Liveness Detection

Determines whether a selfie is from a live person or a spoof (photo, screen, mask).

How It Works

Valydar analyses 6 anti-spoofing signals from the selfie:

1. Texture frequency — skin has natural texture patterns
2. Colour distribution — real skin has specific colour characteristics
3. Edge patterns — natural edges vs synthetic edges
4. Noise pattern — camera sensor noise is consistent
5. Compression artifacts — real photos have expected JPEG patterns
6. Brightness uniformity — natural lighting vs artificial sources

Result

{
  "type": "liveness",
  "result": "passed",
  "score": 0.87,
  "details": {
    "texture_score": 0.91,
    "colour_score": 0.88,
    "edge_score": 0.85,
    "noise_score": 0.82,
    "compression_score": 0.90,
    "brightness_score": 0.86
  }
}

Score Thresholds

Score	Meaning
> 0.7	Likely live person
0.4 - 0.7	Borderline — challenge-response recommended
< 0.4	Likely spoof attempt

Anti-Spoofing Coverage

Attack Type	Detected
Printed photo	✅
Screen replay	✅
Video replay	✅ (partial)
3D mask	⚠️ Limited
Deepfake	❌ (use dedicated deepfake detection)