Skip to main content

List of Possible Fraud Types

NameDefinition
high velocityHigh amount of submissions from a specific device_id for several time periods. Results in Yellow color. Values (amount of submission per period to consider as High velocity) can be configured. Time intervals that are being used to calculate amount of submissions: Last day Last 2 weeks Last 30 days Last 6 minutes
critical velocitySame thing as High velocity, but with higher values and results in Red color. By default it is turned off. Values can be configured.
handwritten charactersHandwriting was detected in specific fields in the submitted document. Returns two granular sub-signals: fully_handwritten_fields (value entirely handwritten - lower risk) and handwritten_tampered_fields (printed value overwritten by hand - higher risk). The original handwritten_fields continues to work as before and returns the union of both.
LCD photoDocument was classified as a photo of a monitor/screen (desktop/tablet/phone). Used for bank checks too.
not a documentSubmitted image is probably not a document. This is an extension of is_document field specifically for fraud detection purposes. Used for bank checks too
screenshotSubmitted document is likely not paper based, but created digitally (like screenshots).
duplicateSubmitted document is probably a duplicate of a previously submitted one. The algorithm is based on image similarity and text. Uses is_duplicate field. Used for bank checks too
similar documentsSubmitted document's text has a high similarity with previously submitted document. Uses meta.duplicates field.
multiple profiles or devicesChecks for using several device_ids per account or several accounts per device_id during the last 30 days.
fraud historyMakes score higher if previously device_id submitted fraudulent documents. Time period is 30 days, and the minimum amount of submissions should be 5. Configurable by "div_coef". Higher - less aggressive this feature is.
emulated deviceImmediately return the red color if the document was submitted via app emulation software.
blocked deviceImmediately return the red color if device_id is in a blocked list.
aspect ratio mismatchThis feature supports bank checks. It compares the front check image's aspect ratio with the back check image's aspect ratio; a larger mismatch produces a higher fraud score.
fraudulent pdfDetects suspicious PDF characteristics, including editing software in metadata, overlapping or covered text, and inconsistent font usage. For example, PDF metadata identifying Photoshop or another editing tool may trigger this fraud type. The fraud score is based on the strongest detected PDF signal.
generated documentDesigned to detect images, created by AI (like chatGPT). Consists of checks: Visual check by a neural network. Exif data check for keywords like ChatGPT Visual check works only on images and not PDFs or other digital documents (screenshots for example). Exif check works when it is preserved in the document metadata and specific keywords are detected. Current set of keywords: chat, gpt, deepseek, gemini, openai
digital tamperingVisually detects cases where fraudsters digitally tampered the photo of a document. They can use software like photoshop or similar and digitally change the number/text/qr code in a document. They also can copy and paste parts of the same photo, for example making a total amount 9999 instead of 99. This feature is designed to detect those types of alterations and will work on photos when those alterations are visible. This Features is used for specific fields that are configurable, for example: "total", "subtotal", "date", "line_items.total", "line_items.price", "tax_breakdown.tax", "tax_breakdown.tax_base",
invalid qr dataQR code data on the document does not match the extracted fields.
stopwordsThe document contains language commonly associated with mock or test receipts (configurable stopword list).
vendor layout mismatchThe document layout does not match the expected template for this vendor.
invalid mrz dataMachine-readable zone data is invalid or inconsistent. This validation runs for passport blueprints.
total in words mismatchThe amount written in words does not match the numeric total.