FraudLens
Document Fraud Detection
- AI-powered document fraud detection capability
- Uncover fraudulent documents and prevent first-party fraud
- X-ray vision capability to recover the original document
- Metadata analysis and file forensics to look for evidence of document tampering
- Our algorithms are informed by over 10M+ documents across multiple financial institutions, various document types, and commonly used PDF editors
Documents are scored based on the findings of the document assessment and are categorized by high, medium, low, and no fraud risk.
FraudLens Indicators
X-ray Vision
Checks the history of files and produces a "recovered version" of the uploaded PDF
Mismatched Dates
Checks if creation and modification dates are the same
Edited Metadata
Checks if metadata itself has been edited and in some cases, recovers the original metadata
Third-party Software Flag
Checks if the PDF was edited using any third-party PDF editor or common editing software
Edited Text
Flags regions where text has been edited but the actual text could not be recovered
PDF Alignment
Detects regions with inconsistent alignment as per document layout
Overlaid Text
Checks if text has been added on top of a scanned PDF/Image
Masked Text
Detects regions with hidden text
Flattened PDF
Checks PDF for certain marks that indicate it was flattened - flattening a file removes all font information and metadata from the document
Font Anomaly
Flags characters or words with inconsistent font in between a contiguous text block/stream
Anomaly Detection
Flags regions such as "Name & Address," "Date Format,""Transaction Amount," or any other regions of interest have been edited
Template Detection
Flags documents created using third-party PDF templates
Copycat Image
Compares uploaded document image with our document database and flags the document if similarities are detected
Copycat Text
Compares document uploaded with our document database and flags the document if similarity is detected
Brush Detection
Detects if an image has been tampered with based on pixelated speckles and distortions
Fingerprinting
Compares uploaded bank statements with genuine templates and flags the anomalies
