4 ms·

If you're going to use something that is permanent and unchangeable for authentication (which is itself dumb), wouldn't it make sense to store one way hashes ra

by beloch 11y ago

If you're going to use something that is permanent and unchangeable for authentication (which is itself dumb), wouldn't it make sense to store one way hashes rather than the fingerprints themselves? There are multiple layers of idiocy at work here.

Wohlf 11y ago
It's not for authentication, it's for background checks. These fingerprints are used to check for criminal connections in the past and future.
vinay427 11y ago
I'm not sure how a likeness comparison would work with a hashed fingerprint. Perhaps it would if a specific subset of the finger scan is always used to compare with the hashed subset.
- Lawtonfogle 11y ago
  My understanding is that some companies like Google have special image comparison algorithms that allow them to store a hash of an image but still perform a likeness comparison of the image. This is useful when filtering out (and alerting authorities about) images where storing the original is illegal, but also catching images in different formats, different resolutions, or even with some extent of color changes to them. While I would not consider it a simple problem, I would think that given enough resources something similar can be made for fingerprints. The problem is that, to my knowledge, nothing like this currently exists and you are not likely to get the needed budget approved for building it, especially in the land of government contracts.
  - cvwright 11y ago
    You're probably thinking of PhotoDNA from Microsoft: https://en.wikipedia.org/wiki/PhotoDNA https://en.wikipedia.org/wiki/PhotoDNA Apparently many of the biggest online services, including MS and Facebook, are using it to scan images uploaded by their users.
Lawtonfogle 11y ago
Last government technical document I read (not federal level) mandated passwords be stored using encryption. Not hashing, encryption. Security by bureaucracy is one of the few things worse than security by obscurity.
cvwright 11y ago
Yeah, protecting the "template" is a pretty basic thing in biometrics. For example, here's a paper from 2005 about a scheme for secure storage of fingerprint templates. http://www.cse.lehigh.edu/prr/Biometrics/Archive/Papers/Tuyls05.pdf http://www.cse.lehigh.edu/prr/Biometrics/Archive/Papers/Tuyl... I don't know much about this scheme in particular; it was just the first search result for "biometric authentication template". But this shows how people have been working on this topic for a long time now.
eli 11y ago
Does that really help? Unless the hashing algorithm is also a secret, an attacker with the full database of fingerprint hashes would still be able to take a random fingerprint, hash it, and then see if they have any OPM records -- isn't that the main threat scenario here? I guess it would make it harder for an attacker to manufacture and plant false fingerprints?
- viraptor 11y ago
  That's where salt idea comes in. Even if you know the hashing algorithm and the key, you still need to rehash it for every single entry in your database to check if you have a match. That assumes you can get a consistent description of a fingerprint though, or have some scheme for fuzzy match with hashing. Now I want to find a paper on this...
ryan-c 11y ago
I doubt that there is an algorithm that can consistently turn any good image of the same fingerprint into the same stream of bytes.
- computator 11y ago
  There certainly are such algorithms. The Henry Classification System is one such algorithm and it's been used for more than century[1]. Even in the 1940s and 50s, before digital computers, you could fingerprint a suspect and through a manual application of the algorithm obtain a unique ID ("the same stream of bytes" as you say). You could then do a look-up to see if he had a criminal record. [1] To be precise, the Henry system uses all ten prints, but single finger classifications also exist. https://en.wikipedia.org/wiki/Henry_Classification_System https://en.wikipedia.org/wiki/Henry_Classification_System
  - ryan-c 11y ago
    The Henry classification system seems to be useful only as an index to reduce the number of fingerprints a human investigator needs to compare against. It looks like it is expected to frequently have collisions.
mattlutze 11y ago
Something you are, something you have, something you know. Part of non-repudiation is that you've established authenticity absolutely, or as near to absolutely as makes identification reliable.

by beloch 11y ago

Wohlf 11y ago

It's not for authentication, it's for background checks. These fingerprints are used to check for criminal connections in the past and future.

vinay427 11y ago

I'm not sure how a likeness comparison would work with a hashed fingerprint. Perhaps it would if a specific subset of the finger scan is always used to compare with the hashed subset.

Lawtonfogle 11y ago
My understanding is that some companies like Google have special image comparison algorithms that allow them to store a hash of an image but still perform a likeness comparison of the image. This is useful when filtering out (and alerting authorities about) images where storing the original is illegal, but also catching images in different formats, different resolutions, or even with some extent of color changes to them. While I would not consider it a simple problem, I would think that given enough resources something similar can be made for fingerprints. The problem is that, to my knowledge, nothing like this currently exists and you are not likely to get the needed budget approved for building it, especially in the land of government contracts.
- cvwright 11y ago
  You're probably thinking of PhotoDNA from Microsoft: https://en.wikipedia.org/wiki/PhotoDNA https://en.wikipedia.org/wiki/PhotoDNA Apparently many of the biggest online services, including MS and Facebook, are using it to scan images uploaded by their users.

Lawtonfogle 11y ago

Last government technical document I read (not federal level) mandated passwords be stored using encryption. Not hashing, encryption. Security by bureaucracy is one of the few things worse than security by obscurity.

cvwright 11y ago

Yeah, protecting the "template" is a pretty basic thing in biometrics. For example, here's a paper from 2005 about a scheme for secure storage of fingerprint templates. http://www.cse.lehigh.edu/prr/Biometrics/Archive/Papers/Tuyls05.pdf http://www.cse.lehigh.edu/prr/Biometrics/Archive/Papers/Tuyl... I don't know much about this scheme in particular; it was just the first search result for "biometric authentication template". But this shows how people have been working on this topic for a long time now.

eli 11y ago

Does that really help? Unless the hashing algorithm is also a secret, an attacker with the full database of fingerprint hashes would still be able to take a random fingerprint, hash it, and then see if they have any OPM records -- isn't that the main threat scenario here? I guess it would make it harder for an attacker to manufacture and plant false fingerprints?

viraptor 11y ago
That's where salt idea comes in. Even if you know the hashing algorithm and the key, you still need to rehash it for every single entry in your database to check if you have a match. That assumes you can get a consistent description of a fingerprint though, or have some scheme for fuzzy match with hashing. Now I want to find a paper on this...

ryan-c 11y ago

I doubt that there is an algorithm that can consistently turn any good image of the same fingerprint into the same stream of bytes.

computator 11y ago
There certainly are such algorithms. The Henry Classification System is one such algorithm and it's been used for more than century[1]. Even in the 1940s and 50s, before digital computers, you could fingerprint a suspect and through a manual application of the algorithm obtain a unique ID ("the same stream of bytes" as you say). You could then do a look-up to see if he had a criminal record. [1] To be precise, the Henry system uses all ten prints, but single finger classifications also exist. https://en.wikipedia.org/wiki/Henry_Classification_System https://en.wikipedia.org/wiki/Henry_Classification_System
- ryan-c 11y ago
  The Henry classification system seems to be useful only as an index to reduce the number of fingerprints a human investigator needs to compare against. It looks like it is expected to frequently have collisions.

mattlutze 11y ago

Something you are, something you have, something you know. Part of non-repudiation is that you've established authenticity absolutely, or as near to absolutely as makes identification reliable.