Synthetic Identity Detection in Credit Applications (SIDCA)

A dataset consisting of historical records and application data, derived from the real-world operations of a major credit card company, is made available for research purposes. This dataset is designed to facilitate the detection and analysis of synthetic identities within financial systems. It comprises features extracted from historical transactional behaviors and personally identifiable information (PII), reflecting the applicant’s interaction history and associated metadata. The data includes indicators such as income, credit risk scores, employment status, and various behavioral metrics, collected both at the time of application and throughout the applicant’s relationship with the financial institution. The dataset is composed of two main components: Appl-Level Features: This component includes static features such as income, employment status, credit scores, and various derived metrics like name-email similarity, customer age, and housing status. These features were obtained at the time of application and serve as a baseline for assessing the applicant’s financial behavior. Behavioral Metrics: The second component includes dynamic features such as transaction velocity, session length, and the frequency of changes in key identifiers (e.g., phone numbers, addresses). These metrics were recorded continuously, reflecting the applicant’s ongoing interactions with the credit card system, and are available at monthly intervals for one year.