3 ms·

Actually, doing away with the signature seems a bit odd, given all the machine learning flying around. An electronic signature that records the speed strokes in

Actually, doing away with the signature seems a bit odd, given all the machine learning flying around. An electronic signature that records the speed strokes in the signature are made, as well as the final result, is probably a really good way of identifying the purchaser. Of course then a lot of the smaller merchants would have to upgrade their 90's era dial-up credit card processing.

cryptoz 9y ago
> An electronic signature that records the speed strokes in the signature are made, as well as the final result, is probably a really good way of identifying the purchaser. Not so sure about that. If I use a different mouse then my "signature" using those metrics will fail, as I'm basically a completely different (inept) person using other mice. Just an example, but I don't think that e-signatures can be trusted by using machine learning techniques, at least not any that I know of today.
- dapreja 9y ago
  Yea no signature pads are pure shit.
- StillBored 9y ago
  Yes, part of it would require some level of standardization of the electronic pad/pen. I'm not really suggesting it as "trust" thing, but a better "pin" code that could be combined with the current shopping history/etc metrics being used to detect fraud.
supertrope 9y ago
Handwriting analysis was pseudo-scientific to begin with. The payments market has to accommodate illiterate customers, disabled customers, and people who have a broken hand at the moment. CC networks currently use previous purchasing patterns, location, time of day, etc. It's the same reasons why biometrics is not a magic security solution. Authentication must be based of something that's hard to fake, cheap to verify, random, and easy to change. What threshold would you set to balance false negatives and false positives? In comparison the private key in your chip card meets all of the criteria.
fulafel 9y ago
Indeed. Chip & PIN system at least in Europe has been a way to shift the abuse risk to the customer away from the banks. It's a lot easier to contest a signature than prove that you haven't been careless with protecting your PIN.
stordoff 9y ago
That assumes you are signing things often enough to have a signature that is reliably the same each time, which for me has _never_ been the case. Also, since developing a disability in my hands, that _can't_ be the case -- there _will_ be variations in stroke speed due to level of pain/stiffness at the time.

cryptoz 9y ago

> An electronic signature that records the speed strokes in the signature are made, as well as the final result, is probably a really good way of identifying the purchaser. Not so sure about that. If I use a different mouse then my "signature" using those metrics will fail, as I'm basically a completely different (inept) person using other mice. Just an example, but I don't think that e-signatures can be trusted by using machine learning techniques, at least not any that I know of today.

dapreja 9y ago
Yea no signature pads are pure shit.
StillBored 9y ago
Yes, part of it would require some level of standardization of the electronic pad/pen. I'm not really suggesting it as "trust" thing, but a better "pin" code that could be combined with the current shopping history/etc metrics being used to detect fraud.

supertrope 9y ago

Handwriting analysis was pseudo-scientific to begin with. The payments market has to accommodate illiterate customers, disabled customers, and people who have a broken hand at the moment. CC networks currently use previous purchasing patterns, location, time of day, etc. It's the same reasons why biometrics is not a magic security solution. Authentication must be based of something that's hard to fake, cheap to verify, random, and easy to change. What threshold would you set to balance false negatives and false positives? In comparison the private key in your chip card meets all of the criteria.

fulafel 9y ago

Indeed. Chip & PIN system at least in Europe has been a way to shift the abuse risk to the customer away from the banks. It's a lot easier to contest a signature than prove that you haven't been careless with protecting your PIN.

stordoff 9y ago

That assumes you are signing things often enough to have a signature that is reliably the same each time, which for me has _never_ been the case. Also, since developing a disability in my hands, that _can't_ be the case -- there _will_ be variations in stroke speed due to level of pain/stiffness at the time.