The time a person spends on different smartphone apps is enough to identify them from a larger group in more than one in three cases, say researchers.
Researchers analyzed smartphone data from 780 people. They fed 4,680 days of app usage data into statistical models. Each of these days was paired with one of the 780 users so that the models learned people's daily app use patterns.
The researchers then tested whether models could identify an individual when provided with only a single day of smartphone activity that was anonymous (匿名的). The models, which were trained on only six days of app usage data per person, could identify the correct person from a day of anonymous data one third of the time.
That might not sound like much, but when the models predict who the data belonged to, it could also provide a list of the most to the least likely candidates. It was possible to view the top 10 most likely individuals that a specific day of data belonged to. Around 75% of the time, the correct user would be among the top 10 most likely candidates.
In practical terms, a law enforcement (执法机构) investigation seeking to identify a criminal's new phone with these models could reduce a candidate pool of approximately 1,000 phones to 10 phones, with a 25% risk of missing them.
Consequently, the researchers warn that software given access to a smartphone's standard activity logging could make a reasonable prediction about a user's identity even when they were logged-out of their account. An identification is possible without monitoring conversations or behaviors within apps themselves.
Therefore, it is important to acknowledge that app usage data alone, which is often collected by a smartphone automatically, can potentially reveal a person's identity. While providing new opportunities for law enforcement, it also poses risks to privacy if this type of data is misused.
