[Infowarrior] - Another insane UK security process
Richard Forno
rforno at infowarrior.org
Fri Jul 17 11:32:49 UTC 2009
False Positives and the Database State
http://www.antipope.org/charlie/blog-static/2009/07/false_positives.html
There is, in the UK (as elsewhere) a prevailing climate of paranoia
about adults interacting with children.
In an attempt to be seen to Do Something, in the wake of a
particularly gruesome multiple murder, the British government
established a new agency, the Independent Safeguarding Authority, "to
help prevent unsuitable people from working with children and
vulnerable adults." Working with the Criminal Records Bureau, the ISA
"will assess every person who wants to work or volunteer with
vulnerable people. Potential employees and volunteers will need to
apply to register with the ISA." For a fee of £64 you apply to the ISA
for a background check. They then certify that you're not an evil
paedophile and a threat to society, and issue you with a piece of
paper that says you're allowed to interact with children in a specific
role. Want multiple roles — driving kids to school in your taxi, and
teaching them karate in the evening? — get multiple certificates.
Authors need to get a certificate before they can visit schools to
deliver readings. MPs need a background check, it seems, before they
can visit schools. (Usually the employer is responsible for getting
the certificate; hilarity ensues when it transpires that MPs aren't
actually employed by Parliament ...)
As you can imagine, the authors are upset. As Philip Pullman puts it,
"It seems to be fuelled by the same combination of prurience, sexual
fear and cold political calculation," the author of the bestselling
His Dark Materials trilogy said today. "When you go into a school as
an author or an illustrator you talk to a class at a time or else to
the whole school. How on earth — how on earth — how in the world is
anybody going to rape or assault a child in those circumstances? It's
preposterous."
He's completely right, in my opinion. But the situation is worse than
he imagines. I'm not going to apply for a CRB check — ever. And not
because I'm a criminal. (My sum total of negative interaction with the
law over the past 44 years has amounted to two speeding tickets, most
recently six years ago.)
Nor am I outraged at the privacy thing. (I'm used to the idea that we
live in a panopticon.)
What I'm worried about is the problem of false positives.
Even the simplest of databases have been found to contain error rates
of 10%. (The HMRC database in this study contains merely first, second
and surname, title, sex, data of birth, address and National Insurance
number — nevertheless 10% of the records contain errors.) Other
agencies are even more prone to mistakes. For example: my wife
recently discovered that our GP's medical records showed her as having
been born outside the UK rather than in an NHS hospital in Manchester.
We don't know why that error's in the system, and we've got the birth
certificate and witnesses to prove that it is an error, but imagine
the fun that might ensue if the control freaks in Whitehall decided to
enforce record sharing between the NHS and the Immigration Agency ...!
(Hopefully they're not that stupid, but who can tell?)
The point is, if 10% of government database records contain an error,
than the probability of a sweep of databases coming up with an error
rises as you consult more sources. And there are a whole bundle of
wonderful ways for errors to show up. If your name and date of birth
are the same as someone with heavy criminal record, a CRB check could
label you as a bad guy. If your social security number is one digit
transposition away from $BAD_GUY, see above. If the previous owner of
your house was a child abuser, see above. If your street address is
one letter/digit away from a street address occupied by a criminal and
some bored clerk mis-typed it, you can end up being conflated with
somebody else. And the more sources the CRB checks, the higher the
probability of a false positive result — that is, of them obtaining a
positive result (subject is a criminal) when in fact the subject is a
negative.
This is not a hypothetical worry. As of last November, the CRB had
falsely identified more than 12,000 people as criminals, according to
the Home Office. (Raw parliamentary answer here.) These are the
disputes that were upheld, that is, ones where the falsely mis-
identified were able to convince the CRB that their record was
incorrect. These are false positives which have been conclusively
identified as such. While the identified false positive rate is around
0.1%, the true figure is certainly much higher: because there will be
a proportion of individuals identified as false positives who are in
the unfortunate position of lacking the documentation to prove their
innocence.
I expect the ISA will be returning many false positives, because
they're looking in multiple places for evidence of misbehaviour, and
the more places they look in, the more likely they are to stumble
across corrupt database records that are superficially incriminating.
The harder they look for evidence of misdeeds, the likelier they are
to find them (even if no such misdeeds took place).
I'm not going near that thing with a barge-pole. The nature of the
precautionary bureaucracy we're establishing in the UK is such that
flags raised by the ISA will almost inevitably be propagated elsewhere
through the police and social security system, sooner or later. I'm
probably as safe as ISA background check applicant can be, because
I've got a unique name, no criminal record (beyond the aforementioned
speeding tickets), and the previous owners of everywhere I've lived in
the past 20 years have been pillars of respectability. However, even
an 0.1% chance of being branded as Evil™ is too damn high, because the
personal cost if you fail an ISA check is potentially enormous going
forward. I assume that in the near future, failing an ISA check will
itself be something that people are required to disclose on job
applications — not to mention ending up in current police intelligence
databases. To put it in perspective, that 0.1% probability of being on
the receiving end of a false positive is of the same order as the risk
of being seriously injured in a road traffic accident at some time in
one's life.
So I won't be doing any readings in schools, or work with youth
groups, in the forseeable future. Sorry — but it's too dangerous.
More information about the Infowarrior
mailing list