Face Compare

What does this system do?

Face Compare is an identity verification system that performs two checks:

1. Liveness Detection — Confirms the person in front of the camera is a real, live human (not a printed photo, screen replay, or mask).

2. Face Comparison — Compares the live person's face against a reference photo to verify they are the same individual.

Together, these checks answer: "Is this a real person, and are they who they claim to be?"

How does the Liveness Check work?

The liveness check uses Amazon Rekognition Face Liveness, a cloud-based AI service from AWS.

When a user initiates the check:

An oval guide appears on screen for the user to position their face
The system runs a challenge: coloured light sequences projected onto the user's face, or guided head movements
The AI analyses the video stream in real-time, looking for depth cues, light reflection patterns, and natural movement
A liveness confidence score (0–100%) is returned

A score above 80% indicates the person is highly likely to be real. The system detects printed photo attacks, screen replays, 3D mask attacks, and pre-recorded video attacks.

How does the Face Comparison work?

After liveness is confirmed, the system compares faces using Amazon Rekognition CompareFaces:

A high-quality reference image is automatically captured during the liveness check
This is compared against the uploaded reference photo (ID card, passport, etc.)
The AI returns a similarity score (0–100%)
A score above 80% is considered a match

The comparison also provides face detection confidence, head pose, image quality metrics, detected emotions, and smile detection.

What technology is used?

Component	Technology
Liveness Detection	AWS Rekognition Face Liveness
Face Comparison	AWS Rekognition CompareFaces
Face Analysis	AWS Rekognition DetectFaces
Server	Go (Golang)
Database	MySQL
Video Processing	FFmpeg
Client Liveness UI	AWS Amplify SDK (React)

What does each verification cost?

Each identity verification involves a liveness check and a face comparison:

API Call	Purpose	Cost (AUD)
Face Liveness	Verifies real person	$0.023
CompareFaces	Matches faces	$0.002
Total	Per verification	~$0.025

Based on approximate exchange rate of 1.55 AUD/USD. Actual charges billed in USD by AWS.

Monthly cost estimates

Verifications/month	USD	AUD (approx)
100	$1.60	~$2.50
1,000	$16	~$25
10,000	$160	~$250
50,000	$800	~$1,240
100,000	$1,600	~$2,480

Volume discounts apply above 500,000 checks/month. Free tier: 1,000 CompareFaces calls/month free for the first 12 months.

How accurate is it?

Liveness Accuracy

>98%

False Accept Rate

<2%

Match Threshold

80%

Two challenge types are available:

Movement + Light — Higher accuracy, uses coloured lights (~6–8 seconds)
Movement Only — Faster (~3 seconds), no lights, slightly lower accuracy

What data is stored?

Comparison results (match/no match, confidence scores)
Reference image filenames and timestamps
No biometric data is permanently stored on the server
AWS processes images in-memory and does not retain them after the API call completes

What does the user need?

A device with a front-facing camera (phone, tablet, or laptop)
A modern web browser (Chrome, Safari, Firefox, Edge)
Minimum camera: 480x640px at 15fps
Adequate lighting

Security & Compliance

All data transmitted over HTTPS/TLS
AWS credentials are never exposed to the client
Temporary credentials with 15-minute expiry used for liveness
Rate limiting prevents abuse
AWS Rekognition is SOC 1/2/3, ISO 27001, and HIPAA eligible