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.

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. ®

Related stories

RSA 2005: complete coverage
IBM buys into spookware
US hit for $548m in fraud losses
Fed: Cyberterror fears missed real threat
Sprint security faulted in Vegas hacks