The financial industry has joined forces to block illegal remittances. President William Lai and Cabinet Minister Cho Jung-tai regard Taiwan’s rampant fraud as a top priority. To further strengthen cooperation, 35 banks have joined the Hawkeye Alliance, established in partnership with the National Police Agency, making it a highlight of cooperation against fraud. Recently, through the alliance, the Agricultural Bank of Taiwan has supported small and medium-sized financial institutions with limited resources to learn more about fraud patterns. Agricultural Bank is the leader in credit for grassroots agricultural and fisheries associations.

Tsai Pei-ling was a prosecutor for the Anti-Money Laundering Office of the Executive Yuan before becoming deputy general manager of Taipei Fubon Bank’s financial security department. She has used the money laundering prevention expertise she has accumulated over her career to develop the so-called “Hawkeye Model.”

Balancing efficiency with fraud prevention

The Hawkeye Model sets parameters for AI to filter. Tsai explains that authorities used to give banks specific characteristics to block transactions. Now, through Hawkeye, AI can make use of a full set of characteristics, weighing them concurrently to find the best balance between efficiency and fraud prevention.

Before Hawkeye, banks used to review 60,000 fraud cases a month, but now, they only review 300. Considering daily transaction needs, the model can initially separate problematic customers out from the rest for manual review, making fraud prevention more efficient.

Even more than the technology, the most essential element of the partnership is sharing information on pattern fraud patterns. Tsai stated that any financial institution can join the Hawkeye Alliance, no matter how small. According to feedback from the banks, the alliance has three main benefits: he lack of capital costs, the technical team, and the fraud patterns which can be accessed by QR code.

Grassroots financial institutions join

The Hawkeye Alliance is now upgrading to version 2.0. Besides allowing grassroots financial institutions to join, another highlight of the new version is joint learning, providing deeper cooperation among allied banks.

The core values of the Hawkeye Alliance are exchange of information and cooperation, so Fubon Bank assists financial institutions that are financially or organizationally deficient to prevent fraud, preventing breaches in joint financial defenses. Small and mid-sized banks, as well as grassroots financial institutions, are good examples. By joining, they can reduce their financial burden through Fubon Bank’s authorization for free use, gaining experience points .

Through information sharing, small and mid-size banks or grassroots financial institutions can see the whole picture of fraud in Taiwan, so that they do not fall victim to patterns already identified and blocked by big banks, which they might not have otherwise shared. Big banks gather more diverse patterns mostly through their larger consumer bases. The public stock banks, for instance, have a variety of services and products with differing policies, which is ideal for modeling, which requires a large training set. Data which is too sparse will affect accuracy, so small institutions cannot effectively make their own models.

Agricultural Bank is regarded as the “central bank of agricultural credit” due to managing the credit departments of many agricultural associations. Tsai said that when its head of compliance brought a team to visit Fubon Bank, she was deeply impressed that almost no fraud has occurred in Agricultural Bank or its credit departments. Agricultural Bank said that the most valuable assets in the financial industry are experience and knowledge, and that fraud prevention today does not equate to fraud prevention tomorrow. A bank which does not get timely information will eventually be defrauded. Fubon Bank has authorized Agricultural Bank to use its models, and it is also currently conducting an inventory of Agricultural Bank’s data systems.

35 banks have now joined the Hawkeye Alliance for joint learning, but the biggest highlights of Hawkeye 2.0 are customer privacy and the premise that data cannot be shared directly. To allow for joint learning and pattern sharing under these conditions, large servers must be built. Eight companies are jointly piloting this project; the results of the trial will be seen as soon as the end of the year.

Another highlight of Hawkeye 2.0 is “Hawkeye Messages,” an inter-agency warning system. After implementation of both this and the joint learning, according to Tsai, “information will circulate much more quickly.” This also includes mutual transaction warnings between financial institutions, which will make the models more efficient.

Non-banks join the fight

Another important area of development in the Hawkeye Alliance is the admission of non-banks to the system, solidifying fraud prevention in the financial industry.

Non-banking industries including securities and life insurance have all expressed desire to join the next stage of “inter-industry defense.” Virtual currency and third-party payments platforms work closely with banks, so they particularly hope to access the mutual transaction warnings of the alliance.

Tsai explain that through observation of fraud patterns, rather than just transaction size thresholds, it is also important to pay attention to other characteristics of suspicious accounts. Banks hope to learn other characteristics from others to more accurately distinguish suspicious accounts.

Hawkeye better than Gray List Platform

Like the Hawkeye Alliance, the Gray List Platform, launched by financial companies under the instruction of the FSC, also features shared warnings, but the latter focuses on the number of scheduled transfers to a certain account. For example, if 100 people around the country transfer money into an account, it will be flagged, and if the transfers exceed a certain amount, every bank will consider the flagged account as suspicious. Hawkeye Alliance, on the other hand, judges the transaction based on the bank’s full understanding of the customer, with a wider scope than just scheduled transfers.

The Police Agency concurred that the problem goes beyond just volume of scheduled transfers. For example, a flagged account might be the fourth or fifth level in a series of pass-through accounts, yet be the only account to get flagged by the platform. The previous accounts act as intermediaries, creating a chain of accounts from as little as two other smaller accounts, but the amounts transferred could be tens of millions of Taiwan dollars. Because the Gray List Platform only looks at scheduled transfers, these pass-through accounts slip though the net. In addition to large transfers, there are also cases where each transaction is small, but the frequency is high, such as 100 small transfers throughout the course of a day. Currently, this type of transaction is not included in the number of scheduled transfers into an account.

The methods of organized fraud have seen considerable innovation recently. In the past, because each transfer increased the chances to getting caught, the typical pattern was a small number of large transactions. The risk of just one transfer was much less than that of ten or more transfers, and if just one victim reported their loss, the whole scheme would be blocked. Now, however, scammers use “many small meals,” disguising fraudulent accounts among normal-looking transfers. The illegal nature of these transactions can be uncovered through exchange and mutual inquiries between banks as well as warnings of suspicious behavior.

Without the Hawkeye Alliance and the messaging mechanism, banks could not detect criminal transactions broken up into parts, but now, they can use a larger model to get a more complete picture of fraud methods.

The author is a senior media executive and a long-time follower of financial news