Computer vision could be a supplemental solution to the issue of computer security among very young or functionally illiterate computer users.
Little kids and illiterate (and/or info-tech unskilled) adults pose a unique challenge to the world of computer security. Remembering, frequently rotating, and pseudo-randomizing one’s passwords is tough enough for computer-educated folks without the additional complexities of not understanding many words and likely not understanding the importance of online identity security. Even assuming that a determined user can memorize a string of characters and generally keep it secret, what recourse is available for lost passwords, and what are the odds of successfully detecting the slight changes in website wording that give away most phishing attempts?
However, digital cameras are getting quite cheap, built-in webcams are available on many laptops, and image-recognition software is getting quite sophisticated. All the components are in place. I imagine that after an initial configuration process (guided by voice or pictograms) that takes a couple of pictures of one’s face and hands (for fingerprints or just distinctive hand-creases), the system would generate a reasonable model of the primary user. Then, the next time someone goes to use the computer (or access administrator-level functions), a fresh photograph would be required.
In order to prevent malicious image-copying by fakers, the system could check for pictures that look a little too much the same. If we’re feeling extra fancy, we could throw in multi-snapshotting or movie analysis to confirm reasonable breathing patterns.
Similarly, the face-model developed could attempt to calculate user age, and use its log of identification pictures to project a growth path so that during a growth spurts the kids won’t find themselves locked out. Same thing for hair-cuts and bloody noses. Multiply redundant logged/verified body parts (hand creases, ear shapes) could help provide a back-up for instances involving major facial-alterations like black eyes or tooth-fairy incursions.
I’m looking at you, OLPC.
