Validating synthetic data is key challenge, says FCA

230621-fca-comments-on-synthetic-data-roundtable-1687453418

The Financial Conduct Authority (FCA), the UK financial regulator, has published insights derived from its roundtable on synthetic data validation, which it hosted in collaboration with the UK’s Information Commissioner’s Office and The Alan Turing institute. The event, which took place in March involved participants from financial institutions, synthetic data vendors, regulatory agencies, and academic institutions.

Synthetic data is generated artificially, and aims to mimic real-world data without containing personally identifiable information. It is commonly used in situations where access to real data is limited or restricted, allowing for model development and information sharing without compromising privacy.

Key insights highlighted by the FCA are:

  • “The challenge of validating synthetic data from utility, fidelity and privacy perspectives remains a critical barrier to synthetic data adoption.”

  • “Understanding the requirements of a specific use case is paramount when assessing both utility and privacy.”

  • “Model generalisability may be one method to increase the utility of synthetic data generators across multiple use cases, however this can lead to challenges with model drift and re-identification.”

  • “Mathematical methods that validate the model that generated the synthetic dataset (the synthetic data generator) need to be supplemented with post-generation validation of the synthetic dataset.”

  • “To drive adoption, the industry could shift to a risk-based model for privacy validation that accepts some level of inherent risk in generating and sharing synthetic data.”

  • “Beyond validation, there are several ways to advance synthetic data adoption including use case documentation, standards and frameworks for adoption, and regulatory guidance.”

The FCA claims that it intends to “further explore these themes through the FCA’s recently established Synthetic Data Expert Group, which comprises 22 members across financial services firms, academia, synthetic data vendors, public sector organisations and others”. Further information on the Synthetic Data Expert Group can be found here.

Full FCA comments on the roundtable are available here.

[Image Source: Financial Conduct Authority]