livelyness_detection 2.0.1

A single function package to detect weather the face visible in the camera is a live face or not.

livelyness_detection

Index

• What is Livelyness Detection?
• Platform Support
• Installation
• Usage
• Configuration
  • DetectionConfig
  • LivelynessStep & LivelynessStepItem
  • DetectionThreshold
  • OvalConfig
  • CameraConfig
  • StepCardStyle
  • InfoScreenConfig
• Privacy & Biometric Data
• Contributors

What is Livelyness Detection?

Liveness detection for face recognition in biometrics is a technique where an algorithm detects if the person in front of the camera is alive and real. The algorithm is able to recognize a live person from presentation attacks - where a bad actor uses someone else's physical characteristics or biometric data (known as "spoofs") for impersonation.

Spoofing attempts using printed photos, recordings, deep fake pictures, and 3D masks poses a serious threat. Facial Liveness Detection incorporates specialized features to identify biometric spoofing attacks. Even though face recognition can reliably answer "Is this the right person?", it cannot answer "Is this a live person?" This is where liveness detection plays a significant role in fraud detection and mitigation.

Platform Support

iOS	Android	MacOS	Web	Linux	Windows
✅	✅	❌	❌	❌	❌

Installation

Flutter Setup

Add livelyness_detection to your pubspec.yaml dependencies.

flutter pub add livelyness_detection

Native Setup

iOS

1. Open the ios/Runner/Info.plist file and add the camera usage description inside the <dict> tag:

<key>NSCameraUsageDescription</key>
<string>Used to verify you are a live person.</string>

The camera runs with audio disabled, so no NSMicrophoneUsageDescription is required. Use a purpose-specific string: App Review rejects vague copy.

2. Open the ios/Runner/Podfile and set the platform:

platform :ios, '14.0'

3. Set the deployment target in the Xcode project accordingly.

Android

1. Set minSdkVersion to 21 in android/app/build.gradle.
2. Add the camera permission to android/app/src/main/AndroidManifest.xml:

<uses-permission android:name="android.permission.CAMERA" />

Usage

final CapturedImage? result = await LivelynessDetection.detect(
  context: context,
  config: DetectionConfig(
    steps: [
      LivelynessStepItem(
        step: LivelynessStep.blink,
        title: "Blink",
        isCompleted: false,
      ),
      LivelynessStepItem(
        step: LivelynessStep.smile,
        title: "Smile",
        isCompleted: false,
      ),
    ],
    startWithInfoScreen: true,
  ),
);

The returned CapturedImage contains:

• imgPath - path to the captured JPEG
• didCaptureAutomatically - whether the capture was auto-triggered
• completedSteps - list of steps the user completed

Demo

Configuration

DetectionConfig

The main configuration object passed to LivelynessDetection.detect().

Parameter	Type	Default	Description
steps	List<LivelynessStepItem>	required	Ordered list of liveness challenges the user must complete.
thresholds	List<DetectionThreshold>?	null	Custom sensitivity thresholds per detection type. Uses built-in defaults when null.
startWithInfoScreen	bool	false	Show an instructional screen before detection begins.
maxSecToDetect	int	15	Timeout in seconds. Detection ends after this duration.
allowAfterMaxSec	bool	false	If true, returns the captured image even on timeout. If false, returns null.
showFacialVertices	bool	false	Render raw face mesh vertices on the camera preview (debug).
displayLines	bool	true	Draw face contour lines on the preview.
displayDots	bool	true	Draw face contour dots on the preview.
contourLineColor	Color	#EAEAEA	Color of the face contour lines.
contourDotColor	Color	#EAEAEA	Color of the face contour dots.
contourLineWidth	double	1.6	Width of contour lines.
contourDotRadius	double	2.0	Radius of contour dots.
dashValues	List<double>?	null	Dash pattern for contour lines. Solid line when null.
captureButtonColor	Color?	null	Color of the manual capture button.
overlayColor	Color	0xCC000000	Color of the area outside the face oval.
progressColor	Color	0xFF4CD964	Color of the progress ring around the oval.
stepCardStyle	StepCardStyle?	null	Styling for the step instruction card.
ovalConfig	OvalConfig?	null	Shape and size of the face oval guide.
cameraConfig	CameraConfig?	null	Camera resolution, lens, and FPS.
infoScreenConfig	InfoScreenConfig?	null	Content for the info screen (when startWithInfoScreen is true).
routeBuilder	Route Function(Widget)?	null	Custom route transition. Uses MaterialPageRoute when null.

Callbacks

Callback	Signature	Description
onStepCompleted	void Function(LivelynessStep)	Fired each time a step is completed.
onFaceDetected	void Function(FaceDetectionModel)	Fired on every frame where a face is detected.
onTimeout	void Function()	Fired when maxSecToDetect is reached.
onComplete	void Function(CapturedImage)	Fired when all steps complete and the image is captured.

LivelynessStep & LivelynessStepItem

LivelynessStep is an enum defining the available challenges:

Step	Description
blink	User must blink both eyes.
turnLeft	User must turn head to the left.
turnRight	User must turn head to the right.
smile	User must smile.

Wrap each step in a LivelynessStepItem:

LivelynessStepItem(
  step: LivelynessStep.blink,
  title: "Blink your eyes",  // displayed to the user
  isCompleted: false,         // always start as false
  detectionColor: Colors.green, // optional: color when this step completes
)

DetectionThreshold

Fine-tune sensitivity per detection type. Pass these in DetectionConfig.thresholds.

SmileDetectionThreshold

Parameter	Default	Range	Description
probability	0.75	0.0 - 1.0	Minimum smile probability to register. Lower = more sensitive.

BlinkDetectionThreshold

Parameter	Default	Range	Description
leftEyeProbability	0.25	0.0 - 1.0	Left eye open probability below which a blink is registered.
rightEyeProbability	0.25	0.0 - 1.0	Right eye open probability below which a blink is registered.

HeadTurnDetectionThreshold

Parameter	Default	Range	Description
rotationAngle	45.0	0 - 180	Minimum head rotation angle (degrees) to register a turn.

DetectionConfig(
  steps: [...],
  thresholds: [
    SmileDetectionThreshold(probability: 0.6),
    BlinkDetectionThreshold(
      leftEyeProbability: 0.3,
      rightEyeProbability: 0.3,
    ),
    HeadTurnDetectionThreshold(rotationAngle: 30.0),
  ],
)

OvalConfig

Controls the face guide oval overlay.

Parameter	Default	Description
widthFraction	0.72	Oval width as fraction of screen width.
aspectRatio	1.32	Height-to-width ratio of the oval.
verticalOffsetFraction	-0.06	Vertical offset (-1.0 to 1.0). Negative = up.
blurSigma	8.0	Blur intensity of the area outside the oval.
ringStrokeWidth	5.0	Width of the progress ring stroke.
ringTrackColor	0x33FFFFFF	Color of the unfilled ring track.
ringPadding	6.0	Padding between the oval and the ring.

CameraConfig

Parameter	Default	Description
resolution	ResolutionPreset.high	Camera resolution preset.
preferredLens	CameraLensDirection.front	Which camera to use.
maxDetectionFps	30	Max frames per second for face detection processing. Lower = less CPU.

StepCardStyle

Customize the instruction card shown during each step.

Parameter	Type	Description
backgroundColor	Color?	Card background color.
borderColor	Color?	Card border color.
borderRadius	BorderRadius?	Card corner radius.
textStyle	TextStyle?	Text style for the step title.
iconColor	Color?	Color of the step icon.

InfoScreenConfig

Shown when startWithInfoScreen: true. Lets you display preparation instructions before detection.

Parameter	Type	Description
title	String?	Screen title.
subtitle	String?	Subtitle / description.
guidanceItems	List<GuidanceItem>?	List of guidance tips (icon + text pairs).
buttonText	String?	CTA button text.
backgroundGradient	Gradient?	Background gradient.
buttonGradient	Gradient?	Button gradient.
lottieAsset	String?	Lottie animation asset path.
lottieFromPackage	bool?	Whether the Lottie asset is from a package.

Privacy & Biometric Data

A captured face is biometric data. Depending on your jurisdiction it may be regulated as a special category of personal data (GDPR Art. 9) or under biometric privacy laws such as BIPA.

What this package does and does not do:

• It performs no network I/O and keeps no persistent storage beyond a single unencrypted JPEG written to the app's temporary directory.
• It returns the file path in CapturedImage. You (the app) are the data controller: obtain explicit consent before launching detection.
• Delete the file as soon as you are done via await capturedImage.delete(). The package does not delete it for you.
• Do not log CapturedImage naively - toString() redacts the path, but imgPath/toJson() still expose it.
• completedSteps and auto-capture results are client-side liveness signals only; they are not tamper-proof. Perform server-side verification for any authentication decision.

Contributors