The government was committed to deploying a secure ID ecosystem with biometric identification as early as 2019. The primary goal was to enhance digital security to facilitate online transactions, financial inclusion, as well as better physical interactions with banking and financial institutions.
Given government’s vision of promoting online transactions through the National Digital ID platform, the existing ID system, which uses only fingerprints, needed an upgrade. The integration of facial recognition modality into the existing ABIS was a requirement. It led to the review of the minimal viable infrastructure setup of the existing solution that was prone to frequent downtimes.
Improvement of national identity management system system with face recognition
New hardware infrastructure for face modality, required capacity and traffic
The new Innovatrics ABIS with facial biometrics modality was integrated on site with the existing ID management system using various third-party legacy hardware.
To support the upgrade of the new national ID featuring face biometrics, Innovatrics has supplied bimodal finger and face ABIS. It features face photo ICAO compliance check for enrollment, face and fingerprint deduplication for identity fraud prevention. It also offers the possibility to update existing records with missing biometric data. A high-throughput and high availability system for 75 million records was deployed on 108 servers in an active-active setup. Moreover, 54 servers for the production environment and 54 servers for disaster & recovery environments are always available.
Efficient identity fraud prevention thanks to face and fingerprint biometric deduplication
54 servers for the production environment and 54 servers for disaster & recovery
Face photo enrollment with ICAO compliance check for high quality data collection
The largest and most complete biometric database in the country was launched in November 2019.
Featuring fingerprint and face recognition, new national civil register consisted of 65 million successfully deduplicated and migrated legacy records.
National ID project now counts 75 million records and boasts a high-throughput performance of 2000 searches in entire database per minute while maintaining high availability system architecture with 0 downtime of the service.
Clean and complete biometric national register
Instant identification and identity verification services also to 3rd party government
Active-active setup with production and disaster & recovery environments always available
Both legacy and new enrolments contain mixed records from children and adults, generating a significant amount of either false positives or false negatives.
During the 2nd year of operation, Innovatrics deployed a customized extraction algorithm that estimates the shape of an adult fingerprint from a child’s scan with impressive accuracy. As a result, the matching algorithm works with the same biometric template for all combinations of datasets: children, adults and mixed. It enables the use of existing Innovatrics ABIS installations while maintaining a high level of accuracy and low latency.
The national ID infrastructure requires 0 downtime deployment. Together with the main – HQ infrastructure, which was configured for high availability to make the cluster resistant to outages, we also deployed a separate, independent cluster of the same size and capacity.
The Disaster & Recovery infrastructure is in an independent deployment in a different geographical location and is synchronized with the HQ in real time and provides full functionality.
As a result, in case of a complete outage of the HQ infrastructure, the NID service keeps running without being affected. Moreover, the active full-featured Disaster & Recovery infrastructure also provides instant identification and identity verification services to 3rd party government and enterprise entities without affecting the performance of the HQ cluster used for national ID.
Legacy fingerprint and face data of poor quality made the requirement of zero FTE (failure to enrol) during legacy data migration very challenging.
We applied a special face recognition algorithm (designed for wild face images and CCTV) to recognize faces in poor quality legacy photos. The new enrolments are subjected to Innovatrics quality assessment during enrollment to guarantee the quality of the data for future operations.