After detecting a face in an image or video, the engine runs an additional check for signs of possible spoofing. It converts the face’s features into a 512-dimensional embedding and uses 1:N vector matching to determine whether the person is the enrolled user.
ziiface / Face authentication
API · SDK · On-premises
Identity verification, built naturally into your service.
From face enrollment to the match decision and the result. ziiface connects the whole authentication flow.
From your service to the authentication result
Your service
ziiface authentication engine
Face detection
Detects faces in images and video
Liveness detection
Additional check for signs of possible spoofing
Feature extraction
512-dimensional face feature vector
1:N vector matching
New enrollments take effect without retraining
Results
- Identity verified · Access granted
- Suspected spoofing · Not enrolled: denied
- Authentication logs · Reports
Returns authentication results to your service’s operational flow.
KISA K-NBTC
Face recognition algorithm performance certification*
New enrollments take effect immediately
Embedding-based, no model retraining
Within 0.5 s on average*
On our standard configuration
Available on-premises
Built and run on your infrastructure
* KISA-BP-2025-009 is a performance certification for ZiPIDA Inc.’s face recognition algorithm. It is not a PAD (presentation attack detection) certification. The average response time of 0.5 seconds or less is measured on our standard configuration.
01 / Authentication engine
How face authentication connects
Compares against enrolled faces, then returns the results you need.
Face detection
Finds the face to authenticate in an image or video.
Liveness detection
Runs an additional check for signs of possible spoofing.
Feature extraction
Converts the face into a 512-dimensional feature vector.
1:N vector matching
Compares against enrolled data and returns the result and logs.
New members, enrolled instantly.
With an embedding-based design, new enrollments take effect without retraining the model. Authentication results and logs are connected to your service’s flow for approving or denying access, or to its identity verification flow.
Face feature vectors are biometric data, too. Every deployment is designed together with policies for storage, access rights, encryption, retention, and destruction. Anti-spoofing that combines rPPG-based passive liveness with active challenges is still in development.
Performance you can verify. A clearly defined scope.
KISA K-NBTC · Algorithm performance certification
99.99%*
TAR-TRR · KISA-BP-2025-009
On our standard configuration
Within 0.5 s*
Average response time
EMBEDDING-BASED ENROLLMENT
Instant enrollment.
New users added without model retraining
* KISA-BP-2025-009 is a performance certification for ZiPIDA Inc.’s face recognition algorithm. It is not a PAD (presentation attack detection) certification. The average response time of 0.5 seconds or less is measured on our standard configuration.
API · SDK · On-premises
Three ways to connect, chosen to suit your service and infrastructure.
02 / Migration
Switching from your existing authentication service
Keep your existing service flow.
BROJ switched its existing AWS Rekognition-based authentication flow to the ziiface API and now uses it for its gym access service. During migration, we work out with you whether to use bulk re-indexing or user re-enrollment, depending on whether the original face images are retained.
Two services. One authentication core.
Gym access without cards or phonesIn service
The AWS Rekognition-based authentication flow was switched to the ziiface API, which connects member face enrollment with on-site authentication. The BROJ service uses the approval or denial result in its access flow.
Identity verification for entering a telemedicine consultation roomPlanned
When patients enter a telemedicine consultation room, ziiface is scheduled to compare them against their enrolled faces. Face authentication will connect to the healthcare service while keeping its workflow intact.