Otukan identifies injected camera input before it reaches identity verification. From inside your app, it checks whether the frames reaching your selfie and document capture come from the device’s own camera. It detects virtual cameras, instrumentation attached to capture code, emulators streaming prepared video, modified app builds and compromised operating systems, and attaches that verdict to the images before your identity provider sees them. Checks run while the camera is open and are built separately for Android and iOS. Otukan does not judge whether a face is synthetic or whether a person is who they claim to be: liveness and identity decisions stay with your verification provider.
The difference between the two problems is covered on camera injection versus deepfake detection.
Threat boundary
An onboarding fraudster has two ways to beat a face check. One is to produce a face good enough to fool the model. The other is to skip the camera and hand the app whatever frames they like. Otukan owns the second route.
Injection is about where the frames come from
Injection attacks replace the live camera feed with media the attacker controls: a recorded video of the real account holder, a stolen selfie, a face swap rendered on a laptop. The media can be crude or perfect. What makes it an injection is the path, not the quality. That is why a capture-path check keeps its value as generative tools improve: a flawless synthetic face still has to enter the app through a route that is not the camera.
Deepfake classification answers a different question
Classifiers inspect pixels for signs of synthesis. Liveness checks ask the user to move, blink or turn. Both work on the content, and both assume the content arrived honestly. Otukan supplies that missing assumption.
| Attack | What reaches your app | Where it is caught |
|---|---|---|
| Printed photo or screen held up to the lens | Real camera frames of a fake subject | Liveness and presentation-attack checks at your identity provider |
| Synthetic face streamed through a virtual camera | Frames that never passed through the lens | Otukan capture-path checks |
| Recorded video fed through hooked capture code | Genuine-looking frames with a tampered origin | Otukan hooking and integrity checks |
| Emulator replaying a prepared file | Video from a device that does not exist | Otukan emulator detection |
Signal path
A capture-integrity verdict is only useful if it reaches the system that decides. Otukan follows the frames from sensor to server.
On the device
On Android, injection usually relies on a rooted device, a virtual camera app or a framework that rewrites camera calls. Otukan checks the capture source, the process and the system. On iOS, the camera feed is harder to replace, so attacks lean on jailbroken devices and tweaks that hook the capture framework; Otukan concentrates on those. Each platform gets its own checks. Otukan never assumes an Android result holds for iOS.
Alongside the capture
Checks run during the capture session itself, because a hook can attach after launch and a virtual camera can be selected at the last moment. The verdict is tied to the specific media produced in that session, so your backend can reject a verdict reused on another upload.
At your identity provider and backend
The images travel to your backend and on to your identity provider as they do today. The integrity verdict travels with them. Your server can refuse the submission, request a new capture or send the application to manual review before any document or face result is accepted. Hooking detection covers how instrumentation is found.
Where app integrity ends
Buyers get the best result when each party owns a clear question.
| Question | Owner |
|---|---|
| Did these frames come from this device’s physical camera, in an unmodified app? | Otukan |
| Is the subject a live person in front of the lens? | Your identity provider’s liveness check |
| Is the identity document authentic and unaltered? | Your identity provider’s document check |
| Is this person who they claim to be, and should the account open? | Your onboarding and compliance decision |
Otukan proves its part with the same discipline your onboarding needs. For each injection route, we reproduce the attack in an authorised lab and confirm the signal. We then run the same journey on a legitimate device and confirm that nothing fires, and repeat install, update, sign-in and the capture flow on each release. Tests with your identity provider are reviewed jointly, and results are recorded with devices, OS versions and the reviewer.
Camera injection detection sits within mobile fraud prevention. The wider onboarding journey is covered on mobile KYC fraud prevention. For claim photos in insurance apps, see insurance application security. For voice steps, voice and audio injection detection checks whether audio comes from the device microphone.