> the cameras in retail are expected to only be used for review, when some event occurs. No one expects it to be used for data mining and behavioral analysis purposes.

Right. Except that this is also not true.

Source: "A crashed advertisement reveals the code of the facial recognition system used by a pizza shop in Oslo..." => https://twitter.com/gamblelee/status/862307447276544000?lang...

The word "reveals" is used because it is not expected by the average person. Otherwise it would be something like "confirms expectation".

No, the word "reveals" is used because the code is revealed. Well, it's more a "log file" than "code", but that's detail. The point is that the "code" (log file) is usually hidden from average people.

Also, the word "reveals" is used to make the headline more sensational, just like any other headline. It doesn't tell you anything about the average person.

How does this show that data mining isn't an abuse of expectation? The reason this is even an article in the first place is presumably because it conflicts directly with expected use.

There is certainly object detection available on the cameras and behavioral analysis applied to this information and POS data in aggregation software at the store and chain level for both loss prevention and marketing reasons. Maybe not used in all cases but I was surprised to learn these systems can be worthwhile in convenience stores, gas stations, and fast food restaurants, let alone larger stores like Target.

But you’re probably right that most people don’t think this is happening but only expect the stream is only flashing on a screen in front of a guard in real time and stored on a tape temporarily.

Worst case scenario is that the stored info is taken by someone more malicious and used for anything unintended beyond the original purpose.

The whole point of "less surveillance" is to limit the amount of information stored, because information stored IS the vulnerability.