Fraud Detection · Risk Management
Swift needed to improve fraud and anomaly detection for global financial transactions without copying or moving sensitive member data out of secure locations.
Reported impact
Swift is creating a privacy-preserving fraud-detection approach that can improve financial-crime prevention while keeping member data confidential and operationally distributed.
Swift and Microsoft are building a federated anomaly-detection platform for financial crime on Azure Machine Learning and Azure confidential computing. Rather than moving member data to a central repository, Swift sends models to participants’ local compute environments, trains against distributed datasets at the edge, and then fuses the training results into a shared model. Confidential computing and supporting governance controls keep data protected during collaboration, while the Azure ML environment manages the training workflows required for the anomaly-detection use case. The design lets more than 11,500 financial institutions contribute insights without copying transactional data out of secure locations, preserving privacy while improving pattern detection. For Swift, the architecture creates a path to detect fraudulent transactions more accurately, with stronger security, privacy, and cost efficiency than traditional centralized approaches.
Transactional data stays in participants’ secure environments, Azure Machine Learning sends models to local compute for training, Azure confidential computing protects execution, and the resulting training outputs are fused into a shared anomaly-detection model.
Rather than bringing the data to a central point, we do the opposite. We send the model for training to the participants’ local compute and datasets at the edge and fuse the training results in a foundation model.
Swift’s network and platform is trusted to provide seamless automated transmission, receipt, and processing of more than nine billion financial messages a year.
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.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Microsoft Sample Nov 13, 2023 Belfius Managed feature store increases agility in building models because users can discover and reuse features instead of starting every time from scratch. It encourages faster experimentation with the ability to do local… Every year Belfius bank processes hundreds of millions of transactions, checking each one for potential money… Azure Databricks · Azure Machine Learning · Azure Synapse Analytics 
In-house Sample Sep 24, 2026 Mastercard Fraudsters are using AI to find vulnerabilities, adapt tactics, and scale attacks across more payment channels, while traditional transaction-by-transaction controls struggle to catch coordinated fraud and authorized… Last year, scam losses globally were estimated at $442 billion 
NVIDIA Sample Jun 2, 2026 Revolut Revolut needed to replace siloed fraud, credit, marketing, and recommendation models that depended on months of feature engineering with a single behavioral foundation model that could learn from raw financial event… 16.1% improvement in credit risk prediction performance from NVIDIA Llama-Nemotron-Embed-1B-v2 NVIDIA H100 GPUs · NVIDIA Llama-Nemotron-Embed-1B-v2 · Nebius AI Cloud 