Fighting computer crooks the Las Vegas way

[R.A. Hettinga]

<http://www.theregister.co.uk/2005/02/23/las_vegas_threat_detection/print.html>

The Register


 Biting the hand that feeds IT

The Register ; Software ; Applications ;


Fighting computer crooks the Las Vegas way
By John Leyden (john.leyden at theregister.co.uk)
Published Wednesday 23rd February 2005 17:38 GMT

RSA 2005 Computing techniques used to identify cheaters in Las Vegas are
being applied to wider computer security and fraud detection problems. SRD,
a Las Vegas software developer which was acquired by IBM last month, is
taking its identity resolution software from the gaming tables into
corporate boardrooms.

SRD's business intelligence software (now renamed DB2 Identity Resolution)
can draw out non-obvious relationships between information stored on a
variety of databases. Las Vegas gaming companies use the technology as a
way to identify customers and their associates worthy of investigation for
possible cheating. The software looks at information on employee records,
suppliers, in-house arrests and incidents and industry-published
professional counters and cheaters' lists to identify action items. For
example, the system can tell if a person who has been arrested for card
counting phones up a relative of a card dealer.

"The casino industry uses our technology to discover individuals who may
have the wrong intent before ferreting out problems with an investigative
team," explained Jeff Jonas, founder and chief scientist of SRD (now
renamed IBM's Entity Analytics Solutions).

SRD's technology minimises the resources needed for an investigation and
works best where a firm owns the data it is processing, according to Jonas.
"It's problematic sharing information between organisations, so we've
developed an anonymous identity resolution approach," he told a workshop
Detecting Asymetric and Insider Threats - Las Vegas style at last week's
RSA Conference in San Francisco.

ID managed and correlated in crypto form

To preserve anonymity, SRD has developed a technique to compare data in
cryptographic form, instead of real information. Raw data can be confusing
- for example, different spellings of Mohammed can be used on records of
the same person, while Social Security numbers can be inputted instead of
driving license IDs. But this can be overcome by pre-processing data to
create a finite number of hashes for comparison. Adding salt to this data
prevents statistical attack.

This approach enables banks in the process of merging, say, to share
anonymized data. Anonymous identity resolution can also help banks identify
possible money-laundering activity or to combat disparate security threats.
According to Jonas SRD software is tuned to minimise false positives,.
Jonas, who describes himself as a data modeller, draws a distinction
between data mining and SRD's "rules based expert system".

Data mining looks for patterns in data that suggest, for example,
individuals who share characteristics with a bank's most profitable
customers. Data mining is tuned to avoid false negatives; but in
applications like fraud detection and counter-terrorism addressed by SRD's
technology false positives are the problem.

"[SRD's technology] It's not probabilistic like data mining. A
probabilistic approach works well if all you are using is 27c stamps but
not if you're pointing guns at people," Jonas said.
-- 
-----------------
R. A. Hettinga <mailto: rah at ibuc.com>
The Internet Bearer Underwriting Corporation <http://www.ibuc.com/>
44 Farquhar Street, Boston, MA 02131 USA
"... however it may deserve respect for its usefulness and antiquity,
[predicting the end of the world] has not been found agreeable to
experience." -- Edward Gibbon, 'Decline and Fall of the Roman Empire'