[Digital Today reporter Ji-young Lee (이지영)] KakaoBank has applied an AI-based financial fraud detection model to its abnormal transaction detection system that analyzes behavioral flows before and after transactions.

KakaoBank said on Monday it developed an AI-based financial fraud detection model, the "sequence detection model", and applied it to its abnormal transaction detection system.

The sequence detection model is an AI model that analyzes not only the outcome of a single transaction such as transfers and withdrawals, but also the behavioral context around transactions. It applies an "attention" mechanism, in which AI understands relationships and flow among data, to assess transaction order, time gaps between actions and device-change behavior in an integrated way.

It analyzes customer behavior as a single flow, capturing patterns in which app access and transactions continue, activity stops at a certain point and then resumes. The company explained that such pauses often align with the process of voice phishing crimes.

Results were also confirmed. KakaoBank said the monthly average number of financial fraud prevention cases through monitoring its abnormal transaction detection system increased 4.4 times compared with before the rollout, after it introduced the model on a pilot basis in November last year.

In the first quarter of this year, when it began full operations, the share of all suspected financial fraud cases detected independently by the sequence detection model reached 49.8 percent.

It also detected cases involving "voice phishing recruitment accounts" and suspected "device transfer" cases in which a phone is changed and then handed over to a criminal organization, pre-emptively blocking transactions.

KakaoBank plans to continue upgrading the sequence detection model to respond to increasingly complex financial fraud methods.

A KakaoBank official said, "With the development and introduction of the sequence model, prediction of abnormal transactions based on behavioral flows has become possible." The official added, "Through technology research and development, we will respond pre-emptively to increasingly sophisticated financial crimes and create a safe financial environment."