Fixing Aadhaar: Security developers' task is to trim chances of data breach

The task before a security developer is not only to reduce the probability of identity breach but to eliminate certain occurrences

I feel no joy when my prophecies about digital identity systems come true. This is because from a Popperian perspective these are low-risk prophecies. I had said that that all centralised identity databases will be breached in the future. That may or may not happen within my lifetime so I can go to my grave without worries about being proven wrong. Therefore, the task before a security developer is not only to reduce the probability but more importantly to eliminate the possibility of certain occurrences.

The blame for fragility in digital identity systems today can be partially laid on a World Bank document titled “Ten Principles on Identification for Sustainable Development” which has contributed to the harmonisation of approaches across jurisdictions. Principle three says, “Establishing a robust — unique, secure, and accurate — identity”. The keyword here is “a”. Like The Lord of the Rings, the World Bank wants “one digital ID to rule them all”. For Indians, this approach must be epistemologically repugnant as ours is a land which has recognised the multiplicity of truth since ancient times.

In “Identities Research Project: Final Report” funded by Omidyar Network and published by Caribou Digital — the number one finding is “people have always had, and managed, multiple personal identities”. And the fourth finding is “people select and combine identity elements for transactions during the course of everyday life”. As researchers they have employed indirect language, for layman the key takeaway is a single national ID for all persons and all purposes is an ahistorical and unworkable solution.

Revoke all Aadhaar numbers that have been compromised, breached, leaked, illegally published or inadvertently disclosed and regenerate new global identifiers. Photo: Reuters

There are many ways in which such an identity monoculture can be prevented. The traditional approach is followed in the US — you could have multiple documents that are accepted as valid ID. Or you could have multiple identity providers providing ID artifacts using an interoperable framework as they do in the UK. Another approach is tokenisation. The first time tokenisation was suggested in the Aadhaar context was in an academic paper published in August 2016 by Shweta Agrawal, Subhashis Banerjee and Subodh Sharma from IIT Delhi titled “Privacy and Security of Aadhaar: A Computer Science Perspective”. The paper in its fourth key recommendation says “cryptographically embed Aadhaar ID into Authentication User Agency (AUAs) and KYC User Agency (aka KUAs) — specific IDs making correlation impossible”. The paper considers several designs for such local identifier where — 1) no linking is possible, 2) only unidirectional linking is possible, and 3) bidirectional linking is possible referring to a similar scheme in the LSE identity report.

Though I had spoken about tokenisation as a fix for Aadhaar earlier, I wrote about it for the first time on the 31st of March, 2017, in The Hindu. The steps would be required are as follows. First, revoke all Aadhaar numbers that have been compromised, breached, leaked, illegally published or inadvertently disclosed and regenerate new global identifiers aka Aadhaar Numbers. Second, reduce the number of KYC transactions by eliminating all use cases that don’t result in corresponding transparency or security benefits. For example, most developed economies don’t have KYC for mobile phone connections. Three, the UIDAI should issue only tokens to those government entities and private sector service providers that absolutely must have KYC. When the NATGRID wants to combine subsets of 20 different databases for up to 12 different intelligence/law enforcement agencies they will have to approach the UIDAI with the token or Aadhaar number of the suspect. The UIDAI will then be able to release corresponding tokens and/or the Aadhaar number to the NATGRID. Implementing tokenisation introduces both technical and institutional checks and balances in our surveillance systems.

On 25th of July 2017, UIDAI published the first document providing implementation details for tokenisation wherein KUAs and AUAs were asked to generate the tokens. But this approach assumed that KYC user agencies could be trusted. This is because the digital identity solution for the nation as conceived by Aadhaar architects is based on the problem statement of digital identity within a firm. Within a firm all internal entities can be trusted. But in a nation state you cannot make this assumption. Airtel, a KUA, diverted 190 crores of LPG subsidy to more than 30 lakh payment bank accounts that were opened without informed consent. Axis Bank Limited, Suvidha Infoserve (a business correspondent) and eMudhra (an e-sign provider or AUA) have been accused of using replay attacks to perform unauthorised transactions. In November last year, the UIDAI indicated to the media that they were working on the next version of tokenisation — this time called dummy numbers or virtual numbers. This work needs to be accelerated to mitigate some of the risks in the current system.

The author is executive director, Centre for Internet and Society