A few weeks ago I had the pleasure of meeting Jeff Jonas [LINK] during a vendor session with IBM. Jeff’s title is "IBM Distinguished Engineer and Chief Scientist, Entity Analytic Solutions, IBM Software Group" where he is "responsible for shaping the overall technical strategy of next generation identity analytics and the use of this new capability in the overall IBM technology strategy".

Jeff easily stood out from the blue suits and gave an energetic and engaging presentation on some of the Identity problems his analytics software has been tackling in Vegas.

Vegas is interested in identifying individuals who have been blacklisted from gaming. Such individuals tend to commit identity fraud to prevent being kicked out of the casino and that’s where Jeff’s technology comes in. It performs real-time analysis of data, tying disparate data sources together to see if there’s a link.

It only takes a few minutes in Vegas to win big so performance is critical. If a guy walks into a casino and begins playing, how do you know he’s legit? He might order a drink while playing and pay for it with a credit card in the name of "Charlie Parker". Jeff’s software will take this data in and link it to where Charlie lives. And anyone else who’s lived at that address. And maybe find that 2 years ago, someone very very bad lived there. Coincidence? Well a few more strikes like that against Charlie and the casino is now in a position to prevent a potential fraud.

As well as real-time analysis, the software can examine existing data sources and tie them together based on a degree of confidence to see if individual identities are in fact the same person.  This kind of analysis begins with everyone being unique. Charlie is not Robert for example. And Robert is not Bob. The software might find that Robert and Bob shared an address at some point. But they went to different high schools. Or that Charlie and Robert were born on the same day. That Charlie and Bob got the same exact grades in College. At some point the system reaches a stage where it’s gathered enough suspicious facts against these folks that it can assert they are actually the same person.

I’m paraphrasing the presentation I heard but hopefully not mis-representing the content.

This approaches the identity problem from a very different angle then I’m used to and I found it fascinating. I’m still pondering the possibilities and consequences.  This perspective, combined with Kim Cameron’s Laws I think form a fairly well rounded definition of the Identity Problem space.

Here’s an interview from 2006 where Jeff talks about how to compare data using hashes and still allow for some fuzziness in the compared data. [LINK]