Letters to the Editor
-
@Arne
Thanks for the clarification.
There is still a huge issue when the same laws are used for both physical tapping and for data mining, though. The latter puts all of the weight on the minimization procedures, since there is no question that the data got collected. In data mining or any data retrieval/recognition domain, "perfect filters" are impossible save for the case of deleting all the information that isn't supposed to be looked for before starting the searches. The minimization procedures then become responsible for dealing with false positives. You mentioned these procedures yesterday, and you've mentioned them before, but very little scrutiny in the debate is focussed on these, all of it is focussed on the warrants. So it is possible for the data mining that people fear to slip by while everybody worries about the other.
One more issue, and hopefully not to confuse: Admissible sets. The best way to picture this is to think of a CT scan. The scanning device makes a spiral (axial) path around the patient, taking a picture at intervals through the patient's body. These "slices" are then put together using what is called an inverse Radon transform. Theory says that a complete 3d picture of the patient can be obtained from a set of pictures taken, even though the results may not be visible in any one picture, as long as the path described forms an admissible set. This means essentially that snapshots have been taken from enough angles to combine them together and see everything.
Data mining works the same way (think of each aggregate statistical category as one snapshot in the CT scan): If a particular object ("target" in the intelligence parlance) is a member of enough categories, the fact that the data is anonymous and sufficient privacy has been maintained in each category separately is irrelevant, the object can be identified, and completely characterized. Such a set of categories is, like the path in CT scans, an admissible set. In the corporate world, arguments on mining the data of consumers raged throughout the 1990s and the EU passed strict laws on compiling databases of personal information -- the US did not. It is not very easy to undo damage from aggregate statistics after the fact, which is why the issue needs to be front and center in the debate. An admissible set of data would easily be enough to argue in court for a warrant. Nobody will notice that the fact that such sets exist has already deleted the 4th amendment privacies. The evidence has essentially already been collected and is used to obtain the warrant that should have been needed to collect the evidence.
Arne, you're absolutely right. This needs to be debated out in the open. And I currently don't trust the Congress to be able to debate it properly.
-
OY GEVALT!!!!!
Glenn Greenwald is as usual, 100% on target, I might add, DEPRESSINGLY SO with his sobering and terrifying look at the Bush-anus-kissing Democrats who roll over and wag their tails and tongues for CHIMPya-Hitler. They should be RIPPING OUT Bush's THROAT with their SLAVERING JAWS and FANGS, the same way those Dobermans RIPPED OUT Gregory Peck's Mengele's THROAT in THE BOYS FROM BRAZIL, but instead, SHOW a MEEK CAPITULATION!!!!! IT REALLY MAKES ME SICK TO MY STOMACH THAT SUCH YELLOW-BELLIED QUISLINGS STILL TOE THE LINE INSTEAD OF FIGHTING Bush EVERY SECOND and WINNING, FOR GOD'S SAKE!!!!! These QUISLING Democrats feel they might lose re-election and the privileges that come with controlling both houses of Congress as well as put a damper on their chances of gaining a PERMANENT MAJORITY in both houses of Congress in the next elections cycle if they are seen as being "soft on terrorism," an OVERUSED and IRRELEVANT MANTRA used by Bush/Cheney, Bushland Uber Allies and repugnicant-repubs to BESMIRCH Democrats and PAINT them as UNPATRIOTIC, when in FACT the repugnicant-repubs, especially the PACK of WORTHLESS JACKALS running for the repugnicant-repub side in 2008, i.e., Giuliani, Romney, McPAIN, Thompson, Paul, Huckabee, Tancredo, Hunter, Brownback ARE THE REAL UNPATRIOTIC, QUISLING, YELLOW-BELLIED, LYING HYPOCRITES!!!!! DO the RIGHT THING, Democrats, and DON'T WORRY ABOUT ANY UNFORSEEN CONSEQUENCES----THE PEOPLE OF AMERICA ARE SMARTER AND MORE UNDERSTANDING THAT EITHER YOU OR those repugnicant-repubs THINK!!!!! OY VEY, OY GEVALT!!!!!
-
@ Ondolette
Data mining works the same way (think of each aggregate statistical category as one snapshot in the CT scan): If a particular object ("target" in the intelligence parlance) is a member of enough categories, the fact that the data is anonymous and sufficient privacy has been maintained in each category separately is irrelevant, the object can be identified, and completely characterized....
I was trying to get that across. Even if single observations don't have the "party-identifying" information sufficient to trigger FISA (or telecomm law), combinations may be sufficient to extract this identifying information. I'd note that under domestic wiretap law, it's easier to get a "call data" court order (the "trap'n'trace" information) than it is to get the full Title III content warrant (the equivalent of a "search warrant"). And data mining/meta-analysis may not get you the "content" (in the traditional, not the FISA ยง 1801(n) sense) where arguably the Fourth Amendment may apply.
... Such a set of categories is, like the path in CT scans, an admissible set. In the corporate world, arguments on mining the data of consumers raged throughout the 1990s and the EU passed strict laws on compiling databases of personal information -- the US did not.
True. I think the gummint should start looking at the privacy implications of todays's digital world, and address this (not only in the "national security" realm) but elsewhere.
As for CT techniques, buddies of mine were working on "metal clip artifact removal" algorithms to reduce artifactual effects ("rays") due to the extreme electron density of metal surgical staples compared to the surrounding brain tissue in brain scans; the scans are not complete (albeit transformed) data sets and not noise-free; the signal from the metal clip overwhelms the signal from surrounding soft tissue and bleeds outward. Knowing that there's a metal clip there (or detecting such), tweaks can be made to the data sets to reduce these artifacts. When MRI imaging started to take off, I suggested that "metal clip artifact removal" there might take a substantially different, but less desirable, form -- primarily mechanical. ;-)
Cheers,