Cross-Document Validation
Cross-document validation is the process of verifying the consistency of information across multiple documents submitted by the same individual or entity. It detects inconsistencies that reveal document fraud or identity theft.
Cross-document validation goes beyond individual verification of each document by analysing the overall consistency of the file. A document may appear authentic when taken in isolation but reveal inconsistencies when compared against other documents in the file. This approach is particularly effective against sophisticated fraud that uses individually convincing but collectively incompatible documents.
Cross-checks cover several axes: identity data consistency (are the name, date of birth, and address identical across all documents?), temporal consistency (are the issue dates, covered periods, and amounts compatible?), logical consistency (is the declared salary compatible with the tax notice? Does the declared address match the bank statement?), and consistency with external databases.
Advanced cross-validation solutions use OCR and NLP to automatically extract structured data from each document, then apply business rules and machine learning algorithms to assess overall consistency. A global consistency score is calculated, and detected inconsistencies are flagged with their criticality level to facilitate decision-making.
Regulations
Real-world examples
- 1.During a mortgage application, the bank detects that the salary shown on the payslips does not match the declared income on the tax notice, revealing payslip falsification.
- 2.A rental application is rejected because the address on the proof of address does not match the bank statements, and the employer name on the employment contract differs from the payslip.
- 3.A leasing company cross-references data from the company registration certificate, financial statements, and bank details to verify the consistency of a business financing application, detecting that the registration number on the certificate does not match the one in the financial statements.