Project AIKYA: Enhancing anomaly detection in financial transactions through decentralized AI

Recent advances in computing power, scalable storage and software tooling have empowered financial institutions to harness historical data for training machine learning (ML) models for a variety of use cases in predictive analytics and anomaly detection. These models unlock actionable insights that could help identify anomalies in payment systems and improve fraud detection, which is particularly relevant in payments ecosystems where critical insights are often distributed across geographies and institutions.

While these models have evolved significantly over the years, there is a potential to further augment their power by training such models on combined data sources from multiple institutions. However, centralized data repository spanning multiple institutions, but legal, regulatory and competitive considerations often render this approach impractical. As a result, decentralized AI has emerged as a class of techniques that move away from a single central entity toward distributed systems and networks. Federated learning (FL) enables decentralization of training and evaluating ML models while enabling privacy-enhancing techniques (PETs) to aid in maintaining a secure environment.

Project AIKYA is an exciting proof-of-concept (PoC) for federated learning, developed through a collaboration between Kinexys by J.P. Morgan and BNY. This exploration demonstrates the power of FL in institutional collaboration, proving that globally aggregated models can outperform individual ones by integrating the unique strengths that they each provide. This approach makes it a viable option for cross-border payments and other complex financial transactions. Following this PoC, several exciting directions are available to explore, including real-world data validation, scaling to multi-participant networks, testing FL models that support participants with heterogeneous technology stacks and/or model-agnostic networks, and testing across multi-region and multi-industry networks.