Onfido logo home page
Get in touch
Arrow back Back to guides

Document report: suggested client actions

Introduction

The Document report employs a variety of techniques designed to identify fraudulent documents. We can support this with expert manual processing to ensure a check is subject to a thorough analysis.

This guide is for clients using API v3 who use the Document report in their service. We look at specific scenarios, examine individual results from the report breakdown, and describe some common suggested actions.

Before using this guide, we recommend that you read our API documentation about the Document report to familiarise yourself with information on individual breakdowns and how they map to individual sub-results for the Document report (see below).

What the sub-result means

The sub_result field indicates a specific detailed result, and is only applicable to Document reports.

Possible values of sub_result are as follows:

  • rejected: We can't process the document image, or the document isn't supported by Onfido for processing. Alternatively, the age of the applicant may be too low (the standard threshold is 16 years old but you can write to your Onfido contact to have this changed).
  • suspected: Document shows signs of suspect fraud.
  • caution:: We can't successfully complete all verifications, but this doesn’t necessarily point to a suspected document (for example, expired document).
  • clear: All underlying verifications pass. There are no indications the document is fraudulent.

Breakdowns and sub-breakdowns in the response object

Individual breakdowns can have the values clear or consider.

Breakdowns are made up of sub-breakdowns. A breakdown will have a consider result when at least one of its sub-breakdowns contains a consider or unidentified result.

You can read how individual breakdowns contribute to the overall sub_result value in the Document report in our API documentation.

You can also review a page that describes the formats that individual fields in the Document report take (for example: arrays of objects, strings).

Example scenarios

The following scenarios are provided as examples, and are based on default configurations. Where breakdowns can be configured to map to alternative sub-results, this is noted. If you wish to discuss such configuration changes, please ask your contact at Onfido or email our Client Support team.



'rejected' sub-result scenarios



Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
image_integrity: supported_document Verify that the document is supported

You can review the full list of documents Onfido supports.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
image_integrity: image_quality Request an additional image of the same document

Any document which is deemed to be “unprocessable” is likely to have key data points which cannot be seen, are cropped, or are in some way obscured. Such documents are rejected for poor image quality.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
age_validation: minimum_accepted_age Block

Asserts whether the age calculated from the document’s date of birth data point is greater than or equal the minimum accepted age set at account level. The standard threshold is 16 years old but you can write to your Onfido contact to have this changed.



'suspected' sub-result scenarios



Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
visual_authenticity: original_document_present (can be configured to map to caution sub-result) a. Ignore
b. Refer to manual team for review
c. Request an additional document

Onfido deploys a number of texture analysis algorithms to detect whether or not an image contains any of the following:

  • screenshots
  • pictures of pictures
  • printouts

Original document scanning presents a difficult challenge that in our experience is best solved via the combination of tech and human expertise. If our algorithms fail to produce a result beyond a certain degree of certainty, documents will be referred to manual analysts to supplement the engine's decision.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
visual_authenticity: picture_face_integrity a. Block
b. Refer to manual team for review

Onfido deploys a number of algorithms designed to analyse whether the picture in a document image may be a physical insertion or otherwise digitally tampered.

If our algorithms fail to produce a result beyond a certain degree of certainty, documents will be referred to manual analysts to supplement the engine's decision.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
visual_authenticity: fonts a. Block
b. Refer to manual team for review

Our machine learning models are constantly refining their ability to recognise, categorise and distinguish between fraudulent fonts and genuine travel document font types such as OCRB.

If our algorithms fail to produce a result beyond a certain degree of certainty, documents will be referred to manual analysts to supplement the engine's decision.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
visual_authenticity: template a. Block
b. Refer to manual team for review

The template breakdown is triggered by algorithms trained to recognise genuine templates and formats of documents, and is supplemented by our manual team’s expert knowledge.

If our algorithms fail to produce a result beyond a certain degree of certainty, documents will be referred to manual analysts to supplement the engine's decision.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
visual_authenticity: digital_tampering a. Block
b. Refer to manual team for review

We analyse a document to recognise if the document is suspected of being created digitally or digitally tampered as opposed to being created physically or tampered in a physical manor.

If our algorithms fail to produce a result beyond a certain degree of certainty, documents will be referred to manual analysts to supplement the engine's decision.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
visual_authenticity: security_features a. Block
b. Refer to manual team for review

The Onfido engine is trained to pick out a number of distinct security features in each document, and is supplemented by the work of our manual team of experts.

If our algorithms fail to produce a result beyond a certain degree of certainty, documents will be referred to manual analysts to supplement the engine's decision.


Potential contributing breakdown Potential client action(s)
data_validation: gender, expiry_date, mrz, date_of_birth, document_numbers a. Block
b. Refer to manual team for review

The Onfido engine asserts whether algorithmically validatable elements are correct. For example, MRZ lines and document numbers.

If our algorithms fail to produce a result beyond a certain degree of certainty, documents will be referred to manual analysts to supplement the engine's decision.



'caution' sub-result scenarios



Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
data_validation: document_expiration a. Ignore
b. Refer to manual team for review
c. Request an additional document

The document_expiration sub-breakdown validates the expiry date extracted from a document and checks its validity. If this is flagged, the document has expired.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
image_integrity: conclusive_document_quality a. Ignore
b. Refer to manual team for review
c. Request an additional document

This breakdown is applied by our expert manual review team when both product and human analysis fail to yield a concrete result either way.

Our manual team will apply this breakdown for documents they receive which are unclassifiable as fraudulent or genuine.


Potential contributing breakdown and sub-breakdown(s) Potential client action(s)
image_integrity: colour_picture (can be configured to map to rejected sub-result) a. Ignore
b. Refer to manual team for review
c. Request an additional document

The Onfido engine deploys color reading algorithms designed to detect whether or not a black and white image has been submitted, flagging these.



'clear' sub-result scenario



We recommend that you allow the user to proceed in this scenario.