Insights
Learn how to identify risks and understand signals from verification checks.
Stripe Identity’s machine learning system considers a variety of signals when verifying a user’s identity. It examines a number of factors to produce insights that can give further clarity into Stripe’s decision. These insights are more nuanced than the top-level verification decisions, and you can use them to assist with manual reviews or customer support processes.
See the following definitions:
- Insights: The name Stripe uses to refer to the collection of all insights.
- Insight: The specific attribute scored (for example, blur, authenticity). It’s of either type
LevelorLabel.- Level: These insights provide a computed level, which is a score that translates to low, elevated, or high. Insights of this type evaluate the potential risk to verification.
- Label: These insights provide a binary value of being either present or absent. Some of these insights evaluate a potential risk to verification, but others might be neutral, with no inherent risk associated with them.
Document insights
These are the insights produced on document checks:
| Name | Type | Description |
|---|---|---|
| Activity on Stripe network | Label | Indicates whether the identity presented fails to match a Stripe-known identity. It’s not uncommon for an identity to be unknown, though it does represent higher risk than a known user. |
| Barcode or machine-readable zone | Level | Indicates if the machine-readable zone or barcode is present and decoded, but was manipulated. |
| Document authenticity | Level | Indicates whether the document presented might have been digitally manipulated or is otherwise inauthentic. |
| Fraud | Level | This comprehensive score evaluates various signals related to document verification, including signals from Stripe Network data. |
| Image blurriness | Level | Indicates the severity of the blurriness of the document image presented during the verification process. |
| Physical document detection | Level | Indicates whether the document presented is a screenshot of a document. |
| Risky behavior on the Stripe network | Label | Indicates whether the identity presented matches a Stripe-known identity with a history of high risk activity. |
Selfie insights
These are the insights produced on selfie checks:
| Name | Type | Description |
|---|---|---|
| Duplicate selfie | Label | Indicates whether the selfie presented matches a selfie from a previous identity verification. |
| Duplicate selfie with a data mismatch | Label | Indicates that the selfie presented matches a selfie from a previous identity verification, but the extracted document data from each doesn’t match. |
| Selfie and document match | Level | Indicates whether the selfie isn’t a high confidence match with the face on the document. |
| Selfie liveness | Level | Indicates that the selfie presented might not have been a live person in front of the camera, such as a photograph or screen presentation attack. |
Dashboard usage
The Dashboard page for a submitted VerificationSession contains a panel showing the insights generated for this session.