September 2023 Summaries

As identity fraud becomes more sophisticated, Socure argues that effective verification requires high-quality, diverse data sources combined with continuously updated machine-learning technology rather than static, rules-based systems. Its investigations identified increased fraudulent applications tied to correctional facility addresses, improving its ability to flag such high-risk locations by 27%, while its deceased-identity database expanded by more than 140 million records using sources such as funeral homes and authoritative obituaries to reduce delays in public records. Specialized models are used to limit false deceased flags, distinguish legitimate applicants from synthetic identities, and help prevent barriers for groups such as immigrants and young people with limited credit histories. Socure also emphasizes address normalization using USPS standards and supplementary property data, enabling it to match over 15 million residential addresses absent from USPS records and identify ownership changes earlier than legacy providers. These capabilities are presented as particularly important for regulated sectors including gaming, broker-dealers, and transfer agents, where accurate addresses support compliance, customer access, and fraud prevention.

Following Berbix’s June 2023 acquisition and the release of Predictive DocV 3.0, the Document Verification team introduced version 4.0 SDKs for mobile and web, redesigning the process of capturing government IDs and selfies. The updates include faster full-screen capture, verification results in under three seconds, improved barcode detection, image validation, data extraction, and immediate prompts to retake images affected by blur or glare. The company reports that these changes have produced a 10% increase in conversion rates and a 15% rise in automatic capture rates, helping reduce abandonment in identity verification flows. The SDKs also support applications including financial-services verification, driver’s license checks, and age-restricted purchase validation, while accessibility features such as audio guidance aim to assist visually impaired users and machine-based decisions are intended to reduce bias.

Socure’s Email RiskScore v10.1 uses generative AI and natural language processing to improve digital fraud detection for financial institutions and merchants, reporting an approximately 4% reduction in false positives and a 2–6% increase in name-email correlation compared with previous versions. The system evaluates email-handle “perplexity” against patterns learned from billions of legitimate email addresses, flagging nonsensical, randomized, fake, or otherwise atypical handles as higher risk. It also incorporates domain analysis, fake and test email detection, fuzzy identity matching, and other identity attributes to assess email legitimacy. Explainable scores and reason codes are intended to provide customers with greater transparency, while the company says the enhanced model can help prevent more fraud and improve approval rates.