"Rolling back privacy protections for airport security", eh?
What's that, Mr. Franklin?
http://www.washingtonpost.com/wp-dyn/articles/A5185-2002Jan31.html
[Note 1: you *can* racially profile using neural nets because the nets
learn categories --like culture-- from apparently benign inputs.
This is why HNC has successes doing predictionsfor industries
prohibited by law from using culture, race, etc.]
[Note 2: HNC has a DoD background; look up Hecht-Neilson]
[Note 3: Blacknet Partners Inc, a subsidiary of Reverse Panopticon
Corp, hereby requests infoon "PROS Revenue Management, which has
access to seating records of virtually every U.S. passenger", and
"Acxiom Corp., one of the world's largest data-marketing companies"]
By Robert O'Harrow Jr.Washington Post Staff
WriterFriday, February 1, 2002; Page A01
Federal aviation authorities and technology companies will soon begin
testing a vast air security screening system designed to instantly
pull together every passenger's travel history and living
arrangements, plus a wealth of other personal and demographic
information. The government's plan is to establish a computer network
linking every reservation system in the United States to private and
government databases. The network would use data-mining and predictive
software to profile passenger activity and intuit obscure clues about
potential threats, even before the scheduled day of flight.
<snip>
"This is not fantasy stuff," said Joseph Del Balzo, a former acting
administrator of the Federal Aviation Administration and a security
consultant working on one of the profiling projects. "This technology,
based on transaction analysis, behavior analysis, gives us a pretty
good idea of what's going on in a person's mind."
...
Government officials and companies also face questions about
privacy. In interviews, more than a dozen people working on two
parallel projects said they were taking pains to protect individual
privacy. They intend to limit the personal information shared with
airlines and security officials. But developers face restrictions on
how much information they can use. Industry officials have already
discussed with lawmakers the possible need to roll back some privacy
protections in the Fair Credit Reporting Act and Driver's Privacy
Protection Act to enable them to use more of the credit and
driver's-license data. Civil liberties activists said they fear the
system could be the beginnings of a surveillance infrastructure that
will erode existing privacy protections. When told about the system,
Barry Steinhardt, associate director of the American Civil Liberties
Union, said it would be "a massive complex system of surveillance."
"It really is a profound step for the government to be conducting
background checks on a large percentage of Americans. We've never done
that before," he said. "It's frightening." Some critics also worry
that law enforcement authorities will be tempted to use it for broader
aims, such as snaring deadbeat parents or profiling for drug couriers.
"If you can profile for terrorists, you can profile for other things,"
said Richard M. Smith, an independent computer security and privacy
specialist. "The computer technology is so cheap and getting so much
cheaper, you just have to be careful: Turn up the volume a little bit,
and we just use the air transportation system to catch everybody."...
The systems under development would include a thousand or more minute
details and computer-derived conclusions about a person's travel,
daily activity over time and whether he or she has coordinated
activity with other passengers, possibly on other flights, according
to the groups developing the systems. Two leading prototypes are
being developed. One group is led by HNC Software, a risk-detection
specialist that works for credit card issuers, telephone companies,
insurers and others. HNC is working with several companies, including
PROS Revenue Management, which has access to seating records of
virtually every U.S. passenger, and Acxiom Corp., one of the world's
largest data-marketing companies, which collects such information as
land records, car ownership, projected income, magazine subscriptions
and telephone numbers. Officials at both HNC and Accenture said they
take care with the personal information their systems collect and
parse. The HNC prototype, for instance, does not link a passenger's
personal information to a passenger's threat index. Officials also
pledged that there will be no racial profiling, in part because
ethnicity often has no bearing on potential risk. The HNC prototype
uses software known as neural networks, which can "learn" subtle
patterns and relationships by processing millions of records, to
predict when a particular transaction is likely to be fraudulent. The
company already uses neural networks software to accurately profile
the activity of millions of credit card owners, telephone callers and
people receiving insurance benefits to crack down on fraud. The
Accenture system also creates a threat index, using massive computing
power and relational database software. It examines travel data to
look for things such as routes involving odd destinations or flying
patterns. To search for threads linking individuals, the system will
sift huge amounts of travel records, real estate histories and "seven
layers" of passenger associates, according to Accenture partner Brett
Ogilvie. For instance, it would note if an individual lived at the
former address of someone considered high-risk. Theoretically, the
system could be calibrated to watch for people with links to
restaurants or other places thought to be favored by terrorist
cells. It might also note phone calls and match individuals against
government watch lists. A potential link to a threatening character or
region could boost a passenger's score, he said
....
Paul Werbos, a
senior National Science Foundation official and a neural networks
specialist, said such systems need to be used carefully. While there
is no doubt that profiling can improve security, Werbos said, "we have
to be very careful not to create punishments, disincentives, for being
different from average."
