As Data Collecting Grows, Privacy Erodes

"The way Mr. Rodriguez’s positive steroid test result became public followed a path increasingly common in the computer age: third-party data collection. We are typically told that personal information is anonymously tracked for one reason -- usually something abstract like making search results more accurate, recommending book titles or speeding traffic through the toll booths on the thruways. But it is then quickly converted into something traceable to an individual, and potentially life-changing."

That's troubling even for someone who is not an A-list celebrity like A-Rod.

Most of us recognize the value of having such detailed data, and also the consequences of it (like in this article). I would prefer not to have to "trust" the company storing the data, and I think leaks like this make it clear this can't be the approach in the future. So how can this problem be solved from a technology perspective?

I'm aware of k-anonymization, where you change the data to make it mathematically impossible to identify an individual data point more specifically than among k entries. So for example

  Age Weight Disease
  18 150 Cancer
  45 203 Diabetes
  37 197 Heart Disease

becomes

  Age Weight Disease
  * * Cancer
  [36-45] [195-210] Diabetes
  [36-45] [195-210] Heart Disease

for k=2, and it's now impossible to know for sure the disease of the 45-year-old, whereas you could deduce that information from the earlier records.

Another approach, used by some major search engines for ad-targeting, is to insert random noise into the data.

What other solutions are there? If you build software that captures sensitive data, how do you deal with it?

I remember a research talk where the speaker said that, empirically, there are 3 reasons consumers give up their privacy: convenience, (the perception of) security, and money.