I always question the effectiveness of obfuscating data using "unique identifiers" against a party determined to de-obfuscate the data. Aside from that, however, this is an encouraging read.
And WRT differential privacy, if enough data is captured and associated with a single identifier, then it's sufficient to get a pretty good idea what the user actually does. That's the point of differential privacy in the first place, since it indicates that a statistically meaningful amount of the data is valid.
For example, if my user id has five visits out of 1000 DP recorded visits to google.com, it's unlikely that I actually visited google.com. However, if there are 200 recorded visits, it can be safely said that yeah, I intentionally visit google.com.
Id't be better if they just didn't store an ID with the data.
> According to Wired, which reported on the story back in 2013, the data is shipped off to Apple's data farm for analysis, where it generates a random number that represents both the user and voice file. Six months later, Apple "disassociates" the user number from the clip, thereby deleting the number. The files are then stored for an additional 18 months, all for the sake of testing and product improvement.
> Siri and Dictation do not associate this information with your Apple ID, but rather with your device through a random identifier. Apple Watch uses the Siri identifier from your iPhone. You can reset that identifier at any time by turning Siri and Dictation off and back on, effectively restarting your relationship with Siri and Dictation.
So it looks like it keeps the same ID until you turn Siri off; if you turn Siri back on, you'll have a new ID.