2.4 Hardware-Assisted Liveness¶

Overview¶

Hardware-assisted liveness leverages specialized sensors beyond the standard RGB camera to provide additional modalities for distinguishing live faces from spoofs.

Structured Light (Apple TrueDepth / Android Equivalents)¶

How It Works¶

A projector emits a known pattern of infrared dots (30,000+ dots for Face ID) onto the face. An IR camera captures the deformed pattern. The deformation encodes 3D depth information.

graph TD
    A["IR Dot<br>Projector"] --> B["30,000+ IR dots<br>projected onto face"]
    B --> C["IR Camera<br>captures deformed<br>pattern"]
    C --> D["Triangulation<br>algorithm"]
    D --> E["Dense 3D<br>Depth Map"]
    E --> F{"3D face<br>geometry<br>consistent?"}
    F -->|"Yes"| G["✅ Live"]
    F -->|"No"| H["❌ Flat surface<br>detected"]

    style G fill:#27ae60,color:#fff
    style H fill:#e74c3c,color:#fff

Property	Details
Accuracy	Sub-millimeter depth precision
Range	25-50cm (optimal for selfie distance)
Speed	Real-time (30 FPS depth maps)
Attack resistance	Defeats all 2D attacks; most 3D attacks except high-quality silicone masks
Availability	iPhone X and later (TrueDepth); select Android flagships (limited)

Limitations for Banking¶

Device dependency: Only ~35% of global smartphone users have structured light sensors
Cannot be mandated: Inclusive banking requires supporting all device types
Use as enhancement: Use when available, fall back to RGB-only liveness when not

Time-of-Flight (ToF) Sensors¶

Emits modulated light and measures the phase shift of returned light to calculate distance per pixel.

Property	Details
Accuracy	1-5mm depth precision
Range	0.2-5 meters
Speed	30-60 FPS
Availability	Samsung Galaxy S20+, Huawei P30 Pro, select others
Limitation	Lower resolution than structured light; susceptible to ambient IR interference

Near-Infrared (NIR) Imaging¶

NIR cameras capture reflectance at 850-940nm wavelength, where materials behave very differently than in visible light.

Material	NIR Behavior	Visible Light Behavior
Live skin	Moderate absorption, subsurface scattering visible	Normal skin appearance
Paper	High reflectance, uniform	Shows printed texture
LCD screen	Very low emission (backlight doesn't emit NIR)	Displays image
OLED screen	Near-zero NIR emission	Displays image
Silicone	Different absorption pattern than skin	May look similar to skin
Latex	High reflectance, different from skin	May look similar to skin

NIR is Extremely Effective

NIR imaging can distinguish live skin from screens with near-perfect accuracy because screens emit virtually no NIR light. This is why ATM and kiosk-based systems prefer NIR cameras.

Recommendation for Banking¶

Deployment Context	Recommended Approach
Mobile app (diverse devices)	RGB-only liveness (passive + active), with hardware features as bonus signals when available
ATM / Kiosk	NIR camera + RGB (controlled hardware, can mandate sensor)
Branch tablet	Structured light if available (iPad Pro), else RGB liveness
Web browser	RGB only (no hardware sensor access)

Next: Comparison Matrix →