The fragility of fingers: UIDAI and debates on the locality of biometrics

If one theme that links the identity crisis of the Shah Rukh Khan film Duplicate to the system of positions of UIDAI is the threat of duplication for both the state and the good citizen, another is the potential destruction of the body itself, eroding or violating the body’s ability to correlate with the uniqueness of the biometric identity card. Remember in the film that the villainous duplicate of the hero burns both his thumb and that of his captive

[ photo source: http://www.findbiometrics.com/articles/i/9041/ ]

double in order to prevent the police from using the ID card to discover who is the original and who the nakli [false, duplicate] registrant of the card.

A similar theme emerges both within the massive literature critical of UIDAI as well as in the technical and entrepreneurial practices organized in relation to the UID. Take the following article published in October of 2011 on the news blog Firstpost [http://www.firstpost.com/politics/aadhar-indias-vanishing-fingerprints-put-unique-identity-in-question-115144.html].

As usual, I add headings [in italics] to edit and comment on the text.

India’s vanishing fingerprints put UID in question

Pallavi Polanki Oct 24, 2011

Fingerprints are temporal, plastic phenomena

A curious situation has come to light at the UID (unique identity) enrolment centres. Call it the phenomenon of vanishing fingerprints. You see, our unique fingerprints don’t necessarily last a lifetime and they can be damaged or destroyed and, in some cases, even non-existent. And that is not the best scenario for the first-of-its-kind project that endeavours to create a unique identity for India’s billion-plus population based on fingerprints and iris scans (or biometric data).

Investigative study in Delhi

To find out more, Firstpost visited five UID centres in the North West district, which incidentally has the highest enrolments (619,571 and counting) among Delhi’s nine districts since the show hit the road in February 2011, and one centre in North Delhi.

Biometric data can be rated for quality: good to poor

The officials at the Unique Identification Authority of India (UIDAI) will tell you that there is no overemphasising the importance of the quality of biometric data to the success of the super ambitious UID, now known as Aadhar project. If the quality of a person’s biometric data is poor, it automatically compromises the authentication of that data by him when he wants to access a service based on his UID.

Poor biometric data is a particularly Indian and socioeconomic situation

So what happens when the data – fingerprints, for instance – are inherently unreliable on account of various biological and socio-economic reasons, some of which are especially relevant to the Indian context.

Fingerprint quality is the “most important” guarantor of effective de-duplication, but we don’t know how it varies at very large scales of data collection and in the “Indian context”

The UIDAI’s Committee on Biometrics in a December 2009 was rather forthright about its reservations on the fingerprint reliability. Titled Biometrics Design Standards For UID Applications (page 4, para 4), it stated, “….two factors however, raise uncertainty about the accuracy that can be achieved through fingerprints. First, retaining efficacy while scaling the database size from fifty million to a billion has not been adequately analysed. Second, fingerprint quality, the most important variable for determining de-duplication accuracy, has not been studied in depth in the Indian context.”

This study suggests poor data — damaged or destroyed fingerprints — is routine in Delhi, especially among the old, among manual laborers, and among children

The findings on the ground were revelatory. Operators and technical experts at the UID enrolment centres confirmed to Firstpost that they routinely came across cases where fingerprints had been damaged/destroyed/underdeveloped. And such cases, they said, were more common among senior citizens, those involved in manual labour (who handle rough objects, for instance) and children (mostly below 10 years of age).

Case study 1: Batuli, a 72 year old woman, wants her “smart card”

Here is an example that Firstpost observed. Batuli, 72, arrived at around noon at the Basti Vikas Kendra, now also a UID centre, in Mangol Puri to get herself a ‘smart card’ everyone has been talking about. (The ‘Aadhar’ brand name hasn’t caught on in these parts of Delhi, with everyone insisting on calling it the ‘smart card’.) The helpful operator with the fancy gadgets helps Batuli to her seat. After her photo is taken, Batuli is asked to place four fingers — one hand at a time — on a green-lit device. Right hand, then left hand. She is, however, asked to repeat the exercise a second time for the left hand. The operator explains. “In some case, we have to scan the fingerprints and Iris multiple times. If it doesn’t pass the required quality percentage of 70 percent (the quality is indicated in percentage terms on the computer screen that is connected to the fingerprint machine), we repeat the exercise up to four times.” Batuli’s hands are then wiped using a cloth and placed back on the device. The exercise is repeated a fourth and final time. But still the same result. ‘Fail’, declares the reading on the operator’s computer. An operator at the next station, says, “In the case of senior citizens, Iris scans also sometime fail. It registers weakly when the retina is damaged.”

Case study 2: Children in Sultanpuri

The project coordinator of the UID enrolment centre working out of the Destitute Welfare Trust, an NGO in Sultanpuri, too, confirmed that he was aware of the problem of damaged fingerprints and the challenge it posed in getting good quality fingerprint data. “The hands of children are very soft and in some cases fingerprints are not yet fully developed. Also, children tend to have sweaty hands and this can interfere with the quality of fingerprints. Extremely dry hands also pose problems. We have to often, wipe the hands or provide lotion to improve the quality of fingerprints,” said a technical expert working with Strategic Outsourcing.

Registrars (private outsourced companies) find multiple failure is routine

While the number of attempts, as prescribed by the UIDAI, to get a stronger finger print when the result reads ‘fail’ is four, operators report that sometimes attempts go up to 10 to 15 (the machine picks up the strongest impression of the attempts made). They say they didn’t anticipate such a problem and it was only when they started enrolments in January that they were confronted with such a scenario. “Now, of course, everybody knows about it.”

Case Study 3: Model town slum: stonecutters

‘Everybody’ implies those who are directly involved in collecting biometric data. An operator from another private company Smartchip working in a JJ (slum) colony in the North Delhi district of Model town, revealed similar problems on being probed.  ”A fingerprint strength of 70 percent or more is pass. About 10 percent of the cases we get every day register the ‘fail’ reading. What can we do? In the case of stone cutters, for instance, the fingers are completely smooth, the fingerprints are completely wiped out.”

Implications: fingerprints are not enough for de-duplication, the iris is needed as well, “residents” need to be educated to know about (and presumably to become responsible for) their “good,” not-yet-degraded fingerprints

So what are the implications of poor quality fingerprints for the UID project? The UIDAI admitted that it could provide challenges to authentication. In an email response to Firstpost, Sujata Chaturvedi, UIDAI’s Deputy Director General for the Delhi region said, “It could provide some challenges in de-duplication although that has been mitigated to a large extent by the decision of the UIDAI to go in for iris an additional de-duplication factor… Also to be noted is the fact that normally not all fingers are equally de-graded. So UIDAI is evolving a protocol to inform the residents about their good quality fingerprints as part of the Aadhaar letter so that they are aware of the finger to use for authentication.”

Expert academic/NGO auditors challenge the accuracy of Indian data for effective de-duplication

But not everyone is convinced. Sunil Abraham is the executive director of Centre for Internet and Society (CIS), which has written seven open letters to the Standing Committee on the Finance Branch (before which is the National Identification Authority of India Bill, 2010) asking the committee to consider their research on the UID project and change aspects of the Bill and the project. The seventh letter sent last week, says Abraham, provides statistical analysis that demonstrates how the UID will never be able to create a unique database. “In CIS’s seventh open letter to the finance committee, statistical analysis reveals that UIDAI tender specification is 1,000 times less accurate than it should be to have a reasonable chance of building a truly unique database. This analysis depends on high quality biometrics. With poor biometric quality the problem of de-duplication is compounded.”

 UIDAI denies the large scale of poor data, but offers no contrastive evidence

When UIDAI was asked what the percentage of the population enrolled (all India and Delhi) had recorded below-standard fingerprint quality, no specific data was forthcoming. Chaturvedi wrote, “Currently the population with very poor quality fingerprints is a very small percentage. It must also be remembered that this population is scattered all over the country.” In India, even a small percentage translates to millions of people. “Small percentage could mean absolutely anything. Why can’t they be more specific? One percent in the Indian context is 12 million people,” said Abraham.

Solution: the iris (in the future); ‘one good fingerprint’ (now); existing [non-biometric] systems

On how the UIDAI was dealing with challenge of poor fingerprint quality, Chaturvedi said, “Even amongst the populace that has damaged/destroyed/underdeveloped fingerprints, chances are very high that they would have at least one good fingerprint that could be used for authentication. Second, UIDAI is also starting to actively develop iris authentication ecosystem. Fingerprint authentication and iris authentication could supplement each other to ensure a universal coverage.” She added that Aadhaar authentication will supplement and work in conjunction with existing authentication systems to strengthen the overall authentication rather than replace completely the existing authentication systems.

But if existing systems shore up the degradation of the body, what is the purpose of the degraded promise of Universal Identification?

That begs the question, as Abraham puts it, “If the UIDAI is not going to replace existing forms of authentication it is not clear why the government is spending all this money on unproven biometric technology.”

‘Proof of concept’ study as guarantor of UID’s promise? UIDAI cites rural implementation as what makes UID achievable

Chaturvedi, however, maintained that the initial PoC (proof of concept) study that was taken up in Andhra Pradesh, Karnataka and Jharkhand collected about 60,000 enrolments indicated that the biometric accuracy levels necessary for de-duplication of all residents of India are achievable. “The PoC results also indicated the time needed for capture of biometrics in typical rural conditions is small enough to support large scale enrolment. Over and above the initial PoC, the UIDAI has currently completed over 5 crore enrolments for which Aadhaars have been generated. This experience reinforces the initial PoC results that the de-duplication accuracy is sufficient and sustainable to enrol the rest of the population.”

—–

3 things to keep in mind:

1) On the degradation of the body/sign, its (failed) temporality as opposed to the Universal Time of the ID card itself (the card may be “smart,” as Batuli notes, but “residents” need to be educated about how to be responsible for their fingerprints)

2) On the pedagogy of parts: this is a theme I take up in my work on transplantation, the idea that the “common man” [sic] needs to be educated toward a correct knowledge of and responsibility toward his or her body as a series of parts

3) On the rural “proof” for the “concept” of Universal ID: I don’t know much here yet. But in an urbanizing country, and one in which the theme of the duplicate is often linked to the journey to the city, how both technically and ideologically to assess such a claim seems important.