In an increasingly digital financial landscape, synthetic data – which is artificially generated, but resembles real data – has become a valuable privacy enhancing technique (PET), says the Synthetic Data Expert Group (SDEG). The advisory group set up by UK’s Financial Conduct Authority (FCA) recently released a report on the use of synthetic data in financial services, providing insights on how it can be harnessed “effectively and responsibly”, as well as the opportunities and challenges it presents.
The value of synthetic data as a PET lies in its ability to enhance data usage and sharing without revealing sensitive underlying information. The report studies the application of synthetic data in three areas of the data lifecycle – data augmentation and bias mitigation; systems testing and model validation; internal and external data sharing.
Findings suggest that synthetic data “plays a crucial role across the data lifecycle” and has “the potential to help augment data quality and quantity”. This rings especially true in machine learning and AI models, where quality is “intricately linked to the quality of their data inputs”. Here, its possible applications include fraud detection, reject inference, and the generation of synthetic identities and customer patterns for testing open banking solutions.
Consider the options
Despite the many use cases, thorough evaluations should be conducted before applying synthetic data to determine if it is indeed the best approach, says SDEG. In some applications, other PETs, whether used alone or in combination with synthetic data may be the better option for improving model quality.
The report cautions organisations to be “thoughtful of ethical considerations and intentional about the use of synthetic data beyond technical development”. Firms should also introduce internal governance processes that “support the responsible and legal use of synthetic data”.