I have been reading a book on privacy (System Error) with a story and recommendation of a demonstration site I thought others might find interesting. The issue considered is how information we think to be shareable can still be used to identify us.
Latanya Sweeney, then an MIT graduate student, and anyone else who wanted, had access to the record of Massachusetts hospital visits by all state employees. These data contained no information assumed to identify individuals – name, address, social security, etc. The governor of Massachusetts contended that the privacy of those associated with these records was secure. Sweeney did not agree. She purchased a list of all registered voters in Cambridge that contained the names of voters as well as their birth date and gender. There were six people in Cambridge with the same birthday as the governor, three of these were male, and one lived in the same zip code as Weld. Sweeney cross referenced the medical records by birthday, zip code, and gender, located the governor’s records and sent him a copy.
Sweeney contends that 87% of Americans can be identified based on zip code, birth date and gender. These are variables we commonly provide without concern.
Sweeney now works at Harvard’s Data Privacy Lab. The lab offers an interesting demonstration. You enter your birthdate, zip code, and gender. I tried the demonstration and found that I could be identified as unique. This may seem amazing, but work through the math. There are likely at least three gender categories and 365 days in the year. I am 73 years old. Even though there are 35,375 individuals in my zip code, the combination of my age, gender, and birthdate identifies one person.
These data do not offer up my name. I suppose this could be determined from voting records as Sweeney demonstrated. For other purposes, this is demonstration of just how easily we can be targeted with ads or information.
Reich, R., Sahami, M. & Weinstein, J. (2021). System error: Where big tech went wrong and how we can reboot. HarperCollins.
You must be logged in to post a comment.