Securing the SWIFT Ecosystem with Synthetic Data

This whitepaper discusses the development and application of synthetic SWIFT messages, particularly focusing on the MT103 message type, within the financial services sector. It outlines the necessity for synthetic data in ensuring regulatory compliance, protecting data privacy, and facilitating effective training and scenario testing. The document details the challenges faced in generating synthetic SWIFT messages, including the need to maintain the structure and information integrity of original messages while masking personally identifiable information (PII). A novel algorithm integrated into Mphasis Synth Studio is presented as a solution to these challenges, enabling the generation of synthetic data that preserves privacy and meets enterprise requirements. The methodology section describes the preprocessing and modeling approaches used, including the Conditional Tabular Generative Adversarial Networks (CTGAN) model, which learns the underlying distribution of features in the dataset. Performance metrics for assessing the quality of synthetic data are also discussed, ensuring that the generated data remains statistically similar to real data while maintaining usability for various applications in the financial industry.