This article suggests a mask can be used to fool Face ID …
And not too long ago it was shown that just by changing a few pixels, image identification tools could be badly fooled
https://
Somewhere around there, real forensics starts how well does some particular method tell faces apart – how many possibilities are there really? How robust is it – how well does it avoid being fooled by unimportant differences? (What happens if we change a pixel?) How well has it been validated – not only only on employees of Apple, or their friends, one hopes? And how well does it adapt to future members of the pool of test subjects – does it have a reasonable life time?
All kinds of identification/individualization claims have been made for various methods and are still being made without any solid base individualization of audio recording, by hair characteristics, by bite marks, by gun bullet markings, as well as their composition … and even by eye witness identification. I shouldn't wonder if identification by keyboard typing characteristics belong there as well.
What disclaimers does face id identification need to make? Too many?
–
Useful reading, though occasionally a little dry (can be found on-line)
Saks, Michael J. and Koehler, Jonathan J., The Individualization Fallacy in Forensic Science Evidence (2008). Vanderbilt Law Review, Vol 61, No. 1, pp. 199-219, 2008. Available at SSRN https://
Section V.B. is probably the most important to read.
Great post! Disclaimers always occur when it comes to responsibility and liability. If users would keep one eye like the give one hand to their mobile and hold it like being life-threatening no unlock mechanisms would be needed. Personal life-balanced-economy related to all resources a person has would result in strongly focus on the important in life. In BigData the principle of 'Gold first' is crucial. This also is valid for information in general, today more digital than audible.
So being focused and concentrated on one or just a few life-processes would result in having the no such problems of femto-second winnings by facial authentication.
BTW Very interesting paper, absolutely worth studying -)
Which forensic suite or tool is the fastest and most reliable to unlock Face ID?