Questions Raised over Privacy, Racial Bias, and Accuracy

 

Today's topic is BBC's “Facebook to end use of facial recognition software”  https://www.bbc.com/news/business-59143323

It is good to see BBC now mention ‘accuracy’ in the report besides other concerns, but not good enough as yet; they still do not refer to the probabilistic nature of biometrics that inevitably brings the trade-off between false acceptance (false positive/match) and false rejection (false negative/non-match) 

Once they start to look at this trad-off issue, they could quickly notice that biometrics is dependent on a default password/pincode deployed in a security-lowering ‘two-entrance’ formation (as against security-enhancing ‘two-layer’ formation).

Then, they would soon be awake to the security-destructive effect of biometrics as summarised in these posts – 

“Step-by-Step Analysis of Why and How Biometrics Brings Down Security”

“Biometrics is to Password what Back door is to Front door”

If there is nothing particularly wrong in using two authenticators in a ‘two-entrance’ deployment for convenience’s sake, it is absolutely wrong and unethical to spread the false sense of security by misleading the citizens to believe that the security has been improved by biometrics.

 

False Acceptance Rates and False Rejection Rates



10

10%

10°




38 (Equa ror Rates)

procucts
(more accurate)






00 100 ar 100 10
FAR (Fale Accegtarce Rates)Worry about a backdoor? - FRR (False Rejection Rages) False Acceptance Rates and False Rejection Rates 10 10% 10° 38 (Equa ror Rates) procucts (more accurate) 00 100 ar 100 10 FAR (Fale Accegtarce Rates)">