THANK YOU BASED APPLE!
>Apple Expands Bet on Cutting Edge Privacy Technology
>Differential privacy allows companies to analyze data without learning too much about users
>Differentially private algorithms blur the data being analyzed by adding a measurable amount of statistical noise. This could be done, for example, by swapping out one question (have you ever committed a violent crime?) with a question that has a statistically known response rate (were you born in February?). Someone trying to find links in the data would never be sure which question a particular person was asked. That lets researchers analyze sensitive data such as medical records without being able to tie the data back to specific people.
https://www.wsj.com/articles/apple-expands-bet-on-cutting-edge-privacy-technology-1499425201
>>61264966
>No. He may have simply flipped a coin, answering yes to the question.
But it is still statistically probable that it was the marijuana question. While it may not be evidence, it is enough to convince most people so in the end it basically doesn't matter.
The problem with the example is that it is clear they are asking personally identifying information along with the questionnaire questions. What point is there in asking if they own a blue car unless you are trying to match data to individuals?
Differential Privacy is clearly a farce. An excuse to try and get people to answer questionnaires more honestly by giving the impression their answers will have greater privacy while still being statistically probable to match the personal data.
Of course, this doesn't matter at all because in Apple's privacy policy it literally says they are allowed to do anything they want with any data you give them that isn't considered "personal information".
>allows companies to
>based
No.
>>61265311
/thread
>>61265311
This.
And it couldn't even be used as evidence before, so this measure is effectively useless
>>61265311
What if the 1 in 10 people flip a coin and answer yes applies to the car question too?
Stats is not even my forte, so I'm sure this is wrong, (not to mention I have no idea on notation) But
Odds you're asked if you own a blue car: 90%
Odds you'll get H if you flip coin: 50%
Odds you're that guy in blue: 1%
so (1-.9)*.5 + .9*.01 - .01*(1-.9) = 5.8%
Apple's the only one that care about privacy.
>>61265311
>But it is still statistically probable that it was the marijuana question
you're retarded. stya in school kid.
>>61265543
The problem would be introducing too much erroneous data to draw conclusions from the questionnaire.
In this case, if the point of the questionnaire was just to find out about how many people smoke marijuana then it would be fine, if you wanted to make a correlation between people smoking marijuana and owning blue cars then you're hurting your data.
So basically, yeah if you are only asking one real question then you could make all the questions Differential, but if you want to make conclusions based on all the data you get then you may as well just generate a random set.