AI Use Case | Swift
With the rapid increase of cross-border transactions and the growth in instant payment networks, the financial services industry is focusing on fighting fraud that can cost hundreds of billions annually. As the leading infrastructure provider for financial messaging services, Swift, the company behind the widely-used Swift numbers for international transactions, has long collaborated with its community of over 11,500 institutions to develop new methods to detect and prevent fraudulent transactions. Now, Swift is working with Microsoft to further innovate in the fight against financial crime. Using federated learning techniques along with Azure Machine Learning and Azure confidential computing, Swift and Microsoft are building an anomaly detection model for transactional data—all without copying or moving data from secure locations. The shared vision is that the model will become the new standard for reducing financial crime while achieving the highest levels of security, privacy, and cost efficiency.
Using Azure Machine Learning, we can train a model on multiple distributed datasets. Rather than bringing the data to a central point, we do the opposite. We send the model for training to the participants’ local computer and datasets at the edge and fuse the training results in a foundation model.
Johan Bryssinck: AI/ML Product and Program Management Lead
Swift
Leading the charge for the industry
Citizens. Countries. Continents. The entire world relies on the efficiency, reliability, and security of Swift’s global financial messaging network. Swift sets the industry standard for how financial institutions and banks communicate with each other. The member-owned cooperative provides secure infrastructure for money and security transfers and related information for over 11,500 financial institutions worldwide. Swift’s network and platform is trusted to provide seamless automated transmission, receipt, and processing of more than nine billion financial messages a year.
Detecting and stopping fraudulent attacks and crime is a priority for Swift’s customers so the cooperative is constantly innovating to strengthen security. “We connect financial institutions around the globe to our network and deliver products and services in over 200 countries and territories. Our standards are used for financial institutions and banks to
communicate with each other, which puts us in a unique position to bring the industry together to try to solve some of the most complex problems,” says Tom Zschach, Chief Innovation Officer at Swift.
As the industry provides faster payment solutions between individuals and organizations across borders, Zschach says the risk footprint increases, as do costs. This is especially true when accounting for secondary impacts like fraud remediation and fund recovery, making fraud a massive issue for all financial intuitions and their customers. “Fraud in payments is a major concern for our clients—in the order of magnitude of billions of dollars every year,” says Isabel Schmidt, Co-Head of Global Payments Products at BNY Mellon, a member of the Swift network. “Preventing financial crime is a problem that we cannot solve individually as a bank, we need to it solve collectively.”
Innovating without sacrificing privacy
With the aim of creating a solution that combats the latest financial security threats, Swift, as part of its robust innovation agenda, has been collaborating with Microsoft. Using federated learning techniques in Azure Machine Learning combined with Azure confidential computing, Microsoft Purview, and a Zero Trust based policy framework, the two organizations are working to build a highly accurate anomaly detection model for financial transactional data. This is being done without copying or moving data from Swift members’ secure locations. Participants’ data remains confidential even while the new model detects anomalies and gains new insights that will help predict and prevent financial crime.
Johan Bryssinck, AI/ML Product and Program Management Lead at Swift, says the organization’s intent is to provide the model so Swift’s banking partners can share their own data to further educate the model. Once data has been collected, Azure confidential computing will securely run new models on the latest data to gain additional insights creating an ongoing cycle of learning that will help drastically improve the rate at which we can detect fraudulent financial transactions across the globe. “Our first ambition at Swift is to build a foundation model for anomaly detection that underpins the detection and prevention of fraud. Our ultimate goal is collaborating with Microsoft and our community to start thinking about how we can stop fraud occurring in payments,” says Bryssinck. “We are exploring the federated learning aspects of Azure Machine Learning where we take a model developed by Swift and further train and enrich it with additional customers’ data through Azure confidential computing.”
“Using Azure Machine Learning, we can train a model on multiple distributed datasets. Rather than bringing the data to a central point, we do the opposite. We send the model for
training to the participants’ local computer and datasets at the edge and fuse the results in a foundation model,” says Bryssinck.
Solving complex problems for the greater good
While the full potential of the model has yet to be reached, the shared vision is that it will become a potent new tool for reducing financial crime while achieving the highest level of security, privacy, and cost efficiency. “Working together with Microsoft and with our banking customers, we can build a model that’s much more accurate and much more performant than I think has ever been seen in financial services,” says Zschach. ”Microsoft is a very important and productive strategic partner for us. There’s a strong alignment in core values like trust and security. They’re able to bring the capability and the expertise to help us solve some of the toughest problems in finance.”
Our first ambition at Swift is to build a foundation model for anomaly detection that underpins the detection and prevention of fraud. Our ultimate goal is collaborating with Microsoft and our community to start thinking about how we can stop fraud occurring in payments.
Johan Bryssinck: AI/ML Product and Program Management Lead
Swift
The text included is sourced from Microsoft. This document is intended for educational and informational purposes only. All trademarks, service marks, trade names, product names, and logos appearing in this document are the property of their respective owners.