Bharat ABIS may help detect duplicate Aadhaar IDs through an open-source biometric identification system developed by researchers associated with UIDAI.
The system combines fingerprint, iris and facial biometrics to search very large identity databases.
The research team includes six UIDAI researchers, a researcher from IIIT Hyderabad and biometrics researcher Anil Jain of Michigan State University.
Their work is detailed in a research paper titled “Towards Billion-Scale Multi-modal Biometric Search.”
Bharat ABIS May Help Detect Duplicate Aadhaar IDs
Bharat ABIS is an Automated Biometric Identification System designed for large-scale biometric search and de-duplication.
It processes fingerprint, iris and face information and combines the resulting data into a single biometric template of about 13.5 KB.
The researchers evaluated the system using a demographically controlled gallery of 220 million de-identified identities randomly sampled from 1.55 billion Aadhaar records.
The study examined how effectively the system could perform biometric searches at very large scale.
How the New Biometric System Works
Bharat ABIS uses different processing stages for fingerprint, iris and facial biometrics, including quality assessment and presentation attack detection.
A similarity-search system then compares a new biometric record against records in the database.
The system also uses a scheduler to distribute search workloads across multiple servers. This approach is intended to make large-scale biometric searches manageable as the number of records increases.
Testing reported promising results. On a 20-million-record gallery, the study reported a false positive identification rate of 0.1% and a false negative identification rate of 0.05%.
The researchers also compared Bharat ABIS with three commercial biometric identification systems used in the Aadhaar ecosystem and reported comparable or better performance in the tested setup.
Bharat ABIS Could Have Wider Uses
The research also examined ways to improve biometric identification when fingerprints are difficult to capture.
The researchers said combining their approach with existing fingerprint technology could reduce authentication failures, including in cases involving worn fingerprints.
The open-source architecture could also make the technology relevant beyond India.
Researchers said countries that cannot afford proprietary biometric identification systems could potentially adapt an open-source approach for large-scale identity programmes.
The system has also demonstrated high search throughput in testing.
According to the research paper, Bharat ABIS achieved about 100 searches per second against a 40-million-record gallery using a single server equipped with eight Nvidia H100 GPUs and 2TB of RAM.
Further Testing Is Still Needed
The research represents a technology demonstration rather than confirmation that Bharat ABIS has already been deployed across the entire Aadhaar database.
The researchers note that searching and operating at billion-record scale presents additional technical challenges.
That distinction is important because performance on a research test environment does not automatically establish how a system will perform in a full national deployment.
Further testing, validation and operational assessment would be needed before drawing conclusions about nationwide use.
Bharat ABIS therefore marks a significant development in open-source biometric research, with its combination of multiple biometric modalities and large-scale search architecture designed to address the challenge of identity de-duplication.
