Leveraging Its Collective Intelligence Capabilities, NICE Actimize’s Federated Learning Approach Tackles Multiple Fraud Typologies, Including Application Fraud, Business Email Compromise and Real-Time Payments Fraud
NICE Actimize, a NICE business and leader in Autonomous Financial Crime Management, is introducing its Federated Learning capability that will provide financial services organizations (FSOs) with higher fraud detection rates across numerous fraud scenarios by leveraging NICE Actimize’s Collective Intelligence network. With this innovative cloud-based approach that uses machine learning analytics, FSOs can protect their institutions more effectively against multiple fraud typologies, including real-time payments fraud, while improving customer experience.
Traditional machine learning approaches require that the entire dataset is centralized. Meaning, there needs to be a specific database where the data resides to give FSOs the ability to build targeted analytical models based on this dataset. However, FSOs are often reluctant or prohibited from sharing datasets from a centralized location. To overcome this challenge, NICE Actimize is applying an innovative method of Decentralized Artificial Intelligence that includes federated learning of models learned in segregated datasets. The models are built for each organization separately, based on its own data, later to be utilized as features in the required context.
Through its application of Federated Learning, NICE Actimize has achieved compelling results in improving value detection rates. Using this approach has also proven to be effective in easing the process of model governance. Additionally, since the method is model agnostic, it can be applied across the range of NICE Actimize fraud solutions for different applications.
Using this Decentralized Artificial Intelligence approach, NICE Actimize research is showing successful results including:
- Improving detection rates and stopping more fraud with real-time payments
- Reducing false positives and detecting more fraud using anomaly detection in low fraud rates environments (like Business Email Compromise and other commercial banking fraud typologies) when trying to determine outliers that often correlate with fraud risk
- Tackling application fraud scenarios, considering monetary and non-monetary transactions as well as data collected from digital channels
“With NICE Actimize’s Federated Learning capabilities, and leveraging its unmatched collective intelligence, financial services organizations can maximize the power of the cloud to fight fraud collaboratively without compromising data security,” said Craig Costigan, CEO, NICE Actimize. “This advanced approach using agile analytics enables financial service organizations to move faster than the fraudsters while providing a frictionless customer experience.”
NICE Actimize monitors more than 3 billion transactions every day through its global network of financial services organizations. This extensive coverage provides significant collective intelligence and a view of multiple fraud scenarios across the market. Based on this collective intelligence, NICE Actimize provides consortium-based fraud analytics models as part of ActimizeWatch, a cloud-based managed analytics service that uses machine learning to proactively optimize analytics for members.