Android app shows how revealing phone metadata can be
When it comes to phone metadata, the government and computer scientists have largely been on opposite sides of the privacy debate.
No doubt ah,ah, the defense of the NSA's program was it's just metadata.
And so, it seemed to us it would be worth the time to try to see whether that claim was true empirically.
Metadata includes details like the number dialed, the time of the call, and duration.
Just how sensitive is that information.
According to research done by Stanford PhD student Jonathan Mayer and his partner.
They created an Android app, MetaPhone, that asked users to volunteer their phone records in an effort to learn what can be uncovered from metadata.
More than 500 people signed up.
We began by identifying.
The organizations associated with the phone numbers in our dataset.
And we did that primarily using phone books provided by Yelp and by Google.
Totally public, totally easy to access.
With the help of Facebook phone directory feature, people searched services and Google.
More than 90% of the numbers were quickly identified.
We noted when a business was.
A firearms dealer.
We noted when a business was a health service provider.
Users also place calls to religious organizations, financial services and marijuana dispensary.
Also NSA surveillance is limited to two or three degrees of separation from an original suspect the MetaPhone App illustrates how the program can reach many people.
Through these numbers are very popular like.
Got T-Mobile's voice mail number.
Or FedEx, or Delta Airlines.
or, one of my favorites, telemarketers.
These, these spam phone calls, which call loads and loads of people.
and, the NSA's rules don't, prohibit the agency from following those hops.
Mayer and his partner plan to examine the data further, to see if other information can be found.
They'll focus on text messages next.
And they're working on something called a dating detector.
They'd like what do computer science people do on a weekend I guess.
We don't go on dates we just build systems for detecting people going on dates.
And so we built a machine learning system for identifying participants who were weren't in a romantic relationship.
It all raises very real privacy concerns about what happens when our phone records and public information reveal very personal affairs.
In Stanford, California, I'm Sumi Das, cnet.com for CBS news.