Detecting Suspicious Transactions with AI... Upbit Protected 120 Billion Won

Blocking Suspicious Transactions After Learning Financial Transaction Information Collection

It has been revealed that the real-time Fraud Detection System (FDS) of Upbit, a virtual asset exchange operated by Dunamu, preserved user assets worth 120 billion KRW.

Blockchain and fintech specialist Dunamu announced on the 7th that its AI-based FDS has blocked virtual asset-related crimes amounting to over 120 billion KRW cumulatively, preventing user damages.

The FDS collects electronic financial transaction information such as users' transaction histories, learns patterns, and detects and blocks transactions similar to abnormal patterns. Banks, securities firms, and virtual asset exchanges operate FDS to respond to increasingly sophisticated external threats.

The FDS commonly used in the banking sector operates based on rules such as "If an amount exceeding 1 million KRW is remitted or transferred, the same amount cannot be withdrawn or transferred from an ATM for 30 minutes after the deposit."

Upbit also introduced the "Virtual Asset Withdrawal Delay System" in November 2020, which restricts withdrawals of virtual assets equivalent to the amount of KRW deposits for 24 hours. In April 2021, it applied the "KRW Withdrawal Delay System," allowing KRW withdrawals only 72 hours after the first virtual asset deposit. Going further, Upbit developed its own FDS that learns real-time abnormal transactions beyond rule learning based on machine learning technology. The AI-based Upbit FDS inspects all deposit and withdrawal attempts in real time and analyzes 'patterns' rather than specific conditions to detect new types of abnormal transactions.

A company official explained, "When an external attacker hijacks an Upbit user account, they impersonate the user with the obtained authentication information. Additionally, users deceived by phishing scams (voice phishing) send assets themselves under external control, making it difficult to distinguish from normal transactions." This is the reason why Dunamu, the operator of Upbit, introduced AI that learns transaction patterns rather than specific conditions into the FDS.

In fact, Upbit's FDS recently detected abnormal transaction patterns of a high-net-worth Upbit user serving in the military based on pattern learning and prevented asset theft. The attacker attempted to withdraw assets after hijacking the user's account but failed as it was caught by Upbit FDS's real-time monitoring.

Meanwhile, Upbit supports an "Account Lock" service for users who have difficulty accessing their accounts due to military enlistment, long-term overseas business trips, or suspected account hijacking. The account lock service can be requested through Upbit's 24-hour dedicated call center. Users who have been asked to withdraw virtual assets by unspecified external parties or have made such withdrawals can also receive assistance through the dedicated call center. Upbit FDS takes preemptive measures on suspicious accounts detected during 24-hour monitoring to prevent damage and shares this information with banks and investigative agencies. Immediate action is also taken on voice phishing reports received by banks.