Synthetic face signals detected before onboarding completes.
ASTERIA KYC deepfake defense runs on every selfie and video capture to detect AI-generated faces, face swap signals, temporal artifacts, and media injection — before the biometric result is returned.
Six deepfake signals. Every capture analyzed.
No single signal is sufficient to confirm or deny a synthetic face. All six signal types run in parallel and combine into one risk assessment.
Synthetic Face Indicators
GAN-generated face artifacts, texture regularity, and feature blending anomalies assessed across the full face region.
Face Swap Detection
Boundary inconsistencies, skin tone discontinuities, and blending artifacts at the face edge detected in composite media.
Temporal Consistency
Frame-to-frame consistency of facial geometry, landmark positions, and micro-expressions assessed across video captures.
Texture Artifact Analysis
Pixel-level texture patterns inconsistent with natural skin. Compression artifacts and generative model smoothing signals detected.
Frequency Domain Analysis
DCT and FFT frequency signatures analyzed for patterns associated with generative model outputs and upscaling artifacts.
Injection Signal Detection
Virtual camera, screen recording, and media injection tool signals detected at the capture layer before biometric analysis.
Deepfake risk score and detected signal breakdown.
Every session returns a deepfake risk score, detected signal types, confidence band, and reason codes for downstream review.
| Field | Description | Values |
|---|---|---|
| deepfake_result | Overall deepfake assessment | clear / uncertain / detected |
| deepfake_risk_score | Synthetic face risk score | 0–100 |
| face_swap_signal | Face swap detection outcome | none / detected / inconclusive |
| texture_signal | Texture artifact assessment | clear / anomaly / failed |
| temporal_signal | Temporal consistency check | consistent / inconsistent / not_applicable |
| reason_codes | Structured reason array | array of strings |
{
"deepfake_result": "clear",
"deepfake_risk_score": 8,
"face_swap_signal": "none",
"texture_signal": "clear",
"temporal_signal": "consistent",
"reason_codes": ["DEEPFAKE_ALL_CLEAR"]
}Three deepfake risk states. One structured response.
Deepfake signals are probabilistic, not binary. Uncertain results route to human review rather than automatic rejection.
Clear
All deepfake signals returned below the risk threshold. No synthetic face, face swap, or injection artifacts detected above the configured level.
Uncertain
One or more signals returned ambiguous results. Session routed to human review with full signal breakdown and capture evidence attached.
Detected
One or more deepfake signals detected above the configured threshold. Reason codes returned with detected signal types and risk score.
Ready to turn identity verification into a controlled compliance workflow?
Use ASTERIA KYC to verify users, screen risk, and preserve evidence from one connected platform.