The Financial Supervisory Service said on the 20th that it expanded its market surveillance system using artificial intelligence (AI) to respond to unfair trading in virtual assets.
The Financial Supervisory Service (FSS) in January developed an AI algorithm that automatically identifies the order book and price involvement intervals of suspected price manipulators, which had relied on investigators' manual work. Then in April, for more effective analysis of allegations, it developed a "suspicious group automatic identification function" that automatically detects multiple linked accounts used for price manipulation. It also built a surveillance system that receives in real time not only price information from exchanges but also various data such as order book, executions, and market alerts.
This time, it expanded the scope of AI use to market surveillance tasks that detect abnormal transactions occurring in the virtual asset market and analyze them to determine whether to launch an actual investigation. In particular, considering the characteristics of the virtual asset market, where thousands of tokens are traded 24 hours across multiple exchanges, it used Generative AI of public large language models (LLMs) and Machine Learning algorithms to capture abnormal transactions in real time and analyze them quickly.
Generative AI was also introduced in the process of analyzing suspected tokens, which had taken considerable time. Accordingly, it automated the market surveillance process from automatic identification of abnormal transactions to pinpointing suspect intervals and generating review reports. First, it established a function that uses real-time data from exchanges to identify at an early stage suspected tokens where unfair transactions are being attempted. The Financial Supervisory Service (FSS), based on its investigative experience, defines the main types of price manipulation in the virtual asset market and captures in real time the suspected tokens that correspond to each type.
The Generative AI model also analyzes the causes of sharp spikes and plunges in prices and trading volume by comparing notices and news related to the identified tokens. The Financial Supervisory Service (FSS), when typical patterns of price manipulation appear or when prices surge without clear grounds and in-depth analysis is needed, receives trading data from exchanges to assess the necessity of an actual investigation.
It also identifies wash trades and matched orders in real time. The Financial Supervisory Service (FSS) applied "Benford's law," used in fields such as accounting, and Machine Learning algorithms to virtual asset market surveillance to crack down on fabricated transactions such as artificially inflated trading volume. It also developed a function that uses publicly available online information to identify suspected tokens. This is to detect unfair trading activities conducted online, such as illegal front‑running through 'reading rooms,' the distribution of videos containing false information, and posts inciting unfair transactions.
Among the abnormal transactions identified during monitoring, for tokens requiring in-depth analysis, AI automates the series of steps from trade analysis to drafting the results report.
Going forward, it plans to additionally develop functions that support fund flows and on‑chain tracing. The Financial Supervisory Service (FSS) said it will continue to strengthen its AI‑based market surveillance and investigation system to protect users and establish a sound market order for virtual assets.