Liveness Detection

The system issues a random, unpredictable stimulus and verifies the correct physiological response. This breaks replay attacks because a pre-recorded sample cannot adapt to a novel challenge.

• Randomized prompts: Blinking patterns, head turns, or specific speech phrases
• Temporal binding: Response must occur within a strict time window
• Non-deterministic: Each session generates a unique challenge set

Algorithms analyze surface properties and sub-pixel noise patterns to differentiate living tissue from synthetic materials. Genuine skin exhibits microscopic pores, sweat ducts, and non-uniform reflectance absent in silicone masks or paper prints.

• Local Binary Patterns (LBP): Captures fine-grained textural descriptors
• Specular reflection: Living skin has a characteristic subsurface scattering profile
• Moiré pattern detection: Identifies the aliasing artifacts of digital screen recapture

Detection of unconscious biological rhythms that cannot be suppressed or accurately forged. These signals provide a continuous liveness guarantee without requiring active user cooperation.

• Micro-saccades: Involuntary, high-frequency eye movements during fixation
• Pulse detection: Remote photoplethysmography (rPPG) extracts heart rate from subtle skin color variations
• Pupillary light reflex: Automatic constriction/dilation in response to illumination changes

Using structured light, time-of-flight sensors, or stereo vision to confirm the sample originates from a three-dimensional object with genuine topography. Flat photographs and screen replays are immediately rejected.

• Point cloud consistency: Validates continuous depth curvature of a real face
• Parallax analysis: Detects the absence of 3D geometry in planar spoofs
• NIR depth mapping: Near-infrared patterns are unaffected by ambient lighting

Examining the dynamic evolution of a signal over time to isolate the transient and steady-state characteristics of a live acquisition. Replayed signals lack the stochastic noise and subtle drift of a live sensor.

• Signal-to-noise ratio (SNR) profiling: Live captures have a characteristic noise floor
• Temporal derivative patterns: Frame-to-frame pixel variation follows natural biological motion
• Compression artifact detection: Identifies the block-boundary signatures of previously encoded video

Combining evidence from heterogeneous sensor streams to create a liveness score resilient to single-modality spoofing. An attacker must simultaneously defeat all sensing channels.

• Visible + NIR + thermal: Each spectrum reveals different material properties
• Audio-visual synchronization: Lip movement must precisely correlate with speech
• Inertial measurement: Gyroscope data detects the micro-movements of a handheld device versus a fixed mount

The standardized ISO/IEC 30107 framework for classifying and detecting spoofing attempts against biometric capture devices. PAD systems differentiate between bona fide presentations (live samples) and presentation attack instruments (PAIs) such as printed photos, silicone masks, or replayed voice recordings. Detection methodologies are categorized into artifact detection (analyzing physical properties like texture or liveness) and challenge-response (requiring a randomized action like blinking or head movement).

A network-based spoofing vector where a valid, previously captured signal or data stream is maliciously retransmitted to impersonate a legitimate source. In the context of RF fingerprinting, an attacker records a transmitter's waveform and rebroadcasts it to bypass physical-layer authentication. Liveness detection counters this by analyzing channel state information (CSI) , timestamps, or nonces to verify the signal is fresh and not a delayed duplicate.

A security paradigm that moves beyond one-time login gates to persistently verify identity throughout an active session. Unlike static liveness checks at enrollment, continuous authentication monitors behavioral biometrics (keystroke dynamics, gait) or physical-layer traits (RF fingerprint stability) in real-time. A session is terminated the moment the live presence signal degrades or deviates, providing zero-effort liveness without user interruption.

A hardware security primitive that derives a unique, unclonable identity from inherent manufacturing variations in silicon. PUFs act as a hardware root of trust, generating a fingerprint that cannot be copied even by the original manufacturer. When combined with liveness detection, a PUF provides a tamper-evident guarantee—any physical probing or alteration to extract the key destroys the very variations being measured, proving the silicon is live and intact.

The machine learning task of identifying inputs that fall outside the statistical distribution of the training data. In liveness detection, OOD detectors flag unknown spoofing artifacts that were not represented in the training set. Techniques include:

The critical security metric measuring the probability that a liveness detection system incorrectly classifies a spoofing artifact as a live, genuine sample. A FAR of 0.1% means one in a thousand attack presentations succeeds. In high-security applications like defense RF authentication, the FAR target is often driven to near-zero, accepting a corresponding increase in the False Rejection Rate (FRR) to maintain a strict security posture.