5대 코인거래소 시세조종 AI로 잡는다…심층분석 ‘3일→1시간’ 단축

South Korean virtual asset exchanges have implemented an automated market surveillance process utilizing artificial intelligence and machine learning algorithms to combat unfair trading practices. This system, dubbed 'Virtual Asset Illicit Trading Market Surveillance Process Automation,' can now analyze trading data in real-time and conduct in-depth analysis within an hour, down from the previous three days.

The Financial Supervisory Service of South Korea (FSS) has incorporated generative AI and machine learning algorithms into this automated system, which now scrutinizes the trading activities of five major virtual asset exchanges – Upbit, Bithumb, Coinone, Homeplus, and Gate.io – across approximately 2,000 virtual assets.

The primary focus of this advanced scrutiny is to detect potential instances of market manipulation, such as pump and dump schemes and order spoofing, which previously took a significant amount of time to investigate due to the extensive volume of data involved. Using AI algorithms, the system can now detect abnormal trading patterns, such as artificial inflation of trading volume or distortion of prices, as well as identify malicious content generated from online news, social media posts, and videos. These AI-driven algorithms can process up to tens of gigabytes of data during the analysis period.

Once the system flags a suspicious activity, it sends the relevant trading records to the virtual asset exchange for further examination. The examination involves a comprehensive analysis of account-specific order and settlement records. The FSS has confirmed that the system not only helps in identifying clear cases of illicit trading but also enables screening of borderline cases that may have been previously overlooked. In some instances, the system has even led to actual investigations based on the detected patterns.