4 ms·
I would imagine they implemented it exactly like Google in that they're stored only on the OS and it needs your code to decrypt them. Anything else just doesn't
by BinaryIdiot 11y ago
I would imagine they implemented it exactly like Google in that they're stored only on the OS and it needs your code to decrypt them. Anything else just doesn't sit right.
- mtgx 11y agoUgh. First off they need to only store hashes of fingerprints, not the fingerprint images themselves. But I hope they are smart enough to do that by now. Second, it's not supposed to be "on the OS". It's supposed to be in TrustZone, the "secure world" separated from the OS. I hope that's what they did, but I worry they may have adopted a more "universal" solution that's a little higher level and less secure (even though TrustZone should be in virtually all ARM chips).
- thaumasiotes 11y ago> First off they need to only store hashes of fingerprints, not the fingerprint images themselves. But I hope they are smart enough to do that by now. That sounds pretty impressive. How would this be done? You're not going to get the same file by scanning the same physical finger twice. Fingerprints are recognized by comparing an image to a stored record of salient features, not to another image. But as far as I can see (and I have no further knowledge of the area), you kind of need the actual record, and the identifying information it contains, in order to see if an image matches it.
- dagw 11y agoThere are lots of algorithms that take an image and produce a so called perceptual hash that matches all images that 'look' the same. phash.org is a popular open source implementation of one such algorithm
- BinaryIdiot 11y agoFYI, good call with the trust zone (I would have hoped they used that): https://www.reddit.com/r/IAmA/comments/3mzrl9/hi_im_hiroshi_lockheimer_here_at_google_with_the/cvjkz9z https://www.reddit.com/r/IAmA/comments/3mzrl9/hi_im_hiroshi_...