the latter insisted that it submit relevant documents.

After examining the association of these nodes with known entities, the researchers found that only 2% of the 200,000 bitcoin transactions were considered illegal, and another 21% were confirmed to be legal, but the vast majority The transaction (about 77%) is still not classified. It is reported that so far, since its launch in 2009, an estimated 440 million bitcoin transactions have been made.

To be clear, this 2% illegal trade comes from the previously unpublished Elliptic dataset, and is only found by MIT researchers through analysis. However, the data is similar to a study by Chainalysis, another analyst firm. Chainalysis estimates that only 1% of bitcoin transactions in 2019 were related to illegal activities.

Since law enforcement agencies around the world often employ Elliptic to identify illegal activities involving cryptocurrencies, this study aims to find ways to help distinguish between illegal use and legal use of Bitcoin, especially for individuals or other unknown entities that do not have a bank account.

Elliptic co-founder Tom Robinson said:

In general, the key issue of compliance is false positives (negative is positive). An important part of this research is to reduce the number of false positive false positives. A key finding is that machine learning technology can be very effective in detecting illegal transactions.

Robinson added that sometimes, based on pre-stored data, ransomware attacks, and other criminal investigations in the dark market, software can find patterns that are difficult to describe but match known entities.

After completing the academic study, Elliptic published a data set to encourage open source.

Mark Weber, a researcher at the Massachusetts Institute of Technology, said:

We are sharing our early experiments on anti-money laundering with domain experts to solicit feedback. At the same time, we also hope that the release of the Elliptic dataset will inspire others to join in the development of new AML technologies and models, which in turn will enhance the security of the financial system.

In April, CNBC reported that an increase in global criminal activity could drive a surge in demand for $100 bills. A 2017 report by the American Economic Research Institute estimates that more than one-third of all circulating US currency is used by criminals and tax fraudsters.

We will continue to update Blocking; if you have any questions or suggestions, please contact us!