Kakao Bank has launched an artificial intelligence (AI)-based financial fraud detection model that analyzes behavioral patterns before and after transactions to predict the risk of fraud. The initiative aims to enhance the detection of financial scams, including voice phishing.
On June 9, Kakao Bank announced the development of the 'Sequence Detection Model,' which has been integrated into its fraud detection system (FDS).
The Sequence Model comprehensively analyzes not only individual transaction outcomes, such as transfers and withdrawals, but also the behavioral patterns surrounding these transactions to assess the risk of financial fraud.
Utilizing an 'attention mechanism' that allows AI to understand the relationships and flow between data, the model examines various behavioral cues, including the order of transactions, time intervals between actions, and whether devices have changed.
For instance, it can analyze patterns where transactions are initiated after a period of inactivity following app access. This pattern of pausing and resuming transactions is often seen in voice phishing schemes, where scammers persuade victims to make additional transfers. The Sequence Model detects these patterns to help prevent fraud.
The results have been promising. Since the pilot introduction of the Sequence Model in November of last year, the number of fraud prevention cases monitored through the FDS has increased by an average of 4.4 times compared to before its implementation. In the first quarter of this year, the Sequence Model accounted for 49.8% of all suspected fraudulent transactions that Kakao Bank prevented.
Kakao Bank plans to further enhance the Sequence Model to refine its FDS system. A company representative stated, "We have moved beyond detecting unusual transactions at specific points in time to predicting risks based on behavioral flows before and after transactions. We will continue our research and development to proactively respond to increasingly sophisticated financial crimes."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.