Pinpointing 'Fake Accounts and Fraud Rings' in Big Data... UNIST Develops Ultrafast Network Analysis Technology
Algorithm Developed to Identify Specific Groups Around User-Defined Targets Without Full Network Data
Applications in Customer Group Analysis, Abnormal Transaction Detection, and More; Selected for Database Conference 'SIGMOD'
A new network analysis technology has emerged that can precisely detect small groups hidden within data, such as voice phishing organizations, suspicious transaction groups, and VIP consumer segments.
An algorithm has been developed that efficiently identifies only truly meaningful groups centered around people or entities of user interest, even in complex, intertwined networks. This technology is expected to make significant contributions to blocking fake accounts exhibiting abnormal behavior or network attacks in real time. It is also anticipated to aid in marketing focused on small, loyal customer groups, tracking related account groups linked to suspected fraudulent accounts, and discovering leads for new drug development through biological protein network analysis.
The research team led by Professor Kim Junghun from the Department of Computer Science and Engineering at UNIST announced on May 28 that they have developed a new community search technique that always includes a user-specified target and identifies only meaningful groups within a predefined size.
Research team, Professor Kim Junghoon (left) and Researcher Kim Dahi. Courtesy of UNIST
View original imageCommunity search is a data analysis technique that finds groups with strong internal connections within vast network data. Existing technologies have been difficult to use when network data is large or when only partial relationships can be seen due to privacy concerns, and there have been cases where groups included members with weak connections.
The research team’s technology can select groups that are tightly connected internally and well distinguished from the outside, all within the user-defined size, even without having access to the entire network information.
This is achieved by starting from a user-specified node and sequentially expanding the group by checking neighboring candidates. At each step, the algorithm assesses whether adding a candidate improves the overall result, while incorporating size constraints so that the score does not increase easily as the group becomes unnecessarily large.
To compensate for relationships that might be missed when selecting candidates individually, the algorithm also considers small clusters of connected nodes in the vicinity. This helps capture candidates that may not be noticeable alone, but clarify the group’s characteristics when grouped together.
In experiments using real network data, the new method achieved F1 scores up to 1.39 times higher and ARI scores up to 5.95 times higher than previous state-of-the-art techniques. This indicates that the technology more accurately identifies the target group while including fewer weakly connected members.
Professor Kim Junghun stated, “In real-world network analysis, it is difficult to obtain all the data at once, and the size of the groups actually needed is usually predetermined. This research focuses on quickly identifying only meaningful relationships near user-specified targets, and can be applied to various fields such as customer group analysis, abnormal transaction detection, and biological protein network analysis.”
Research figure, comparison of community detection performance validated using real network data.
View original imageThis research was conducted by Researcher Kim Dahee at UNIST as the first author and supported by the National Research Foundation of Korea.
Hot Picks Today
"Two Spend 500,000 Won in a Flash, Yet Wait Two Hours?" ... 'Dreamland' Shunned by Kids and Foreign Tourists
- The Era of Getting 200 Million Won After 30 Years Is Over... A Japanese Bank's "Perpetual Department Head" Who Just Clocked In Cries Out
- Spain Hit by Its Worst Tragedy... Lee Kangin, Moreno and Real Madrid Join Soccer World in Mourning
- They Said, “Make It Your Home for Life” - Then Suddenly Kicked 200 Passengers Off... What Happened to the Around-the-World Cruise?
- "130 Rice Bowls Sold Out in a Day"... 'Unfiltered Face' Adds 8 Million Won to Night Market Sales: China's Bungeunwoo
The results have been accepted for presentation at SIGMOD 2026, which will be held in Bengaluru, India, from May 31 to June 5. SIGMOD (ACM Special Interest Group on Management of Data) is one of the most prestigious conferences in the field of databases.
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.