Model independently detected 49% of all suspected fraud cases in Q1; proactively flags suspected burner accounts and device handoffs
Kakao Bank said Monday it has developed an AI-based fraud detection model called the "sequence detection model" — which analyzes behavioral patterns before and after financial transactions to predict fraud risk — and has integrated it into its fraud detection system, or FDS.
Unlike conventional approaches that examine only the outcome of individual transactions such as transfers or withdrawals, the sequence model is an advanced AI model that also analyzes the behavioral context surrounding each transaction. It applies an "attention mechanism," which allows AI to understand correlations and flow between data points. Using this mechanism, the model organically links and evaluates the order in which transactions occur, the time intervals between actions and device-switching behavior — enabling more precise detection of fraud attempts disguised as normal transactions.
The model treats a customer's behavior not as isolated events but as a single continuous flow. It can capture subtle patterns such as a pause in activity mid-session — after an app login and a series of transactions — followed by a resumption. In many cases, such "activity interruption periods" coincide with the moment a voice phishing criminal is persuading a victim or inducing additional transfers. The sequence model analyzes this behavioral context in aggregate to assess risk.
After Kakao Bank piloted the sequence model in November last year, the number of fraud cases prevented through FDS monitoring rose by an average of 4.4 times per month compared with the period before its introduction. In the first quarter of this year, when full operations began, the model independently detected 49.8 percent of all suspected fraud cases blocked by Kakao Bank.
Detection results also showed notable success against new fraud techniques designed to evade conventional FDS systems.
One prominent example is the detection of "voice phishing collection accounts," commonly known as burner accounts, which have surged recently. When funds are repeatedly deposited from multiple unrelated parties but no withdrawals follow, conventional systems struggle to flag the account unless rapid withdrawals occur after deposits. The sequence model analyzed deposit patterns alongside time-of-use data to accurately classify such accounts as burner accounts, enabling preemptive action before stolen funds could be funneled to criminal networks.
The model also caught suspected cases of "device handoffs," in which a user changes their phone and then passes the device to a criminal organization. The sequence model detected anomalies by analyzing the device change immediately before a transfer, followed by subsequent app usage and transaction flow. Under a conventional rule-based system, such cases would likely have passed undetected due to their resemblance to normal past transaction patterns, but the new model allowed Kakao Bank to protect the funds in advance.
Kakao Bank said it plans to continue refining the sequence model to make its FDS framework more intelligent and to stay ahead of increasingly sophisticated fraud schemes.
"We are now able to detect sophisticated fraud methods in advance — ones that were difficult to catch from a single transaction record alone — by analyzing the continuous behavioral flow before and after transactions," a Kakao Bank official said. "We will continue to make our detection capabilities even more intelligent so that we can proactively protect our customers' valuable assets from ever more cunning financial fraud."
won@heraldcorp.com
