1 hr ago
Bharat ABIS Tackles Large-Scale Biometric Deduplication in India
Bharat ABIS is a computer system that checks whether identity records may belong to the same person.
It uses fingerprints, iris patterns and faces together.
Combining these details can make searches stronger than using only one biometric type.
The system turns the information into a small digital template.
It then searches a very large collection of identity records.
A scheduler divides the work among several servers so the search can run at national scale.
Researchers tested it using 220 million records chosen from the Aadhaar database.
Their results showed that the system could find duplicates while keeping mistaken matches relatively low.
The researchers said the open-source system could offer a less expensive option for some countries.
Researchers associated with UIDAI developed Bharat ABIS, an open-source system combining fingerprint, iris and facial biometrics.
The system was tested on 220 million identity records sampled from Aadhaar’s 1.55 billion-record database.
Bharat ABIS creates a combined biometric template of about 13.5KB and distributes searches across multiple servers.
In adult-record testing, it reported a 0.3% false negative rate at a 0.5% false positive rate.
Researchers said Bharat ABIS performed comparably or better than three commercial ABIS platforms evaluated in the Aadhaar ecosystem.
- Who
- Researchers associated with the Unique Identification Authority of India, along with researchers from IIIT Hyderabad and Michigan State University’s Anil Jain, developed and evaluated Bharat ABIS.
- What
- They developed and tested an open-source, multimodal Automated Biometric Identification System for large-scale biometric deduplication.
- Where
- The system was evaluated using records from Aadhaar, India’s identity database, including a 220-million-record test set.
- When
- The articles do not specify when the testing or research was conducted; the findings were published in the paper “Towards Billion-Scale Multi-modal Biometric Search.”
- Why
- It was designed to identify possible duplicate identity records at large scale while balancing accuracy, computing demands and cost.
Key facts
- System
- Bharat ABIS, an open-source Automated Biometric Identification System
- Biometric inputs
- Fingerprints, iris data and facial information
- Test scale
- 220 million records sampled from Aadhaar’s 1.55 billion-record database
- Combined template size
- About 13.5KB per person’s biometric representation
- Adult-record result
- 0.3% false negative identification rate at a 0.5% false positive identification rate
- Smaller-gallery result
- 0.05% false negative identification rate at a 0.1% false positive identification rate on a 20-million-record gallery
- Comparative evaluation
- Researchers reported performance comparable or better than three commercial ABIS platforms used in the Aadhaar ecosystem
Sources
Related news
Facial Recognition Systems Can Flag Matches, But Not Prove Identity
UIDAI Plans Major Increase in Daily Aadhaar Authentication Capacity
UIDAI launches SDK to simplify Aadhaar face authentication
Aadhaar Authentication Set to Nearly Triple as Digital Checks Expand
Report Finds 25 Jail-Held Accused Flagged at Jantar Mantar Protest
