proveface #

A Flutter package for real-time face liveness detection.

Challenges users with randomised blink, smile, and head-turn checks powered by Google ML Kit, plus passive anti-spoofing to reject printed photos, screens, and video replay attacks. Camera permission is requested automatically — no setup needed in the consuming app.

Features #

• ✅ Blink detection (eyes-closed → eyes-open confirmation)
• ✅ Smile detection (geometry-verified, multi-frame)
• ✅ Head turn — left, right, up, down (with ghost-face guide overlay)
• ✅ Passive anti-spoofing via ML model
• ✅ Screen / video replay detection via pixel texture analysis
• ✅ Distance, centering, lighting, sunglasses, mask checks
• ✅ Camera permission handled internally
• ✅ Fully configurable challenges and timeouts
• ✅ Returns typed ProveFaceResult with base64 JPEG on success

Platform support #

Installation #

dependencies:
  proveface: ^1.0.2

flutter pub get

Setup #

Android #

In android/app/build.gradle:

android {
  defaultConfig {
    minSdkVersion 23
  }
}

iOS #

In ios/Runner/Info.plist:

<key>NSCameraUsageDescription</key>
<string>Used for face liveness verification.</string>

In ios/Podfile:

platform :ios, '12.0'

Usage #

import 'package:proveface/proveface.dart';

final result = await Navigator.push<ProveFaceResult>(
  context,
  MaterialPageRoute(
    builder: (_) => ProveFaceDetector(
      appBarColor: const Color(0xFF0D4582),
      config: const ProveFaceConfig(
        enableBlink:     true,
        enableSmile:     true,
        enableTurnLeft:  true,
        enableTurnRight: true,
        enableTurnUp:    false,
        enableTurnDown:  false,
      ),
    ),
  ),
);

if (result?.success == true) {
  final base64Image = result!.capturedImageBase64!;
  // upload to your server
}

Configuration #

ProveFaceConfig #

ProveFaceDetector widget #

Result #

class ProveFaceResult {
  final bool    success;
  final String? capturedImageBase64; // non-null on success
  final String? errorMessage;        // non-null on failure
}

Result scenarios #

How anti-spoofing works #

Three independent layers run simultaneously:

1. Active challenges — Randomised sequence a static photo or looped video cannot pass.
2. Passive ML model — Scores every 6th frame; rejects if rolling average falls below threshold.
3. Screen texture analysis — Pixel variance and banding checks flag phone screens or printed photos.

License #

MIT — see LICENSE