Re:Homeland Security's 'Pre-Crime' Screening Will Never Work
by tptacek on Wednesday April 18, 2012 @12:51PM
(#3858886)
This is a good article marred by, I think, a poor story angle. FAST sounds like a terrible idea, but not because it's Orwellian.
Homeland Security is not primarily a law enforcement agency. Here are it's departments; departments with a principle law enforcement focus are starred:
Immigration Services
(+) ICE
(+) Customs
(+) TSA
Coast Guard
(+) Secret Service
FEMA
Homeland Security Advisory
National Infrastructure Advisory
Homeland Security Science/Technology
Critical Infrastructure Partnership Advisory
Interagency Coordinating Council on Emergency Preparedness
Task Force on New Americans
Domestic Nuclear Detection Office
Federal Law Enforcement Training Center
National Protection and Programs Directorate
(+) Federal Protective Service
National Communications System
Directorate for Science and Technology
Directorate for Management
Offices of Policy; Immigration Statistics; Health Affairs
Offices of Intelligence and Analysis; Operations Coordination
Office of the Secretary
National Cyber Security Center
(Yes, DHS is a monster, everyone knows that; we should definitely slay it).
People who have worked on the intrusion detection problem in computer systems are well familiar with the Base Rate Fallacy (what this article appears to be talking about when it mentions the "false positive fallacy"). To anyone who has actually monitored a deployed IDS, the base rate problem is a killer. Most of the events generated by systems like this are mathematically certain to be bogus.
So it's great to see an Atlantic article trying to tackle the numeracy issue here: DHS policymakers looking for research funding are bragging about a terrorist detection system with a 70% accuracy rate, and 70% sounds great when you leave out the base rate of terrorism.
On the other hand: of the LEO functions of DHS, most of them fall into one of two categories:
* Departments that already have total discretion to ignore civil liberties, like Customs: it's hard to imagine any new program making Customs more intrusive, because Customs operates under the border search rule (since at least 1976, Customs can set up any search program it wants to protect the integrity of the US border).
* Departments that operate under administrative search regimes, like TSA, where "behavioral monitoring" is if anything less Orwellian than the systems already in place.
DHS research projects do not automatically imply that local and state police will be deploying "behavioral" technology to "pre-crime" people. Again the theme: when you argue against something, you need to be both right and correct. "Pre-crime" is an inaccurate framing device.
Homeland Security is not primarily a law enforcement agency. Here are it's departments; departments with a principle law enforcement focus are starred:
(Yes, DHS is a monster, everyone knows that; we should definitely slay it).People who have worked on the intrusion detection problem in computer systems are well familiar with the Base Rate Fallacy (what this article appears to be talking about when it mentions the "false positive fallacy"). To anyone who has actually monitored a deployed IDS, the base rate problem is a killer. Most of the events generated by systems like this are mathematically certain to be bogus.
So it's great to see an Atlantic article trying to tackle the numeracy issue here: DHS policymakers looking for research funding are bragging about a terrorist detection system with a 70% accuracy rate, and 70% sounds great when you leave out the base rate of terrorism.
On the other hand: of the LEO functions of DHS, most of them fall into one of two categories:
* Departments that already have total discretion to ignore civil liberties, like Customs: it's hard to imagine any new program making Customs more intrusive, because Customs operates under the border search rule (since at least 1976, Customs can set up any search program it wants to protect the integrity of the US border).
* Departments that operate under administrative search regimes, like TSA, where "behavioral monitoring" is if anything less Orwellian than the systems already in place.
DHS research projects do not automatically imply that local and state police will be deploying "behavioral" technology to "pre-crime" people. Again the theme: when you argue against something, you need to be both right and correct. "Pre-crime" is an inaccurate framing device.