Whatever comes to mind, and then the interesting directions that thought will take you on the web.

Saturday, February 20, 2010

Detective Work

I'm fascinated by the concept of information. How much data do you need to derive information. How can you inadvertently be providing information?

In Austin Powers, the following interaction takes place:

Austin Powers: Your boss, Number 2. I understand that cat's involved in big underground drills.
Alotta Fagina: How did you know?
Austin Powers: I didn't, baby. You just told me.

Alotta Fagina inadvertenly gives Austin Powers information!

I was watching the TV Show "Designed to Sell". They were featuring a house, but didn't say where it was located, no information.

But in 2 consecutive shots, I got some data:
  • On the real estate sign, I could make out the name of the realtor, "Sean", and the first part of a phone number - "818 999"
  • I could see the number of the house as "21716"
So, was this enough data to derive the location?

Using Google, I searched for 818 999 sean realtor

This gave me a realtor called Sean, in Woodland Hills with a number starting with 818 999

Next, I searched Google for 21716 "woodland hills"

This gave me a Zillow Page, that had an address 21716 Costanso St, Woodland Hills

This is definitely the house on "Designed to Sell". The image on the left is from the TV, the image on the right is from Google Street View. Interestingly, the Google Street View car was driving through the neighborhood at the time the house was being sold.

From blockshopper we can see that the owners (the people on the TV show) were Adam and Charity Brockman.

And from Zillow, are the new owners, the Cohens and Levins, regretting the $760K purhcase price?

Another example of inadvertently giving out information happened to Netflix, and it landed them in legal trouble. A few years ago, Netflix released ran a contest to create a recommendation engine that was 10% better than their existing one. The contestants used a file that contained millions of ratings that Netflix users had given to movies. The file contained just id numbers to identify people, but researchers were able to correlate the movie ratings with ratings that a person had made on another website (IMDB), identifying the person. Check this out.

No comments: