Slashdot Mirror


User: Dataminer

Dataminer's activity in the archive.

Stories
0
Comments
1
First seen
Last seen
Profile
(view on slashdot.org)

Comments · 1

  1. Use existing technologies instead on CAPPS 2 Back to the Drawing Board · · Score: 1
    All of this just means that we need to identify terrorists BEFORE they get into airport, not wait until everyone checked in and then try to figure out if someone is a threat.

    The thing I am most concerned about is US government steadfast desire to build the huge comprehensive single system for identifying threats. This does not work. There are two problems about any "lets build a huge database to tackle XYZ":

    1) There are inherent assumptions built into any such system, which means it will be looking at the things in a particular way. If this way of thinking is identified, terrorists will adapt accordingly and slip through.

    2) Current state-of-the-art in data mining technologies is very good for data analysis, but bad for making final decisions. This is the last bit between collating and analyzing information and assigning color codes to passengers, which is trouble. Why force the technology into something it is not designed to? Lets instead provide collated information to security officer looking at a particular person for him/her to make the final decision

    Instead, US government should consider taking the best of commercial and semi-commercial systems used by different agencies and in different fields, apply them to the same data and try to cross-reference the results. Each one will go its own way and one might score. From the top of my head:

    There is a CopLink from Knowledge Computing Corp. Granted - it looks at existing criminal records in US, but it is a start.

    There is LAS NameHunter for name recognition (but CAPPS 2 is probably using it)

    There are piece of technology which are used in domain other then security, but could/should be easily adapted to do the job:

    There is "Anonymous Entity Resolution" from SRD, which is used by casinos to cross-reference the entry identification information.

    There is HyperDossier/QueryConstructor from Intellidos which are being used by pharmaucetical companies to sift through huge sets of drug activity data

    There is Infoglide's Bladeworks, which tracks similarities between different online identities (I understand eBay is using it to stop rampant scams in its auctions)

    There are dozens of other companies working in the field.