An open-source Bharat ABIS tool could assist address the major issue of duplicate Aadhaar


A biometric database is considered one of the most secure methods of verifying identity. However, when such a system operates at enormous scale, involving data belonging to more than 1.5 billion people as in the case of Aadhaar, it can face significant limitations. Researchers linked to the Unique Identification Authority of India (UIDAI) have developed Bharat ABIS, an open-source biometric verification system that appears to offer improvements in both speed and accuracy compared with systems currently in use.

The research team consisted of six UIDAI researchers, a researcher from IIIT Hyderabad, and noted biometrics expert Anil Jain from Michigan State University. Their study is titled “Towards Billion-Scale Multi-modal Biometric Search.”

The research looks at how Bharat ABIS, an open-source multi-modal biometric system, performs when dealing with massive databases while maintaining speed and accuracy. In particular, it examines how the system can add new identities to a large database without creating duplicate records or incorrectly flagging entries.

For context, the Ministry of Electronics and Information Technology (MeitY) told the Lok Sabha in 2022 that close to 6,00,000 duplicate Aadhaar cards had been detected and cancelled by May that year.

According to the researchers, ABIS was tested against a gallery containing 220 million Aadhaar records. The system itself was trained using a much smaller dataset containing between 1,00,000 and 2,00,000 entries. The data was de-identified before being provided to the system.

What is Bharat ABIS?

ABIS stands for Automated Biometric Identification System. Bharat ABIS brings together an individual’s fingerprints, iris and facial information and converts them into a single biometric template measuring 13.5kb.

The template is then compared with the database to determine whether the person already has an existing credential or ID. To enable the system to operate efficiently at large scale, ABIS includes a scheduler that distributes search tasks across multiple servers.

The system can also work with individual biometric identifiers, such as fingerprints or iris scans alone. However, its accuracy improves significantly when fingerprints, iris and facial information are used together.

How good is Bharat ABIS?

According to the study, Bharat ABIS records a False Positive Identification Rate (FPIR) of 0.1 per cent. FPIR measures instances where the system incorrectly identifies an entry as a duplicate. UIDAI has set a target rate of 0.5 per cent for vendors, meaning ABIS records a lower error rate.

The False Negative Identification Rate (FNIR) for Bharat ABIS is 0.05 per cent. FNIR measures cases in which the system fails to detect a duplicate, while the target set by UIDAI is 0.1 per cent. Both of these figures apply to a database containing 20 million records.

The researchers also compared Bharat ABIS with three existing tools used within the Aadhaar ecosystem. On the 20-million-record database, the study found that ABIS delivered “comparable or better” accuracy than all three systems.

The study says that Bharat ABIS’ DeepPrint fingerprint model, when used alongside vendor tools, can cut authentication failures by half. The largest improvements are seen among older people and manual labourers, whose fingerprints are more likely to be worn down. This could be particularly useful across millions of daily transactions.

The study also found that Bharat ABIS can perform 100 searches per second on a database containing 40 million records using a single server equipped with eight Nvidia H100 GPUs and 2TB of RAM. Researchers believe the system could potentially serve as a biometric identification solution in around 160 countries with populations below 20 million. According to them, some of these countries may not be able to afford commercially available biometric systems, while an open-source solution such as ABIS could provide a more accessible alternative.

However, the researchers acknowledge that Bharat ABIS still needs to undergo testing at much larger scales before its ability to handle databases the size of India’s Aadhaar system can be conclusively established.

In other words, while Bharat ABIS appears to represent a promising development for preventing duplicate identities such as Aadhaar IDs, it is still too early to determine whether the open-source platform can actually be deployed for systems of that scale.


 

buttons=(Accept !) days=(20)

Our website uses cookies to enhance your experience. Learn More
Accept !