As the overhaul of the virtual asset taxation system gains momentum, the construction of a next-generation abnormal transaction detection system combining AI and a graph database (GraphDB) is being pursued. It is noteworthy for applying a technology infrastructure capable of tracking even dozens of roundabout transactions in real time to the public tax sector.

Sky Worldwide (SKAI) announced on the 6th that it will participate as a key partner in the National Tax Service's 'Virtual Asset Integrated Analysis System' construction project and begin developing an abnormal transaction detection system.

This is due to the need to establish an advanced detection system to analyze vast amounts of transaction data as the implementation of taxation on income from the transfer and lending of virtual assets in January 2027 coincides with the mandatory submission of transaction records by Virtual Asset Service Providers (VASPs).

The core systems that Sky Worldwide will establish in this project include: △establishment of a virtual asset transaction flow analysis system △analysis of fund flow relationships based on GraphDB △design of a data model for abnormal transaction detection △multi-stage transaction tracking and pattern analysis △establishment of an abnormal pattern detection environment △linkage analysis of blockchain and relational data.

The company explained that existing FDS systems used by tax authorities and the financial sector operate on relational databases (RDB) and rule-based structures, which limits their ability to respond to sophisticated money laundering methods such as smurfing (dividing high-value assets into small amounts) or layering (transferring assets through dozens of steps).

It has been pointed out that in multi-stage transaction structures, analysis speeds can drastically decrease or real-time tracking becomes impossible.

To address this, Sky Worldwide applies its hybrid database management system (DBMS), AgensGraph, as a core technology.

AgensGraph is a GraphDB platform that stores and analyzes transaction entities and fund flows in a node-edge structure, and the company stated that by combining it with Graph Neural Network (GNN)-based AI technology, it can visualize multi-stage fund flows and precisely detect anomaly patterns.

New complex cash flows of 5 or more stepsThe explanation is that it is possible to track quickly and preemptively detect abnormal transactions.

AgensGraph has previously been applied to FDS projects in the public and financial sectors, including the Kookmin Bank AI Voice Phishing Monitoring System (2021), the Kookmin Bank Integrated Continuous Audit Monitoring System (2024), and the National Health Insurance Service Integrated Collection Information System Reconstruction (2024–2025).

Shin Jae-hyuk, CEO of Sky Worldwide, stated, “This project is a prime example of applying graph database-based abnormal transaction analysis technology to the public sector,” adding, “We plan to expand the scope of application to include Anti-Money Laundering (AML) and Fraud Detection Systems (FDS) in the financial sector, as well as digital asset monitoring.”