Resolving Identity: The Importance of Who's Who

Despite the advances in technologies and solutions, the concern that remains central to the objective of providing security to the nation is the question of “who is who?” The federal homeland security laws directed toward enhancing the security of the United States — and the federal government’s confidence in the adequacy of such efforts — is, at the most basic level, grounded in the process of identity determination and the use of powerful databases to accurately determine individual identity to best evaluate threat and risk.

Much of government’s effort to protect the borders, minimize the risk to infrastructure and ensure the American public can travel safely, freely and efficiently — both domestically and abroad — is based on a complex system of computer-based data analyzers that search government records and provide information to help guide security personnel.

....Identity resolution, for the government and the private sector, is really a data-based intelligence process.

From the paper:

“Secure identification should begin in the United States. The federal government should set standards for the issuance of birth certificates

Indeed. In fact, it should start with the President, shouldn't it?

***

The paper has given the name "data resolution" to what I call "data triangulation." At some point I'll develop the idea a bit. In a nutshell, it means comparing as many databases as possible for common information and determining your identity based upon that. It no longer matters who you say you are. It matters who the algorithm determines you are.

“It matters who the algorithm determines you are.”

So true. I’ve worked in databases 26 years. It is amazing the amount of blind faith people have in the database content, even after repeatedly being burned.

The tendency is for a primary source to have a small number of errors. Each time it is merged with other databases, the rate or errors multiplies ... sometimes it seems exponentially.

For example, each creditor (bank_x, bank_y, store_z) might have 5% error rate in its database. Those are merged into a credit bureau which then has an error rate of 15% or more. A credit card marketing outfit then buys a list from the credit bureau and merges it with its own database to separate existing customers from new prospects for marketing. That ends up with a 25% error rate ... or worse.

One credit card shop bought over 50 million records a month from Equifax. After cleansing the 5% of 555-1212, eat-shit. etc it sent the phone numbers to separate vendors for telemarketing and direct mail marketing.

Even after cleansed, over 25% of the apparently valid phone numbers turned out to be invalid. The most common improbability in Equifax data was an implausible area code for the address given. For example, over 25% of Illinois zip codes did not have an area code of Illinois or an adjacent state. 1% is plausible. But not 25%.

Of course, the telemarketing vendor complained about the garbage they were being asked to work. They could not make any money off of so many bad phone numbers. The direct mail vendor didn’t complain. They got paid the same, whether they sent the junk to a valid or invalid address.

Go through a Credit Bureau database and you’ll be surprised at the number of blatant factual errors... situations that are physically impossible or a pattern in 25% of the data that is only plausible in 1% of the data.

Us consultants who bounce from shop to shop talk to each other. Generally speaking, the impression of consultants is that the IRS has the most error prone of any big database, even worse than that of Equifax.

Political algorithms are the other type. I interviewed for the CTA-ChicagoTransitAuthority for a workers comp system. They wanted someone to develop a system that contained a code to indicate which lawyer was handling the claim. Claims were paid based on who the lawyer was, not on the validity of the injury. I worked on the City of Chicago sytem that tracked the Building Department, Water Department, etc. The database kept track of who was “protected” from code enforcement and who was “targeted” for code enforcement based on political connections.

If Emmanuel, Axelrod, etal have anything to say about the National databases, they also will track who is politically “protected” and who is politically “targeted”.

that's a really interesting perspective. And there is the statistical, probabalistic algorithims -- that is, presumptions -- to consider. This is a problem when a probability, a guess, something fluid -- has absolute and tangible consequences, even life and death consequences.

I watched default rates that were supposed to have low correlations (and, in fact, under certain conditions, did have low correlation) suddenly become highly correlated (thereby immediately ruining the balance sheets of many major financial institutions) -- look at the very absolute and tangible consequences of those missed guesses.

Death to databases! And God help the fools that worship them. And, of course, the rest of us.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.