Aadhaar Duplicate Detection Update: Researchers associated with the Unique Identification Authority of India (UIDAI), along with academic experts, have developed an open-source biometric identification system called Bharat ABIS. The technology is designed to detect whether a person's biometric identity already exists in a large database, potentially helping prevent duplicate identity enrolments.

ABIS stands for Automated Biometric Identification System. Bharat ABIS combines multiple biometric characteristics—including fingerprints, iris information and facial data—to create a compact biometric template that can be searched against existing records.

The research, titled "Towards Billion-Scale Multi-modal Biometric Search," examines how such a system could perform when dealing with extremely large biometric databases.

The research team reportedly included six researchers associated with UIDAI, one researcher from IIIT Hyderabad and prominent biometrics researcher Anil Jain of Michigan State University.

However, Bharat ABIS should not be confused with a newly implemented Aadhaar rule or a tool that has already replaced UIDAI's existing biometric systems. The research itself indicates that further large-scale testing would be required before conclusions can be drawn about its suitability for a database as large as Aadhaar.

What Is Bharat ABIS?

Bharat ABIS is a multimodal Automated Biometric Identification System.

Instead of depending on only one biometric characteristic, the system can combine a person's fingerprints, iris and facial information.

According to the study cited in the report, these biometric signals can be represented through a compact template of around 13.5 KB.

When a new identity is submitted, this template can be compared against records already stored in the database.

The objective is straightforward: determine whether the person may already have an existing identity before allowing another one to be created.

This process is particularly important for large identity systems because manually comparing a new enrolment against millions—or potentially billions—of records would not be practical.

How Does Bharat ABIS Search Large Databases?

Searching an enormous biometric database requires significant computing power.

Bharat ABIS reportedly uses a scheduler that can distribute biometric-search tasks across different servers. This architecture is designed to allow searches to be performed efficiently even as the number of records increases.

The system can work using individual biometric modes such as fingerprints or iris information. However, combining fingerprints, iris and facial information can improve identification performance because the system has multiple biometric signals available for comparison.

This multimodal approach could be particularly useful in situations where one biometric characteristic is difficult to capture accurately.

Tested Using 220 Million Aadhaar Records

According to the report, researchers evaluated Bharat ABIS using around 220 million Aadhaar records.

The data was reportedly de-identified before being used in the system, meaning identifying information was removed for the research process.

Interestingly, the system's underlying models were reportedly trained using much smaller datasets containing approximately 100,000 to 200,000 entries.

The study then examined how the technology performed when applied to a much larger biometric-search environment.

This distinction is important because testing a technology on a large dataset does not necessarily mean it has been deployed operationally across the Aadhaar ecosystem.

What Are FPIR and FNIR?

Two important measures used to evaluate biometric identification systems are the False Positive Identification Rate (FPIR) and False Negative Identification Rate (FNIR).

FPIR measures how often a biometric system incorrectly identifies an entry as a potential duplicate when it should not.

According to the study, Bharat ABIS achieved an FPIR of approximately 0.1% in the cited test conditions. The report states that UIDAI's target for vendors is 0.5%.

FNIR, meanwhile, measures cases in which the system fails to identify an actual duplicate.

Bharat ABIS reportedly recorded an FNIR of 0.05%, compared with the cited target of 0.1%.

These figures, however, were reported for a 20-million-record database and should not automatically be assumed to represent performance at Aadhaar's full database scale.

How Does It Compare With Existing Biometric Tools?

Researchers reportedly compared Bharat ABIS with three existing tools used in the Aadhaar biometric ecosystem.

According to the study, its accuracy on a database of 20 million records was comparable to or better than the three systems included in the comparison.

The research also examined a fingerprint-recognition model called DeepPrint.

When DeepPrint was combined with a vendor's system, the study reportedly found that the number of authentication failures could potentially be reduced by around half.

Such improvements could be particularly useful for people whose fingerprints are difficult for biometric systems to read.

This may include elderly individuals and people engaged in physical or manual labour, whose fingerprint patterns can become difficult to capture clearly over time.

How Fast Can Bharat ABIS Search Records?

The study also provides an indication of the computing infrastructure required for high-speed biometric searches.

According to the reported results, Bharat ABIS could perform around 100 searches per second against a gallery containing 40 million records on a single server.

The server used for this setup reportedly had eight Nvidia H100 GPUs and 2TB of RAM.

These specifications underline the computational demands involved in searching very large biometric databases at high speed.

For a nationwide identity database, performance would need to remain reliable not just in controlled testing but also under large-scale operational conditions.

Why Is Duplicate Aadhaar Detection Important?

Preventing duplicate identity records is one of the key challenges faced by any large biometric identification programme.

According to information previously provided by the Ministry of Electronics and Information Technology (MeitY) in the Lok Sabha, around 600,000 duplicate Aadhaar cards had been identified and cancelled by May 2022.

Biometric deduplication systems are intended to detect situations in which the same person's biometric information may already be associated with an existing record.

A system capable of making such comparisons quickly and accurately could strengthen the deduplication process while reducing incorrect matches.

Bharat ABIS Could Have Uses Beyond India

One significant feature of Bharat ABIS is that it has been developed as an open-source system.

The researchers suggest that this could make the technology relevant for countries that want biometric identification infrastructure but may find proprietary commercial systems expensive or difficult to implement.

According to the report, researchers believe Bharat ABIS could potentially be useful for around 160 countries with populations below 20 million.

A country with a relatively smaller population could potentially use such a system without needing infrastructure at the enormous scale required by India's Aadhaar programme.

Open-source availability could also give governments and researchers greater ability to study, adapt and evaluate the underlying technology.

Is Bharat ABIS Already Being Used for Aadhaar?

This is where an important distinction needs to be made.

The research results do not mean that Bharat ABIS has already become Aadhaar's new biometric identification system.

The study itself reportedly acknowledges the need for additional large-scale testing before conclusions can be drawn about how effectively the system would operate on a database as large as India's Aadhaar ecosystem.

Performance results obtained using 20 million, 40 million or even hundreds of millions of records cannot automatically establish that the system will maintain identical speed and accuracy at billion-person scale.

Therefore, Bharat ABIS should currently be viewed as a significant research and open-source biometric technology development, rather than a confirmed replacement for UIDAI's existing Aadhaar infrastructure.

If future large-scale testing demonstrates consistent accuracy, speed and reliability, the technology could potentially play a role in strengthening biometric deduplication systems. For now, however, claims that duplicate Aadhaar cards will immediately disappear because of Bharat ABIS would be premature.