Global payments messaging network Swift has partnered with Google Cloud to develop anti-fraud technologies using advanced artificial intelligence (AI) and federated learning.

The partnership aims to better combat cross-border payment fraud, as current fraud detection methods struggle to keep up with increasingly sophisticated criminal tactics.

Swift said it plans to launch a sandbox with synthetic data to prototype learning from historical fraud in the first half of 2025.

Rachel Levi, Swift's head of artificial intelligence, said that the project aims to help banks and other stakeholders validate whether federated learning technology can help financial institutions stay one step ahead of bad actors by sharing fraud tags, enabling them to provide a better cross-border payment experience to their customers.

The move will enable multiple financial institutions to train a robust fraud detection model while preserving the confidentiality of sensitive transaction data.

The system will use federated learning and confidential processing techniques, such as Trusted Execution Environments (TEE), to enable secure, multi-party machine learning without displacement of training data.

The approach, which combines privacy enhancing technologies (PET) to deliver collaborative intelligence without compromising proprietary data, involves collaboration between Swift, 12 global financial institutions, and Google Cloud as a strategic partner.

Swift allows a copy of its anomaly detection model to be sent to each participating bank, with each financial institution training the model locally on its own data.

Only the learning from the training, and not the data itself, is transmitted back to a central aggregation server, managed by Swift, which aggregates the learning to enhance Swift's global model.

Rhino Health and Capgemini will also collaborate on the project in both the implementation and deployment phases.

The initiative builds on Swift's existing Payment Controls Service (PCS), a system designed to increase the security of financial transactions by detecting and intercepting payment anomalies in real time.

It follows a pilot project with financial institutions in Europe, North America, Asia and the Middle East.

Andrea Gallego, managing director, global GTM incubation, Google Cloud, said that the collaboration with Swift illustrates the transformative potential of federated learning and confidential computing.

“By enabling secure collaboration and knowledge sharing without compromising data privacy, we are fostering a safer and more resilient financial ecosystem for everyone,” she added.

Google Cloud explained in a blog post how federated learning can offer powerful solutions for the collaborative training of artificial intelligence models without compromising privacy and confidentiality.

The post, written by Google Cloud’s heads of technology and incubation Vineet Dave and Arun Santhanagopalan, said: “To be effective, this approach requires data sharing across institutions, which is often restricted because of privacy concerns, regulatory requirements, and intellectual property considerations.”

Sudhir Pai, chief technology and innovation officer, financial services, Capgemini said that federated learning as a means to combat fraud is growing.

“Payment fraud stands as one of the greatest threats that undermines the integrity and stability of the financial ecosystem, with its impact acutely felt upon some of the most vulnerable segments of our society,” he said.

---