APPENDIX 6
Memorandum from Dr John Daugman, University
of Cambridge
INQUIRY ON USE OF SCIENTIFIC EVIDENCE IN
POLICY FORMATION AND ASSESSMENT
EXECUTIVE SUMMARY
Public discussion of scientific issues related
to biometrically-enabled ID cards has been of poor standard. This
is because public debate about the proposed biometric ID cards
has been dominated by a single document, the "LSE Report,"
which had no natural scientists amongst its putative or actual
authors. Persistent errors of fact are repeated both in that Report
and in the parallel press campaigns run by its organisers. This
Memorandum documents some of the misinformation and recommends
that in future the broad scientific basis for policy proposals
should be assembled in an on-line web resource containing balanced
and critical empirical documents. This may prevent future hijacking
of public discussion by scientifically misinformed assessments
that are spearheaded by activists under academic cover.
1. I thank the Science and Technology Committee
for inviting my comments about the way in which the Government
obtains and uses scientific advice, specifically in connection
with the proposals about identity cards. The aspect on which I
wish to comment is the way in which the public debate, and to
a significant extent the Parliamentary debate, on this issue has
been influenced by scientific misinformation from lobbyists opposed
to the proposals. In some cases (which I will document here) it
could even be called disinformation. Of relevance under the Terms
of Reference of the present Inquiry is whether such tactics have
influenced policy formation or assessment, and whether in its
public communications the Government has adequately challenged
the scientific misinformation.
2. Immediately prior to every Reading of
the ID cards bill in either House, a report ostensibly prepared
by senior academics at the London School of Economics was widely
disseminated. The putative LSE authors included no scientists.
Moreover the LSE Reports were spearheaded and apparently written
not by the LSE Professors whose names appear on them, but by Simon
Davies, who is Director of Privacy International, a political
lobbying organisation fiercely opposed to the concept of citizen
identification.
3. Although recent debate has shifted mainly
to questions of cost, a major focus earlier was the scientific
and technical feasibility of biometric identification of persons
across a national database. Both the LSE Report and a wider press
campaign by the same source to influential media (including The
Economist; New Scientist; and the broadsheets) asserted
repeatedly that biometric identification simply would not and
could not work. Arbitrary statistics about False Match rates were
fabricated from thin air and presented as scientific facts in
that media campaign, contradicting all available scientific evidence,
as I shall detail more fully in paragraph 9.
4. However ambiguous or contrived may be
the authorship of the LSE Report, the absence of any natural scientists
from amongst even its putative authors may explain the persistent
errors of scientific fact that appear within it. Many of these
arise from confusing the iris with the retina. (The iris lies
near the front of the eye, in front of the lens. The retina lies
at the very back of the eye.) These simple errors when assessing
the feasibility of the iris biometric, for example if the lens
of the eye becomes cloudy from cataract, occur equivalently both
in the "Interim" release of the LSE Report dated 23
March 2005 and in its final release dated 27 June 2005, and so
henceforth I shall refer to both releases collectively as "the
LSE Report."
5. Glaucoma, diabetes, cataracts, blindness,
and pregnancy were all incorrectly said to affect the iris pattern,
or its visibility: "People with glaucoma or cataracts may
not be reliably identified by iris recognition systems."
"People with diabetes . . . will not be able to use this
biometric method." In fact glaucoma affects the retina, not
the iris. Cataract clouds the lens, which lies behind the iris
and which therefore does not affect the visibility of the iris.
Diabetes may affect the retina, not the iris.
6. It is informative to trace the origins
and promulgation of so many basic misunderstandings. Invariably
in the biometrics debate the sequence is that statements which
began as speculation or simple errors in earlier reports or in
the press, become cited as established facts in later documents
without further investigation. This is the ubiquitous standard
of scientific evidence in the LSE Report. For example it is taken
for granted that blind persons, or those with visual disabilities,
must lack eyes or lack visible irises. (The blind former Home
Secretary David Blunkett successfully used an iris recognition
system.) When I pursued one such confusion about eyes with the
authors of a document submitted to the Commons Home Affairs Committee
by the British Computer Society, I learned that they further believed
that the iris "shatters at birth."
7. Most bizarrely, the LSE Report asserts
that "Pregnancy . . . can affect the recognition of irises;"
and that "Patterns in the eye may change over time because
of illness," and that "using the iris image for health
diagnostics" is a concern. This practice, and these beliefs,
are called Iridology. All published scientific tests of Iridology
(see bibliography at http://www.CL.cam.ac.uk/users/jgd1000/irido1ogy.htin1)
have dismissed it as medical fraud. Yet this belief in systemic
changes in iris patterns seems to be part of the basis for the
LSE high cost estimate for the ID cards scheme, as the Report
asserts that the biometrics would need to be re-enrolled frequently
for these reasons.
8. Besides scientific inaccuracies such
as those cited above, the influential LSE Report was extremely
selective in the data that it cited. It ignored completely the
very positive test data about large-scale biometric capabilities
reported for example by the US National Institute for Standards
and Technologies. In particular, it ignored eight published studies
conducted over the past decade about the accuracy of iris recognition,
each one finding no False Matches. Two of those reports were particularly
germane to the contemplated large-scale UK deployment, as they
showed that with reasonable thresholds it was possible to perform
two billion iris cross-comparisons without making any False Matches
(IBG ITIRT Report 2005); and indeed 200 billion iris cross-comparisons
were performed without encountering any False Matches (University
of Cambridge Technical Report UCAM-CL-TR-635, 2005). The origin
of such resilient performance is the mathematical principle of
binomial combinatorics embedded into the iris recognition algorithms,
a topic which again has eluded any public discussion. The scale
of this huge number of iris cross-comparisons (200 billion) without
making False Matches is not widely appreciated. It is larger than
the estimated number of stars in our galaxy; it is larger than
the estimated number of galaxies in the universe; and it is larger
than the estimated number of neurones in the human brain.
9. Yet in earlier phases of the campaign
against ID cards, several influential journals (including The
Economist and New Scientist) and press were told by
the organiser and author of the LSE Report that iris recognition
has a "False Match Rate of 1%;" that "for every
100 scans, there will be at least one False Match," and that
therefore in a nation of 60 million persons, "each person's
scan will match 600,000 other records in the database." (Simon
Davies, New Scientist 180, no 2422, page 13.) This statistic
was simply conjured out of thin air with no basis in fact, and
obviously it contradicted dramatically all of the above-mentioned
studies. Nonetheless it was published as a fact without further
investigation.
10. Every day today some seven billion iris
comparisons are performed in a national security deployment covering
all 27 air, land, and sea ports of entry into the United Arab
Emirates, comparing arriving passengers against a central database
of iris patterns. (About 9,000 daily arrivals are each compared
by real-time exhaustive search against an enrolled database of
800,000 IrisCodes, making 7.2 billion iris comparisons per day.)
According to the UAE Ministry of Interior, over the past 4.5 years
this system has caught some 50,000 persons trying to enter or
re-enter the UAE under false travel documents. If the putative
1% False Match rate were correct, then the daily volume of seven
billion iris comparisons would be producing 70 million False Matches
per day. If this were true, I should have thought someone would
have noticed.
11. Both the LSE Report, and the parallel
media campaigns arguing that biometric identification cannot work,
have been highly influential. The Leader of the Opposition, David
Cameron, stated on 15 January 2006 (BBC, Andrew Marr's Sunday
AM Programme) that he based his objection to the ID Card proposals
primarily on the LSE Report's conclusion that the system would
be unworkable. Commons MP and Home Affairs Select Committee member
the Rt Hon Bob Russell (Lib Dem) declared that iris cameras would
cause epileptic fits, and that trained medics would need to be
standing by at each one. New Scientist (180, no 2422, page
13) asserted that the iris is a kind of thermometer, changing
its pattern with temperature. The conclusion of the LSE Report
about the technical feasibility of biometrics (page 184) was that
"Implementing biometrics [in the UK] could bring the country
to a standstill."
12. Conclusion: The science of biometric
pattern recognition and its underlying mathematics has been not
only ignored but contradicted and overwritten by political campaigners
against the Government's ID cards proposals. In effect this important
part of the discussion has been hijacked, as have also been some
academics. Under these unusual circumstances, with public debate
so steered by a scientifically misinformed document and a parallel
press campaign, the public interest may have been better served
by a more robust presentation from Government of the scientific
basis and technical capabilities of biometrics. One mechanism
whereby this might be achieved in future would be to create and
maintain an on-line website resource containing a balanced and
critical collection of scientific papers and reports that inform
and address public policy proposals.
January 2006
|
