Secure your identity flow with liveness detection
Liveness detection ensures a real, live person is present, protecting against fraud like deepfakes and spoofing attacks. It’s vital for secure identity verification in digital onboarding and authentication.
Advanced Tech Behind Seamless and Secure Verification
Our solution leverages deep neural networks to detect injection and presentation attacks. Real-time environment analysis and micromotion tracking enhance security against sophisticated threats.
Deep neural networks to detect injection and presentation attacks.
Original camera verification to detect video stream spoofing and virtual camera attacks
Real-time analysis of environmental conditions (lighting, blur, glare)
AI trained on thousands of real-world spoofing attempts, including 3D masks
Automatic detection and adjustment of face position relative to the camera
Adaptive lighting via SDK for low-light environments
Micromotion tracking and liveness signals (e.g., reflections, pulse, eye movement)
Intelligent user guidance in unfavorable capture conditions
False Acceptance Rate (FAR)*
Conversion rate among users 60+
+70%
better results than our top competitor
First Attempt Conversion Rate
97%
Use Cases
Enhance your security and streamline identity verification
Schedule your demo
Ready to enhance security without increasing friction?
Certified to prevent biometric fraud
What is Oz Liveness and how does it work?
Oz Liveness is a face liveness detection service that confirms a real, live person is in front of the camera, not a photo, screen replay, mask, or deepfake. During capture, Oz’s SDK records a short video and an on-device engine checks quality and authenticity in real time; then a server-side AI ensemble of around 50 models analyzes the footage and metadata for spoofing and injection signals. The default gesture, selfie, is a passive liveness and takes about 0.7 seconds to capture, so most users just look at the camera. Oz Liveness is certified to ISO/IEC 30107-3 Level 1, Level 2 and Level 3 by iBeta and BixeLab with 0% ACPER/FAR, and EU CEN/TS 18099 at 0% APCER/FAR across its Web, Android, and iOS SDKs.
What’s the difference between passive and active liveness detection, and which does Oz use?
Passive liveness works silently: the user just looks at the camera, while active liveness asks the user to blink, smile, or turn their head. Oz supports both but recommends passive liveness as the default: it converts better and, crucially, is the stronger defense today. Active gestures were meant to prove a live human through movement a photo couldn’t fake, but modern deepfake and face-reanimation tools now synthesize a convincing blink, smile, or head turn on demand, so demanding the gesture adds friction without real assurance and can create a false sense of security. Instead of asking the user to prove liveness through motion, Oz analyzes the video itself for the traces an attack leaves behind (screen and re-capture artifacts, deepfake blending distortions, and temporal inconsistencies across frames), defending against both presentation and injection attacks. Active gestures remain available where regulation requires them, but as a security control, relying on active movement is outdated.
Is passive liveness secure enough, or do I need active gestures?
Passive liveness is secure enough for most onboarding and authentication when it’s properly implemented, certified and backed by injection attack defense, which is the case with Oz Liveness (iBeta ISO/IEC 30107-3 Level 1, 2 and 3, 0% APCER; CEN/TS 18099 at 0% APCER for injection), but not all Liveness technologies can offer the same level of security as Oz’s does. Passive also converts better because there’s nothing for the user to do. Active gestures are worth adding for regulatory requirements that mandate a challenge, or when you want a visible step-up. With Oz you can mix them: passive selfie by default, active gestures where regulation requires it.
Does Oz Liveness require any special hardware or sensors?
No, Oz Liveness does not require any special hardware or dedicated sensors; it operates using standard device cameras found on smartphones, tablets, and webcams. This software-only approach keeps deployment simple and broadens accessibility, since businesses don’t need to invest in additional capture hardware to use certified liveness detection. It also means the same engine works consistently across the SDKs: iOS, Android, Flutter, and Web, without device-specific hardware dependencies.
How does Oz Liveness detect deepfakes and AI face swaps?
Oz Liveness detects deepfakes with a ~50-model AI ensemble that analyzes a sequence of video frames, not a single image, looking for blending seams, unnatural textures, motion and optical-flow inconsistencies, and re-capture artifacts that betray synthetic or replayed faces. It’s trained on millions of attack samples, including deepfakes generated with LoRA fine-tuning and LivePortrait, plus real attacks seen in production. It also uses out-of-distribution detection to flag novel attack patterns it hasn’t encountered before, which is what gives it resilience against new, zero-day deepfake tools.
How does Oz stop injection attacks such as virtual cameras and emulators?
Oz stops injection attacks with a multi-layer defense. First, at capture time, the SDK inspects the device environment for virtual cameras, emulators, rooting/jailbreaking, debuggers, and hooking frameworks, and the Web SDK adds a dedicated Camera Authenticity Layer, rejecting the session before any video is recorded. Second, video is recorded in a proprietary format that preserves injection artifacts, and a server-side deep neural network analyzes it for injection signatures. An attacker has to defeat the three layers independently, which is why BixeLab certified Oz at 0% APCER under EU CEN/TS 18099 across its Web, Android, and iOS SDKs.
What liveness certifications does Oz Liveness have?
Oz Liveness is certified to ISO/IEC 30107-3 Level 1, Level 2 and Level 3 by iBeta and BixeLab, both NIST-accredited labs, with 0% ACPER/FAR (no spoof accepted). For injection attacks, it passed BixeLab’s evaluation under EU CEN/TS 18099 at 0% APCER across Web, Android, and iOS. A key detail: the certifications cover the full architecture: the SDK capture layer and the server-side presentation attack detection together, rather than a single isolated component.
Which SDKs and platforms does Oz Liveness support?
Oz Liveness ships SDKs for iOS (Swift, minimum iOS 11), Android (Kotlin/Java), Flutter (iOS 13+ and Android SDK 21+), and Web (a client-side Web Plugin plus a server-side Web Adapter). The Web SDK requires HTTPS and a license tied to your specific domain(s). All SDKs run in portrait mode, include a deep on-device security stack (tamper, root/jailbreak, emulator and hook detection, SSL pinning), and support seven UI languages. They connect to the Oz API in Full or Instant (no-storage) modes.
How long does an Oz liveness check take, and what’s the user experience?
For the user, the default passive selfie check takes about 0.7 seconds to capture: they simply look at the camera, with no gesture required. On the backend, a server-side analysis takes roughly 2 seconds per check. The SDK guides the user in real time on lighting and face positioning and only records when conditions are met, which keeps false rejections low and avoids retries. If you choose active gestures, each gesture adds a 3–5 second window. Oz Forensics is known for the swift, seamless experience its liveness detection offers, which translates to frictionless experiences for final users and higher conversion rates.
What are FAR, FRR, APCER, and BPCER, and how does Oz perform?
These are the standard accuracy metrics for liveness systems. FAR (False Acceptance Rate) and APCER (Attack Presentation Classification Error Rate) measure how often a spoof is wrongly accepted, lower is better, and Oz recorded 0% ACPER/FAR in both ISO/IEC 30107-3 and CEN/TS 18099 certifications. FRR and BPCER measure how often a real user is wrongly rejected: too high and you lose genuine customers. Oz’s technology is known for having lower FRR/BPCER than competitors in real-world scenarios, achieving up to 3x better performance when compared to other solutions.
How does Oz Forensics mitigate bias in liveness detection?
Oz Forensics mitigates bias in liveness detection by training its ~50-model AI ensemble on tens of millions of real videos and attack samples drawn from diverse demographics, and by testing performance across ages, genders, and ethnicities rather than relying on a single benchmark population. This is combined with out-of-distribution detection, which flags unfamiliar patterns instead of forcing a pass/fail decision based on a narrow training set, helping keep false rejection rates low and consistent across different user groups. Fair, equitable performance is treated as a core design requirement, not an afterthought.
Is Oz Liveness suitable for high-security environments like government applications?
Yes, Oz Liveness is well suited to high-security environments, including government and public-sector applications, because of its certifications against presentation and injection attacks with 0% APCER, and its support for fully on-premise deployment. For institutions that require an audit trail and cannot send biometric data to a third-party cloud, Oz’s licensing model and on-premise architecture (via Docker or Kubernetes) provide the data-sovereignty guarantees that public-sector deployments typically require. Oz Forensics also processes over 100M liveness checks per month in operations that require high throughput, uptime and scalability.
How is Oz Liveness different from iProov, FaceTec or Zoloz?
Oz Forensics, iProov, FaceTec and Zoloz are all strong, independently certified liveness vendors, and the right choice depends on your priorities. Oz differentiates on four points: it is Presentation Attack Detection certified in accordance to ISO/IEC 30107-3 Levels 1, 2 and 3 and Injection Attack Detection certified in accordance to CEN/TS 18099, and the certifications cover the full SDK-plus-server architecture (not just one module); it offers genuine deployment flexibility: SaaS and on-premise, and is also available through marketplaces such as AWS and Hauweii; it uses OzCapsula, a tamper-evident container with an immutable event log that provides cryptographic proof of media integrity for audit and legal use; and it also leads with passive selfie liveness at ~0.7 seconds capture for minimal user friction, leading to up to 3x lower BCPER/FRR than these competitors.
