interesting-people message

Subject: [IP] -- more on -- Millions more travelers could be flagged for intensive airport searches

From: Dave Farber <dave@farber.net>
To: ip@v2.listbox.com
Date: Tue, 26 Aug 2003 12:43:21 -0400

Date: Tue, 26 Aug 2003 08:32:16 -0700 (PDT)
From: Seth Grimes <grimes@altaplana.com>
Subject: Re: [IP] -- more on -- Millions more travelers could be flagged for
 intensive airport searches
X-X-Sender: ap2@whirlwind.he.net
To: BSteinhardt@aclu.org, Dave Farber <dave@farber.net>
Cc: jensen@cs.umass.edu, Seth Grimes <grimes@altaplana.com>

Reply for the list if you wish --

I attended a very interesting presentation yesterday, Information
Awareness: A Prospective Technical Assessment, at the SIGKDD conference.
The presenter was David Jensen; SIGKDD is the Special Interest Group on
Knowledge Discovery and Data Mining of the Association for Computing
Machinery.  Co-authors, all from the computer-science dept. at U Mass.,
were Matthew Rattigan and Hannah Blau.  Cribbing from their paper in the
conference proceedings --

Their thesis is that false positives can be vastly reduced by assuming
Relational data sources "providing connections between individual data
records" rather than Proposition, "in which each instance is characterized
by a set of simple propositions (e.g., age=32, gender=male)" where "each
individual is assumed to be statistically independent of any other."

Secondly, the authors believe that use of Ranking rather than Binary
classifiers (decision variables), where outputs are scores rather than
overly-simple true/false values, can also significantly reduce the
false-positive rate.

Lastly, they advocate use of multi-pass rather than single-pass inference.
[As an aside, I wonder that multi-pass wouldn't produce low additional
information due to the correlation inherent in "relational" (per the first
point) rather than statistically independent data.]

The authors believe that these three steps will rein in data requirements,
which can help address privacy concerns, in addition to reducing the rate
of false positives.

Jensen did acknowledge that information awareness is being pursued in an
adversarial environment with countermeasures per the identity-theft
example in Barry Steinhardt's e-mail.  He immediately got the drift of my
mention of Star Wars during Q&A:  After 20 years of research, defenses can
intercept in-coming missiles only in highly-controlled circumstances with
an expansive definition of the word "intercept."  He responded that it's
important for the technical community to contribute fair, expert views on
information awareness to forestall the government's taking the program
"black," i.e., secret.

Jensen did not address -- and wasn't asked to -- programs like CAPPS-II
that are being deployed despite technical design that is less than optimal
according to his improvement criteria.

                                        Seth

--
Seth Grimes   Alta Plana Corp, analytical computing & data management
              Intelligent Enterprise magazine (CMP), Contributing Editor
grimes@altaplana.com       http://altaplana.com    301-873-8225

On Tue, 26 Aug 2003, Dave Farber wrote:

>
> >Date: Tue, 26 Aug 2003 09:56:13 -0400
> >From: Barry Steinhardt <BSteinhardt@aclu.org>
> >Subject: Re: [IP] Millions more travelers could be flagged for intensive
> >  airport searches
> >X-Sender: Bsteinhardt@exch1.aclu.org
> >To: dave@farber.net
> >
> >
> >Dave,
> >
> >I thought IP readers my be interested in the fact sheet the ACLU put
> >together on CAPPS  II  for yesterday's press event The Five Problems With
> >CAPPS II: Why the Airline Passenger Profiling Proposal Should Be
> >Abandoned.http://www.aclu.org/SafeandFree/SafeandFree.cfm?ID=13356&c=206.
> >

> >Reason # 2 -- It won't work-- may be of particular interest to IPreaders.

> >
> >  Effectiveness: This System Will Not Make Us Any Safer
> >Why build CAPPS II if it won?t make us any safer?
> >
> >Even a known, wanted terrorist could sail right through this system simply
> >by committing identity theft (which as we all know is all too easy today)
> >and obtaining a false driver's license or passport (which is even easier).
> >For example, such a terrorist might present a driver?s license with their
> >own photograph, but the name, address, phone number and DOB of an innocent
> >person (data that for most Americans can be purchased online for about
> >$50). Nothing in the CAPPS II program would stop such a terrorist. This
> >system is like a Maginot line  the heavily fortified defensive frontier
> >constructed by the French before World War II, which was rendered useless
> >when Hitler?s army simply went around it.
> >And even a tiny error rate would create huge problems. Each year, 100
> >million Americans fly, many of us more than once. Total passenger
> >transactions each year have been estimated to be as high as one billion.
> >CAPPS II would check every one of those transactions. Even if we assume an
> >unrealistic accuracy rate of 99.9%, mistakes will be made on approximately
> >one million transactions, and 100,000 separate individuals. Those mistakes
> >will result in not only a lot of innocent people coming under suspicion 
> >or worse  but will make it extremely hard to find the handful of real
> >terrorists amid the ocean of false positives.
> >While these shortcomings are being ignored now, once this system is sold
> >to the American people, they will inevitably be used to justify demands
> >for an airtight, cradle-to-grave, biometric national identity and tracking
> >system that would change what it means to live in America (but in all
> >likelihood still fail to thwart terrorism).
> >
> >The full event can be viewed on CSPAN at http://www.cspan.org/.
> >
> >Barry Steinhardt
> >
> >
> >
> >Director Technology and Liberty Program
> >American Civil Liberties Union
> >125 Broad Street,NYC 10004
> >212 549 2508 (v) 212 549 2629 (fax)
> >BSteinhardt@aclu.org


-------------------------------------
To manage your subscription, go to
 http://v2.listbox.com/member/?listname=ip

Archives at: http://www.interesting-people.org/archives/interesting-people/