An open-source ABIS: freedom, not independence
This version is an AI-assisted translation. Read the original in French
Source: Biometric Update, “Open-source-based Bharat ABIS shows potential for national-scale deduplication” (Chris Burt, 23 September 2026).
As reported by Biometric Update, a team of researchers mostly affiliated with India’s Unique Identification Authority (UIDAI), joined by a researcher from IIIT Hyderabad and by Anil Jain of Michigan State University, has published Towards Billion-Scale Multi-modal Biometric Search. The paper presents Bharat ABIS, a multimodal automated biometric identification system (fingerprint, iris and face) built on an open-source architecture.
Tested on 220 million records sampled from Aadhaar, it reports a false negative identification rate of 0.3% at a false positive identification rate of 0.5% on adult records, and on a 20 million gallery an FPIR of 0.1% at an FNIR of 0.05%. Benchmarked against the three commercial ABIS used in Aadhaar, its performance is described as comparable or better. The authors position this approach as viable for the many countries that cannot afford a commercial ABIS and have populations of 20 million or fewer.
Why it matters: a fully open stack
Governments building foundational identity on open platforms still rely on a proprietary biometric core. The deduplication engine is often the one component that remains closed, licensed and hard to replace. A credible open ABIS would be a game changer: it would make a fully open-source identity stack possible, without a proprietary heart.
Good enough is often enough
Most ABIS on the market are already sufficient for deduplication. NIST rankings surface the fastest engines, but a system outside the top three would serve many use cases, particularly for countries of 20 million people or fewer. For most programmes, the selection criterion should be fitness for purpose, sustainability and total cost, not a podium position.
A familiar name at the table
Seeing Anil Jain involved is no surprise. He is an emblematic figure of the biometrics field, with brilliant work on biometrics for infants. ABIS designed with inclusion and sustainability in mind would be good news for governments.
Why open ABIS never reached industrial scale
Open alternatives already existed. They did not reach industrial scale for two reasons: the R&D investment needed to compete on accuracy and throughput, and the barrier of certification and independent evaluation.
Freedom, not independence
Open source does not mean independence. It means freedom. Operating an ABIS at national scale still requires international expertise: tuning, quality management, capacity planning, fraud handling. The question to watch is how the builders of this solution will address capacity building. AI could be a decisive factor here, hiding complexity and lowering setup and maintenance costs.
This is the lesson we drew from Ethiopia’s Fayda journey: seven years to build sovereign capacity with MOSIP. Read it here: End and start of a quest to sustainability.
Sources: Biometric Update, Open-source-based Bharat ABIS shows potential for national-scale deduplication (23 September 2026); paper on arXiv.
Want to comment, ask a question, or join the discussion? Continue the conversation on LinkedIn.