Interesting People mailing list archives
Re: The Economist article on data mining
From: David Farber <dave () farber net>
Date: Mon, 29 Sep 2008 20:36:41 -0400
Begin forwarded message: From: "Hasan Diwan" <hasan.diwan () gmail com> Date: September 29, 2008 8:27:04 PM EDT To: dave () farber net Subject: Re: [IP] The Economist article on data mining Dr. Farber,
Data-mining may be bad for national security as well as for civil liberties. The software is often modelled on the fraud-detection applications used by financial institutions. But terrorism is much rarer. So spotting conditions that may precede attacks is harder. Mike German, a former FBI agent who now advises the ACLU, says intelligence agencies too readily believe in the "snake oil" of total information awareness, which drains effort from moreuseful activities such as using informers and infiltrators.
Data-mining's effectiveness depends on the level of detail given to the algorithm that performs the function. Too much detail, and some "good" information is marked "bad". Too little, and some "bad" information is marked "good". Irrespective of what approach is taken, it still must be reviewed, in the end, by some arbitrer of good and bad. A sample approach is to mark people named "Hasan" as bad and all others as good. This would not be useful, because it would have let Osama bin Laden slip through, while subjecting me to search (as my name is "Hasan"). However, you could have a profile matching bin Laden exactly, down to his DNA fingerprint. This would similarly be bad, because it would let Timothy Mcveigh, a non-Arab terrorist, slip through. In conclusion, data mining is no panacea. Indeed, it doesn't even begin to address the problem. According to [1], "there is no easy way to identify those who become involved in terrorism" -- data mining is sifting through data, looking for certain characteristics. If there are no characteristics binding this lot together, it isn't the proper tool for the task. I do not know what the proper tool would be, but even if I (or any other individual) did, would the government listen? -- Cheers, Hasan Diwan <hasan.diwan () gmail com> 1. http://www.guardian.co.uk/uk/2008/aug/20/uksecurity.terrorism1 MI5 Report Challenges Views on Terrorism ------------------------------------------- Archives: https://www.listbox.com/member/archive/247/=now RSS Feed: https://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com
Current thread:
- The Economist article on data mining David Farber (Sep 29)
- <Possible follow-ups>
- Re: The Economist article on data mining David Farber (Sep 29)