Document Liveness

This is a service that checks whether the photo on the document is a black and white photocopy, a color photocopy or a screenshot.

This is a AI service integrated with CAF, and we are continuously improving its accuracy. Currently, the accuracy may be lower for less common documents, such as passports, RNE, functional documents, and others. The service is more effective at identifying documents like RG and CNH. Remember that the result provided is a probability that the document is a screen photo or a photocopy, not a final decision.

If the decision attribute is false and the confidence level (confidence) is LOW, the rule will be approved. Make sure to carefully review these cases, as a low confidence level may indicate uncertainty in the provided result.

Section: documentLiveness

Required files: any document images

Schema

Attribute	Type	Description
statusCode	String	Query status (details)
data	Object	Return data
data.message	String	Return message
data.front	Object	Object containing the image result description (front of document)
data.front.decision	Boolean	Decision result for the front of the document
data.front.confidence	String	Confidence level for the front decision (e.g., "LOW", "MEDIUM", "HIGH")
data.front.indicators	Array	Array of indicators for the front of the document
data.front.indicators[x]	Object	Indicator object at index x
data.front.indicators[x].id	String	Indicator ID for the front of the document (e.g., "real_doc", "screen_photo_doc", "grayscale_copy_doc", "colored_copy_doc")
data.front.indicators[x].message	String	Message associated with the indicator for the front (e.g., "The document is probably real", "The document is probably a colored copy", "The document is probably a grayscale copy", "The document is probably a screen photo")
data.back	Object	Object containing the image result description (back of document)
data.back.decision	Boolean	Decision result for the back of the document
data.back.confidence	String	Confidence level for the back decision (e.g., "LOW", "MEDIUM", "HIGH")
data.back.indicators	Array	Array of indicators for the back of the document
data.back.indicators[x]	Object	Indicator object at index x
data.back.indicators[x].id	String	Indicator ID for the back of the document (e.g., "real_doc", "screen_photo_doc", "grayscale_copy_doc", "colored_copy_doc")
data.back.indicators[x].message	String	Message associated with the indicator for the back (e.g., "The document is probably real", "The document is probably a colored copy", "The document is probably a grayscale copy", "The document is probably a screen photo")

Example

{
  "documentLiveness": {
    "statusCode": "01",
    "data": {
      "data": {
        "front": {
          "decision": true,
          "confidence": "MEDIUM",
          "indicators": [
            {
              "message": "The document is probably real.",
              "id": "real_doc"
            }
          ]
        },
        "back": {
          "decision": true,
          "confidence": "HIGH",
          "indicators": [
            {
              "message": "The document is probably real.",
              "id": "real_doc"
            }
          ]
        }
      }
    }
  }
}

Validations Rules

Rule	Description
is_document_liveness_screen_photo_doc	This rule checks whether the document provided is a screen photo, which could indicate an attempt at fraud. If the document is identified as a screen photo, the rule will be considered invalid, indicating that the document is not authentic.
is_document_liveness_colored_copy_doc	This rule checks whether the document provided is a color photocopy, which could indicate an attempt at fraud. If the document is identified as a color photocopy, the rule will be considered invalid, indicating that the document is not authentic.
is_document_liveness_grayscale_copy_doc	This rule checks whether the document provided is a black and white photocopy, which could indicate an attempt at fraud. If the document is identified as a black and white photocopy, the rule will be considered invalid, indicating that the document is not authentic.