Liveness Verification

Detect presentation and injection attacks in GBG GO by confirming a real, live person was in front of the camera when the selfie was captured.

The Liveness Verification module determines whether a real, physically present person was in front of the camera at the point the selfie photograph was taken. It detects presentation attacks such as:

• Attempts to spoof the verification process using a printed photograph
• A screen replay
• A video recording
• A 3D mask
• A digitally injected (deepfake) image.

The module returns a confidence score indicating the likelihood that the captured selfie represents a genuine, live person.

The module has two variants. Both analyse a selfie image to confirm genuine presence, but they differ in result type and detection scope.

Variant	Result type	Detection scope	Review tier
V2	Numeric score (0–100)	Presentation attacks	Available
Injection Attack Detection	Binary (Live / Not Live)	Presentation attacks + injection attacks	Available

Liveness Verification V2

The V2 variant analyses a selfie image captured during the onboarding journey and produces a score that reflects the system’s confidence in the subject’s genuine presence at the point the selfie photograph was taken. It detects presentation attacks such as printed photographs, screen replays, video recordings, and 3D masks, and is also supported as a standalone module.

Capabilities

The module returns a single capability: a liveness score.

Liveness score

Returns an integer score from 0 to 100 indicating the system’s confidence that the selfie was captured from a real, physically present person at the time of the photograph. A higher score indicates greater confidence that the selfie is genuine. A score of 0 indicates the check could not produce a result and should be treated as inconclusive rather than a definitive failure.

Detail	Description
Type	Integer
Range	0–100
Default	0

A higher score indicates greater confidence that the selfie is genuine. A lower score indicates a higher likelihood that the image is a presentation attack or does not represent a live person.

The score reflects the combined output of multiple analysis signals, which may include

• Texture analysis: Detecting the flat, reflective surface of a printed photo or screen.
• Depth estimation: Distinguishing a three-dimensional face from a two-dimensional image.
• Edge and boundary detection: Identifying the borders of a held-up photograph or device screen.
• Digital injection detection: Identifying images that weren’t captured by the device camera but inserted into the capture pipeline programmatically.

A score of 0 indicates that the liveness check could not produce a result. This may occur if the selfie image was corrupted, if no face was detected in the image, or if a system error prevented analysis. Evaluation rules should treat a score of 0 as an inconclusive result rather than a definitive failure.

Default outcomes

The module uses score thresholds to classify the liveness result into one of three outcomes. These thresholds are pre-configured as default outcomes and can be used in evaluation and routing logic within the journey designer.

Outcome	Condition	Description
Success	Score ≥ 75	The system has high confidence that a real, live person was present when the selfie was captured. The journey can proceed to downstream biometric checks such as facematch verification.
Review	Score ≥ 60 and < 75	The system’s confidence is moderate. The selfie may be genuine, but the analysis detected some signals that could indicate a presentation attack, or the image quality reduced the system’s ability to make a definitive determination. Common causes include poor lighting, extreme camera angles, or partial face occlusion. This outcome typically routes to manual review, where a human operator can visually inspect the selfie and the surrounding session data.
Fail	Score < 60	The system has low confidence in genuine presence. The selfie is likely a presentation attack, a photograph of a photograph, a screen replay, a mask, or a digitally injected image. This outcome typically routes to rejection or escalation, depending on your fraud policy.
ERROR	Default (no conditions matched)	The liveness check could not be completed. This fallback outcome typically indicates a system-level issue, such as a service outage or image processing failure, rather than a spoofing attempt.

Input payload

The following is a sample payload used to submit a selfie image to the Liveness Verification V2 variant for processing.

JSON

{
  "resourceId": "<resourceId>",
  "context": {
    "config": {
      "delivery": "api"
    },
    "subject": {
      "biometrics": [
        {
          "type": "Selfie",
          "selfieImage": "<base64-encoded string>"
        }
      ]
    }
  }
}

Field	Required	Description
type	Yes	Identifies the biometric object. Set to Selfie.
selfieImage	Yes	Base64-encoded selfie image captured from the end user.
biometrics	Yes	Accepts at least one item.

Sample response

The following is a sample response for a successful liveness check from the Liveness Verification module.

JSON

{
  "response": {
    "advice": {
      "contains": {
        "score": 98
      }
    },
    "outcome": "Success"
  }
}

Liveness Verification with Injection Attack Detection

The Liveness Verification with Injection Attack Detection variant extends the standard liveness check with an additional layer of security that detects injection attacks at the capture level, where an image is programmatically inserted into the capture pipeline rather than captured through the device’s camera. Requires an EncryptedSelfie input with selfieImageEncryption set to true.

Capabilities

The module returns two capabilities: a face liveness result and an injection attack detection result.

Face liveness result

Indicates whether the face in the selfie is determined to be live, meaning a real, physically present person was in front of the camera at the time of capture.

Value	Description
Live	The face is live. The system has determined that a real, physically present person was in front of the camera when the selfie was captured.
Not Live	The face is not live. The system has determined that the selfie does not represent a genuine, live person. This might indicate a presentation attack such as a printed photograph, screen replay, video recording, or 3D mask. This is the default value.
Error	The face liveness check could not be completed due to a technical issue.

Injection attack detection result

Indicates whether an injection attack was detected during the selfie capture process. An injection attack occurs when an image is programmatically inserted into the capture pipeline rather than being captured through the device’s camera in real time.

Value	Description
Live	No injection attack detected. The selfie image was captured through a legitimate device camera. This is the default value.
Not Live	The face is not live. The system has determined that the selfie does not represent a genuine, live person. This might indicate a presentation attack such as a printed photograph, screen replay, video recording, or 3D mask. This is the default value.
Error	The injection attack detection check could not be completed due to a technical issue.

Default outcomes

The module combines the face liveness result and the injection attack detection result to produce an overall outcome. These default outcomes can be used in evaluation and routing logic within the journey designer.

Outcome	Condition	Description
Success	Face liveness result is Live and injection attack detection result is Live	The subject is confirmed as genuinely present and the selfie was captured through a legitimate device camera. The journey can proceed to downstream biometric checks such as facematch verification.
Review	Face liveness result is Live or injection attack detection result is Not Live	The face liveness check passed but an injection attack was detected, or vice versa. The results are mixed, so this outcome typically routes to manual review for further inspection.
Fail	Face liveness result is Not Live and injection attack detection result is Not Live	Both checks failed. The subject is not confirmed as genuinely present and the selfie appears to have been injected. This outcome typically routes to rejection or escalation.
ERROR	Default (no conditions matched)	An unexpected error occurred during processing. This fallback outcome typically indicates a system-level issue rather than a spoofing attempt.

Input payload

The following is a sample payload used to submit an encrypted selfie image to the Liveness Verification with Injection Attack Detection variant for processing.

JSON

{
  "resourceId": "<resourceId>",
  "context": {
    "config": {
      "delivery": "api"
    },
    "subject": {
      "biometrics": [
        {
          "type": "encryptedSelfie",
          "selfieImageEncryption": true,
          "selfieImage": "<base64-encoded string>"
        }
      ]
    }
  }
}

Field	Required	Description
type	Yes	Identifies the biometric object. Set to encryptedSelfie.
selfieImageEncryption	Yes	Must be set to true to indicate the selfie image has been encrypted for injection attack detection.
selfieImage	Yes	Base64-encoded encrypted selfie image.

Sample response

The following is a sample response for a successful check from the Liveness Verification with Injection Attack Detection variant.

JSON

{
  "response": {
    "advice": {
      "equals": {
        "liveness": "Live",
        "injectionAttackDetection": "Live"
      }
    },
    "outcome": "Success"
  }
}

Liveness Verification

Capabilities

Liveness score

Returns an integer score from 0 to 100 indicating the system’s confidence that the selfie was captured from a real, physically present person at the time of the photograph. A higher score indicates greater confidence that the selfie is genuine. A score of 0 indicates the check could not produce a result and should be treated as inconclusive rather than a definitive failure.

Detail	Description
Type	Integer
Range	0–100
Default	0

Input payload

The following is a sample payload used to submit data to the Liveness Verification module for processing.

JSON

{
  "context": {
    "subject": {
      "biometrics": [
        {
          "faceImage": "<base64-image>"
        }
      ],
      "identity": {}
    }
  }
}

Sample response

The following is a sample response returned by the module.

JSON

{
  "response": {
    "advice": {
      "score": 99
    },
    "outcome": "Fail"
  }
}