Deepfake Defense

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.

ASTERIA KYCDeepfake Defense Result
CLEAR
Risk Score8 / 100
Face Swapnone
deepfake_resultclear
deepfake_risk_score8
texture_signalclear
temporal_signalconsistent
Signals Analyzed6
Session IDakyc_df_b7e4d1a0
6Signal Types
GAN+SwapAttack Models
FrameLevel Analysis
<1sResult Return
Detection Signals

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.

01

Synthetic Face Indicators

GAN-generated face artifacts, texture regularity, and feature blending anomalies assessed across the full face region.

02

Face Swap Detection

Boundary inconsistencies, skin tone discontinuities, and blending artifacts at the face edge detected in composite media.

03

Temporal Consistency

Frame-to-frame consistency of facial geometry, landmark positions, and micro-expressions assessed across video captures.

04

Texture Artifact Analysis

Pixel-level texture patterns inconsistent with natural skin. Compression artifacts and generative model smoothing signals detected.

05

Frequency Domain Analysis

DCT and FFT frequency signatures analyzed for patterns associated with generative model outputs and upscaling artifacts.

06

Injection Signal Detection

Virtual camera, screen recording, and media injection tool signals detected at the capture layer before biometric analysis.

Defense Outputs

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.

FieldDescriptionValues
deepfake_resultOverall deepfake assessmentclear / uncertain / detected
deepfake_risk_scoreSynthetic face risk score0–100
face_swap_signalFace swap detection outcomenone / detected / inconclusive
texture_signalTexture artifact assessmentclear / anomaly / failed
temporal_signalTemporal consistency checkconsistent / inconsistent / not_applicable
reason_codesStructured reason arrayarray of strings
Sample Output
{
  "deepfake_result": "clear",
  "deepfake_risk_score": 8,
  "face_swap_signal": "none",
  "texture_signal": "clear",
  "temporal_signal": "consistent",
  "reason_codes": ["DEEPFAKE_ALL_CLEAR"]
}
Risk States

Three deepfake risk states. One structured response.

Deepfake signals are probabilistic, not binary. Uncertain results route to human review rather than automatic rejection.

clear

Clear

All deepfake signals returned below the risk threshold. No synthetic face, face swap, or injection artifacts detected above the configured level.

uncertain

Uncertain

One or more signals returned ambiguous results. Session routed to human review with full signal breakdown and capture evidence attached.

detected

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.

ASTERIA KYC | Compliance-First Identity Verification Infrastructure