I have been involved in facial recognition (FR) with law enforcement authorities for the last 10 years. In my opinion for sharing purpose, there are 2 key considerations for mass surveillance which looks at Technology + Human + Process.
The process & procedure for handling the images that are being collected which can be separated into people not on any watchlist and people on the watchlist. The ability to track people of interest extends before the days of FR where FR in this instance is only an enabler. People is the weakest link here and those handling these systems must be validated and govern with strict security procedure.
Images collected are just images that do not link to any individual unless the system is being fed with face images that contain personal information. If not, the operator will only know them as a matched or a non-matched. Handling of information, in this case, can be separated to prevent the ability to hold onto too much information by an individual. In this case, the history of cameras that captured this individual against other personal information.
Eventually, the weakness of FR lies in the need for face and the challenge right from the start is to ensure that the camera collects those faces (which also deal with human factors) and the platform can process them.
Thank you.
RE: Should Large Scale Facial Recognition Be Allowed?