[lovelearning]

Although the medical data itself may be "anonymized", surely FB is in a position to associate that data with actual people, given that they know so much about a person's schedule, location, searches and private messages.

[sveme]

Deanonymization of medical data is actually pretty easy if you know a little about your target (age group, height, a set of pre-existing medical conditions limit the set of potential people considerably).

[downandout]

Just an aside...of the many things I have entered into Facebook over the years, I am 100% certain that I have never given them my height or pre-existing medical conditions.

[froindt]

There may be some other ways to link it up, at least with a degree of confidence. It really depends on what information is shared from the medical community and by the patient on Facebook.

A couple points of speculation:

* Facebook may possess a machine learning algorithm which can estimate weight from pictures. Getting within 5 pounds would eliminate most other people.

* Facebook could make photos of you and estimated weights into a time series, and pair up appointment dates with photos shared.

* Given enough photos with you and other people, they could probably estimate your height reasonably well. We know height distributions by age and race. If you're a Caucasian 21 year old female and consistently on average 10% shorter than the Caucasian males you're standing next to, that gives some info.

* Many people have willingly given the familial relationships to Facebook (tagging people as mom, dad, cousin, etc.) which will only help in being confident of race and the various risk factors which are higher in each race.

* Facebook knows your gender, which cuts out about half of the people. Such a basic fact would almost certainly be shared by the medical community.

* Facebook either has your birthday or could estimate it based on how you look. Again, being 98% confident of your age +/- 3 years cuts out most people.

All these fuzzy signals added up could lead to a reasonably confident matching up.

Anonymous data release is difficult. About 87% of people are uniquely identifiable by their date of birth, zip code, and gender.

https://news.ycombinator.com/item?id=2942967

[throwaway84742]

Not to mention location metadata on pics taken at or near doctors offices, and “private” messages in Messenger which they admitted they read.

[lovelearning]

The scary thing about data is that a lot of it can be inferred from other innocent data, such as photos. Height, can be inferred simply from the photos and videos you upload to social media.

[1]: https://graphics.ethz.ch/~edibra/Publications/HS-Nets%20-%20...

[fapjacks]

Yeah, it's the ole "encrypted in transit" (implying unencrypted at rest) corporate spin gimmick. The data is "anonymized"... By itself. But if you just so happen to have some kind of reference dataset or index, well...