API reference / Document AI

Document forgery check

Scores one or two document images (public https URLs) for tampering. threshold (0-100, default 20) sets the sensitivity; text_fields lets you check printed values.

POST/v1/read/forgery
Authentication
Bearer token
Scope
read:document

Before you integrate

  • The answer is returned in this reply; there is nothing to poll for later.

Request body

FieldTypeRequiredDescription
urlsarrayYesOne or two public https URLs of document images to check, each up to 2040 characters.
thresholdintegerNoSensitivity from 0 to 100; defaults to 20. A lower number flags more images.
text_fieldsobjectNoPrinted values to compare with the image, as name and value pairs; names are 1 to 40 letters, digits or underscores and values up to 200 characters.

Sample code

Set your key in the GRAPPI_API_KEY environment variable. Sample values are placeholders.

curl -sS -X POST 'https://grappi.ai/v1/read/forgery' \
  -H "Authorization: Bearer $GRAPPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "urls": [
    "https://example.com/kyc/pan-front.jpg"
  ]
}'

Sample response

{
  "request_id": "req_01HZY3K8M2",
  "result": {
    "image_results": [
      {
        "copymove_detection": {
          "info": "No copy-move forgery found",
          "score": 0,
          "locations": [],
          "regions_of_concern": []
        },
        "photocopy_detection": {
          "info": "Not a photocopy",
          "score": "0.02"
        },
        "webcheck_detection": {
          "info": "Not found on the web",
          "score": "0.01"
        },
        "screenshot_detection": {
          "info": "Not a screenshot",
          "score": "0.03"
        },
        "splice_detection": {
          "info": "No splicing found",
          "forged": "no",
          "score": "0.04"
        }
      }
    ],
    "image_correlation_results": {}
  }
}

Response fields

FieldTypeDescription
image_resultsarrayOne entry per image, in the order you sent them.
image_results[].copymove_detectionobjectCheck for areas copied and pasted within the image, with info and score.
image_results[].photocopy_detectionobjectCheck for a photocopy of a document, with info and score.
image_results[].webcheck_detectionobjectCheck for an image taken from the web, with info and score.
image_results[].screenshot_detectionobjectCheck for a screenshot, with info and score.
image_results[].metadata_detectionobjectWhat the image's embedded details show, such as software and camera.
image_results[].splice_detectionobjectCheck for parts joined from other images, with info, forged, score and region_of_concern.
image_results[].textmatchobjectComparison of your text_fields with the printed text, with a score for each field.
image_correlation_resultsobjectComparison between the two images, when two were sent.

Errors

Every failure has the same shape. A call that fails on our side or at the data source is refunded to your wallet; a "no record found" answer is charged. See the error reference.