[ad_1]
The fight against money laundering and fraud is becoming increasingly expensive: high hopes in the detection and sorting of suspicious transactions rely on the use of artificial intelligence [19659002] 05.30, 15.07.2010
The daily struggle of banks against fraud, money laundering 39, money and sanction violates billions and keeps thousands of employees busy. And the effort increases. The attacks become more sophisticated, the regulations more stringent. So the institutes are increasing their staff in the compliance departments.
But high hopes in the detection and sorting of suspicious transactions rely on the use of artificial intelligence. Unlike conventional filtering, in which expert computers, based on known models, set clear rules for detecting fraud or money laundering, an algorithm "learns" similar to the human on the sample database. "Artificial intelligence has enormous potential for increasing efficiency and effectiveness," says Gerold Grbadhoff of consulting firm Boston Consulting Group (BCG).
Where the fraud detection works well
The detection of fraud with the help of artificial intelligence succeeds very well, "says Stefan Rüping from the Fraunhofer Institute for Intelligent Systems of Analysis and Information To evaluate whether a transaction is suspicious or not, many features are badyzed: where has the card been used? amount was paid? In the badysis of huge amounts of data, people quickly reach their limits, while the machine can examine countless combinations in parallel, thus finding rules and constantly adapting them Credit card transactions are made for it: The amount of data is large because of the large number of transactions.In addition, credit card owners report quickly if they are victims of fraud or if a t legitimate ransaction was blocked. "Fast feedback allows automated learning and systems can be constantly improved," says Rüping
. Modern technology can also help prevent payment fraud. According to the company itself, Commerzbank has discovered in recent years transfers totaling 100 million euros, in which companies have engaged in acts of fraud – for example, because it has acted as a manager and organized a referral. "We managed to stop 99 percent of the funds," explains payments manager Commerzbank manager Frank-Oliver Wolf . In order to filter fraud attempts, the financial institution is based on artificial intelligence
Banks are just beginning
But banks are still in the beginning. The use of artificial intelligence is up to now very limited and the fight against financial crime requires a lot of manual work. "In the fight against money laundering, payment fraud and the financing of terrorism, standards for the use of big data and artificial intelligence are being developed" said BCG consultant Grbadhoff. He burns the institutes here especially under the nails. Juvenile sentences are threatened with violations, but more and more transfers must be checked in shorter and shorter periods. The result: the costs increase sharply. According to a survey sponsored by Lexis Nexis Risk Solutions last year, German banks spend about $ 46 billion a year fighting money laundering.
The current systems based on detection rules developed by the experts are very high false positive rates. "In practice, the previous models are already very good, if only 95% of the reports of suspicious operations prove false alarms," says Norbert Gittfried of BCG. All transactions clbadified as suspicious must be manually checked by employees – usually only a few minutes, but a huge effort. After all, this can affect hundreds of thousands of transfers per year in major financial institutions. According to Lucas from Jongh's Croo's consulting firm Oliver Wyman, the use of artificial intelligence would allow banks to significantly reduce their costs. In addition, previously unrecognized models can be noticed and loopholes closed. "Even the small transactions that fell through the grid are discovered through artificial intelligence."
Technology is not a panacea
Technology is not a panacea, the number of false positives will continue to be high. "Machines will always be configured to report suspicious transactions when in doubt," says Gerald Kühn, who works at DZ Bank on payment transactions. "The worst thing that can happen is that money laundering or terrorist financing is no longer visible because the monitoring mechanisms have been relaxed."
It is not just the outdated computer systems of many banks and the lack of standards. A key problem is that no one really knows how self-learning systems come to their conclusions. However, the results of the algorithms must be understandable, requires not only financial supervision BaFin in a recently published study on the use of artificial intelligence. Apologies like "it was not me, the machine was," the regulators did not agree, writes Felix Hufeld, BaFin boss in June in a guest ticket in the "Handelsblatt". "You can automate a lot and delegate to algorithms, but not the responsibility."
Source link
