Re: [liberationtech] [sunlightlabs] need advice on using hashes for preserving PII's utility for disambiguation while protecting sensitive info
Arggh. Wrong link. Apologies to all and thanks to James McKinney. That's what I get for having that many tabs open. https://sunlightfoundation.com/blog/2014/03/20/a-little-math-could-make-identifiers-a-whole-lot-better/ On Thu, Mar 20, 2014 at 5:44 PM, James McKinney ja...@opennorth.ca wrote: Do you mean this post? https://sunlightfoundation.com/blog/2014/03/20/a-little-math-could-make-identifiers-a-whole-lot-better/ On Mar 20, 2014, at 3:44 PM, Tom Lee t...@sunlightfoundation.com wrote: Thanks again to everyone who helped me think through how government's approach to disclosing identifiers could be improved through checksums, tokenization and related techniques -- it was extremely helpful. The resulting post is here: https://sunlightfoundation.com/blog/2013/07/25/the-sunlight-foundations-comments-on-the-faas-proposed-open-data-policy/ I'd be grateful for any feedback -- or, especially, corrections -- that might occur to you. On Thu, Feb 6, 2014 at 3:49 PM, Tom Lee t...@sunlightfoundation.comwrote: We've been kicking around an idea at Sunlight that aims to use cryptographic ideas to resolve some of the concerns around the publication of publicly identifiable information in government disclosures. I could use some smart people to tell me what's dumb about it. We often face challenges related to disambiguating entities: is the John Smith who gave political donation A the same John Smith that gave political donation B? One obvious solution to this problem is to push to expand the information that's collected and disclosed -- if we had John's driver's license number (DLN), for instance, it'd be easy to disambiguate these records. But that could introduce privacy concerns for John. One approach to this problem (which I don't think government has tried) is employing a one-way hash. Obviously the input key space for DLNs and most other personal ID numbers is so small that reversing this with a dictionary attack would be trivial. You can add a salt, but only on a per-entity basis (not a per-record basis) if you want to preserve the capacity to disambiguate. That in turns calls for a lookup table in which the input keys are stored, which kind of defeats the point of using a hash (you might as well just assign random output IDs for each input ID). I would worry about government's ability to keep this lookup table secure, and I worry about the brittleness of such a system. Alternately, you can use a single system-wide secret (or set of secrets) to transform inputs into reliable outputs. I think this is less brittle and maybe easier to preserve as a secret, but this system might be too easily reversible given the ability to observe its outputs and know the universe of possible inputs. I'm unsure of the cryptographic options that might be appropriate here. For all I know, the lack of implementations using this kind of one-way transformation isn't about government sluggishness but rather about its feasibility. I'd be very curious to hear folks ideas on this score, though. My general hunch is that something must be possible -- even a few bits' worth of disambiguating information would be hugely useful to us, and presumably you're not leaking important amounts of information by, say, sharing the last digit of a DLN. So there must be a spectrum of options. But as is probably apparent, I don't think I've got a handle on how to think about this problem rigorously. Tom -- You received this message because you are subscribed to the Google Groups sunlightlabs group. To unsubscribe from this group and stop receiving emails from it, send an email to sunlightlabs+unsubscr...@googlegroups.com. To post to this group, send email to sunlightl...@googlegroups.com. Visit this group at http://groups.google.com/group/sunlightlabs. For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups sunlightlabs group. To unsubscribe from this group and stop receiving emails from it, send an email to sunlightlabs+unsubscr...@googlegroups.com. To post to this group, send email to sunlightl...@googlegroups.com. Visit this group at http://groups.google.com/group/sunlightlabs. For more options, visit https://groups.google.com/d/optout. -- Liberationtech is public archives are searchable on Google. Violations of list guidelines will get you moderated: https://mailman.stanford.edu/mailman/listinfo/liberationtech. Unsubscribe, change to digest, or change password by emailing moderator at compa...@stanford.edu.
Re: [liberationtech] [sunlightlabs] need advice on using hashes for preserving PII's utility for disambiguation while protecting sensitive info
-BEGIN PGP SIGNED MESSAGE- Hash: SHA256 On 06/02/14 20:56, Margie Roswell wrote: For all I know, the lack of implementations using this kind of one-way transformation isn't about government sluggishness but rather about its feasibility. I'd be very curious to hear folks ideas on this score, though. My general hunch is that something must be possible -- even a few bits' worth of disambiguating information would be hugely useful to us, and presumably you're not leaking important amounts of information by, say, sharing the last digit of a DLN. So there must be a spectrum of options. But as is probably apparent, I don't think I've got a handle on how to think about this problem rigorously. Even if you had a perfect method of anonymising the individual records, they might be reidentifiable by examining the whole dataset: http://33bits.org/2010/06/21/myths-and-fallacies-of-personally-identifiable-information/ http://randomwalker.info/social-networks/ http://www.cs.utexas.edu/~shmat/shmat_oak08netflix.pdf At the level of individual records, you could use modular exponentiation to anonymise the data. You pick a prime modulus p, and each organisation that's going to publish anonymised data picks a random secret value. Organisation X with secret value x anonymises a piece of data d by publishing d_x = d^x mod p, and organisation Y with secret value y anonymises the same data by publishing d_y = d^y mod p. If X and Y want to know which records they have in common, X takes the data published by Y and calculates d_x' = d_y^x mod p = d^(yx) mod p, and Y takes the data published by X and calculates d_y' = d_x^y mod p = d^(xy) mod p. For each record in common, d_x' = d_y', but neither can de-anonymise records published by the other that they don't have in common. This can be extended to more than two organisations: pass the records round in a circle, and when they get back to you they've been exponentiated by all the secret values (order doesn't matter). Now you can see which records you have in common with all the other organisations. (Maybe. IANAC.) Cheers, Michael -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.10 (GNU/Linux) iQEcBAEBCAAGBQJS9NWaAAoJEBEET9GfxSfMyWcH/1Au9/066O/3AaPkkid8nBhq 2uuNjjLgDWzE+5aTIQGMzk9yy85TRKlXKdC4c9/n0UXxJjAUYxkLSoNkAD33ej36 s/oi3pI0C9OQ1MffJVCSImA+NwQ0QqDG6DOUBNPRoBUTr/nd5efbBRwWVtLSn50D 0QlLJYXUGGB+fSMZKyy368rrx5Ue8ICQOzIUyNJ3sWZsQEJo0nE8WJd1+89GlR45 XPRSUUma/5DCECl9gWBFq5pVuEtf29KoXV6QLCzagWCaAa2dNlCspoGp4bVlkBz9 UWMJRFHYDj9AxzUKt5Vi++uh6nYrTu++a7bXqOGJHb9y8VL54JHweEXNW2xWyog= =BrUY -END PGP SIGNATURE- -- Liberationtech is public archives are searchable on Google. Violations of list guidelines will get you moderated: https://mailman.stanford.edu/mailman/listinfo/liberationtech. Unsubscribe, change to digest, or change password by emailing moderator at compa...@stanford.edu.
Re: [liberationtech] [sunlightlabs] need advice on using hashes for preserving PII's utility for disambiguation while protecting sensitive info
PII = personally identifiable information (Anyone who can address the question probably already knows that... but I was curious, and figured I'd spare others the look-up.) -- http://FarmBillPrimer.org http://www.BaltimoreUrbanAg.org (Please send events; This site is hungry.) http://www.ExcellentNutrition.org http://www.packtpub.com/drupal-5-views-recipes/book On Thu, Feb 6, 2014 at 3:49 PM, Tom Lee t...@sunlightfoundation.com wrote: We've been kicking around an idea at Sunlight that aims to use cryptographic ideas to resolve some of the concerns around the publication of publicly identifiable information in government disclosures. I could use some smart people to tell me what's dumb about it. We often face challenges related to disambiguating entities: is the John Smith who gave political donation A the same John Smith that gave political donation B? One obvious solution to this problem is to push to expand the information that's collected and disclosed -- if we had John's driver's license number (DLN), for instance, it'd be easy to disambiguate these records. But that could introduce privacy concerns for John. One approach to this problem (which I don't think government has tried) is employing a one-way hash. Obviously the input key space for DLNs and most other personal ID numbers is so small that reversing this with a dictionary attack would be trivial. You can add a salt, but only on a per-entity basis (not a per-record basis) if you want to preserve the capacity to disambiguate. That in turns calls for a lookup table in which the input keys are stored, which kind of defeats the point of using a hash (you might as well just assign random output IDs for each input ID). I would worry about government's ability to keep this lookup table secure, and I worry about the brittleness of such a system. Alternately, you can use a single system-wide secret (or set of secrets) to transform inputs into reliable outputs. I think this is less brittle and maybe easier to preserve as a secret, but this system might be too easily reversible given the ability to observe its outputs and know the universe of possible inputs. I'm unsure of the cryptographic options that might be appropriate here. For all I know, the lack of implementations using this kind of one-way transformation isn't about government sluggishness but rather about its feasibility. I'd be very curious to hear folks ideas on this score, though. My general hunch is that something must be possible -- even a few bits' worth of disambiguating information would be hugely useful to us, and presumably you're not leaking important amounts of information by, say, sharing the last digit of a DLN. So there must be a spectrum of options. But as is probably apparent, I don't think I've got a handle on how to think about this problem rigorously. Tom -- You received this message because you are subscribed to the Google Groups sunlightlabs group. To unsubscribe from this group and stop receiving emails from it, send an email to sunlightlabs+unsubscr...@googlegroups.com. To post to this group, send email to sunlightl...@googlegroups.com. Visit this group at http://groups.google.com/group/sunlightlabs. For more options, visit https://groups.google.com/groups/opt_out. -- Liberationtech is public archives are searchable on Google. Violations of list guidelines will get you moderated: https://mailman.stanford.edu/mailman/listinfo/liberationtech. Unsubscribe, change to digest, or change password by emailing moderator at compa...@stanford.edu.
Re: [liberationtech] [sunlightlabs] need advice on using hashes for preserving PII's utility for disambiguation while protecting sensitive info
Just one thought to throw out: Something that sprang to mind is the idea of a check digit or simplified hash that would be redundant enough to collide very often if you were trying to reverse, but would still provide enough disambiguation that you'd be able to appropriately determine who you're dealing with. You could probably use something similar to the Luhn algorithm for that, although I'm not sure how uniform that is: http://en.wikipedia.org/wiki/Luhn_algorithm - also, that only ends up with a single check digit, which is probably too small for good disambiguation. The approach in general might still be helpful though. -Chris On Thu, Feb 6, 2014 at 3:49 PM, Tom Lee t...@sunlightfoundation.com wrote: We've been kicking around an idea at Sunlight that aims to use cryptographic ideas to resolve some of the concerns around the publication of publicly identifiable information in government disclosures. I could use some smart people to tell me what's dumb about it. We often face challenges related to disambiguating entities: is the John Smith who gave political donation A the same John Smith that gave political donation B? One obvious solution to this problem is to push to expand the information that's collected and disclosed -- if we had John's driver's license number (DLN), for instance, it'd be easy to disambiguate these records. But that could introduce privacy concerns for John. One approach to this problem (which I don't think government has tried) is employing a one-way hash. Obviously the input key space for DLNs and most other personal ID numbers is so small that reversing this with a dictionary attack would be trivial. You can add a salt, but only on a per-entity basis (not a per-record basis) if you want to preserve the capacity to disambiguate. That in turns calls for a lookup table in which the input keys are stored, which kind of defeats the point of using a hash (you might as well just assign random output IDs for each input ID). I would worry about government's ability to keep this lookup table secure, and I worry about the brittleness of such a system. Alternately, you can use a single system-wide secret (or set of secrets) to transform inputs into reliable outputs. I think this is less brittle and maybe easier to preserve as a secret, but this system might be too easily reversible given the ability to observe its outputs and know the universe of possible inputs. I'm unsure of the cryptographic options that might be appropriate here. For all I know, the lack of implementations using this kind of one-way transformation isn't about government sluggishness but rather about its feasibility. I'd be very curious to hear folks ideas on this score, though. My general hunch is that something must be possible -- even a few bits' worth of disambiguating information would be hugely useful to us, and presumably you're not leaking important amounts of information by, say, sharing the last digit of a DLN. So there must be a spectrum of options. But as is probably apparent, I don't think I've got a handle on how to think about this problem rigorously. Tom -- You received this message because you are subscribed to the Google Groups sunlightlabs group. To unsubscribe from this group and stop receiving emails from it, send an email to sunlightlabs+unsubscr...@googlegroups.com. To post to this group, send email to sunlightl...@googlegroups.com. Visit this group at http://groups.google.com/group/sunlightlabs. For more options, visit https://groups.google.com/groups/opt_out. -- Liberationtech is public archives are searchable on Google. Violations of list guidelines will get you moderated: https://mailman.stanford.edu/mailman/listinfo/liberationtech. Unsubscribe, change to digest, or change password by emailing moderator at compa...@stanford.edu.
Re: [liberationtech] [sunlightlabs] need advice on using hashes for preserving PII's utility for disambiguation while protecting sensitive info
It's been a while since I dug into it, but something like an 8-bit CRChttp://en.wikipedia.org/wiki/Cyclic_redundancy_checkwould probably provide enough disambiguation but would collide often enough to not be much of a concern for reversing - 256 different values. On Thu, Feb 6, 2014 at 4:10 PM, Chris Dary umb...@gmail.com wrote: Just one thought to throw out: Something that sprang to mind is the idea of a check digit or simplified hash that would be redundant enough to collide very often if you were trying to reverse, but would still provide enough disambiguation that you'd be able to appropriately determine who you're dealing with. You could probably use something similar to the Luhn algorithm for that, although I'm not sure how uniform that is: http://en.wikipedia.org/wiki/Luhn_algorithm - also, that only ends up with a single check digit, which is probably too small for good disambiguation. The approach in general might still be helpful though. -Chris On Thu, Feb 6, 2014 at 3:49 PM, Tom Lee t...@sunlightfoundation.comwrote: We've been kicking around an idea at Sunlight that aims to use cryptographic ideas to resolve some of the concerns around the publication of publicly identifiable information in government disclosures. I could use some smart people to tell me what's dumb about it. We often face challenges related to disambiguating entities: is the John Smith who gave political donation A the same John Smith that gave political donation B? One obvious solution to this problem is to push to expand the information that's collected and disclosed -- if we had John's driver's license number (DLN), for instance, it'd be easy to disambiguate these records. But that could introduce privacy concerns for John. One approach to this problem (which I don't think government has tried) is employing a one-way hash. Obviously the input key space for DLNs and most other personal ID numbers is so small that reversing this with a dictionary attack would be trivial. You can add a salt, but only on a per-entity basis (not a per-record basis) if you want to preserve the capacity to disambiguate. That in turns calls for a lookup table in which the input keys are stored, which kind of defeats the point of using a hash (you might as well just assign random output IDs for each input ID). I would worry about government's ability to keep this lookup table secure, and I worry about the brittleness of such a system. Alternately, you can use a single system-wide secret (or set of secrets) to transform inputs into reliable outputs. I think this is less brittle and maybe easier to preserve as a secret, but this system might be too easily reversible given the ability to observe its outputs and know the universe of possible inputs. I'm unsure of the cryptographic options that might be appropriate here. For all I know, the lack of implementations using this kind of one-way transformation isn't about government sluggishness but rather about its feasibility. I'd be very curious to hear folks ideas on this score, though. My general hunch is that something must be possible -- even a few bits' worth of disambiguating information would be hugely useful to us, and presumably you're not leaking important amounts of information by, say, sharing the last digit of a DLN. So there must be a spectrum of options. But as is probably apparent, I don't think I've got a handle on how to think about this problem rigorously. Tom -- You received this message because you are subscribed to the Google Groups sunlightlabs group. To unsubscribe from this group and stop receiving emails from it, send an email to sunlightlabs+unsubscr...@googlegroups.com. To post to this group, send email to sunlightl...@googlegroups.com. Visit this group at http://groups.google.com/group/sunlightlabs. For more options, visit https://groups.google.com/groups/opt_out. -- Liberationtech is public archives are searchable on Google. Violations of list guidelines will get you moderated: https://mailman.stanford.edu/mailman/listinfo/liberationtech. Unsubscribe, change to digest, or change password by emailing moderator at compa...@stanford.edu.
Re: [liberationtech] [sunlightlabs] need advice on using hashes for preserving PII's utility for disambiguation while protecting sensitive info
On 02/06/2014 03:49 PM, Tom Lee wrote: Obviously the input key space for DLNs and most other personal ID numbers is so small that reversing this with a dictionary attack would be trivial. You can add a salt, but only on a per-entity basis (not a per-record basis) if you want to preserve the capacity to disambiguate. That in turns calls for a lookup table in which the input keys are stored, which kind of defeats the point of using a hash (you might as well just assign random output IDs for each input ID). I would worry about government's ability to keep this lookup table secure, and I worry about the brittleness of such a system. And yet a lookup table mapping inputs to random outputs might be the best worst option. Even if the right cryptographic method (hash, encryption, etc.) can be found and is mathematically sound, I'd have /very/ low confidence that it would be implemented correctly. Maybe one office does it right, the next office says hey that's a great idea but forgets that hashing a four digit pin doesn't provide any obscurity, etc. (That's not a jab at government. Crypto is so hard.) I'd ask, for a particular case, what data does the data source already have? If they /already/ have DLNs in their database, there's no added privacy concern in creating a random mapping to unique identifiers for public consumption. (Besides the mosaic effect, but that aside.) Assuming the data source can make the distinction at all internally, they must have /something/ already in their database. HTH, - Josh Tauberer (@JoshData) http://razor.occams.info -- Liberationtech is public archives are searchable on Google. Violations of list guidelines will get you moderated: https://mailman.stanford.edu/mailman/listinfo/liberationtech. Unsubscribe, change to digest, or change password by emailing moderator at compa...@stanford.edu.
Re: [liberationtech] [sunlightlabs] need advice on using hashes for preserving PII's utility for disambiguation while protecting sensitive info
I don't know how these government databases are maintained in the US, but in Canada it's not infrequent for such databases to be more-or-less write only - the government fills up a database with names, donation amounts, postcodes, etc. and then publishes it somewhere for others to consume. In a subsequent year, it fills up a fresh database - maybe it maintains the same database schema, but in every other respect it's as if the old database didn't exist. If we go with the solution of generating a new ID for each donor, there will have to be better coordination within and between agencies to store this information centrally in order for them to share IDs across time and location. That's a security risk. Can we guarantee that each agency will have the same private information to create identifiers from? If so, as Chris mentions, a CRC can be used to disambiguate, i.e. match donors on name, etc. and resolve collisions by looking at the CRC. On 2014-02-06, at 4:19 PM, Chris Dary wrote: It's been a while since I dug into it, but something like an 8-bit CRC would probably provide enough disambiguation but would collide often enough to not be much of a concern for reversing - 256 different values. On Thu, Feb 6, 2014 at 4:10 PM, Chris Dary umb...@gmail.com wrote: Just one thought to throw out: Something that sprang to mind is the idea of a check digit or simplified hash that would be redundant enough to collide very often if you were trying to reverse, but would still provide enough disambiguation that you'd be able to appropriately determine who you're dealing with. You could probably use something similar to the Luhn algorithm for that, although I'm not sure how uniform that is: http://en.wikipedia.org/wiki/Luhn_algorithm - also, that only ends up with a single check digit, which is probably too small for good disambiguation. The approach in general might still be helpful though. -Chris On Thu, Feb 6, 2014 at 3:49 PM, Tom Lee t...@sunlightfoundation.com wrote: We've been kicking around an idea at Sunlight that aims to use cryptographic ideas to resolve some of the concerns around the publication of publicly identifiable information in government disclosures. I could use some smart people to tell me what's dumb about it. We often face challenges related to disambiguating entities: is the John Smith who gave political donation A the same John Smith that gave political donation B? One obvious solution to this problem is to push to expand the information that's collected and disclosed -- if we had John's driver's license number (DLN), for instance, it'd be easy to disambiguate these records. But that could introduce privacy concerns for John. One approach to this problem (which I don't think government has tried) is employing a one-way hash. Obviously the input key space for DLNs and most other personal ID numbers is so small that reversing this with a dictionary attack would be trivial. You can add a salt, but only on a per-entity basis (not a per-record basis) if you want to preserve the capacity to disambiguate. That in turns calls for a lookup table in which the input keys are stored, which kind of defeats the point of using a hash (you might as well just assign random output IDs for each input ID). I would worry about government's ability to keep this lookup table secure, and I worry about the brittleness of such a system. Alternately, you can use a single system-wide secret (or set of secrets) to transform inputs into reliable outputs. I think this is less brittle and maybe easier to preserve as a secret, but this system might be too easily reversible given the ability to observe its outputs and know the universe of possible inputs. I'm unsure of the cryptographic options that might be appropriate here. For all I know, the lack of implementations using this kind of one-way transformation isn't about government sluggishness but rather about its feasibility. I'd be very curious to hear folks ideas on this score, though. My general hunch is that something must be possible -- even a few bits' worth of disambiguating information would be hugely useful to us, and presumably you're not leaking important amounts of information by, say, sharing the last digit of a DLN. So there must be a spectrum of options. But as is probably apparent, I don't think I've got a handle on how to think about this problem rigorously. Tom -- You received this message because you are subscribed to the Google Groups sunlightlabs group. To unsubscribe from this group and stop receiving emails from it, send an email to sunlightlabs+unsubscr...@googlegroups.com. To post to this group, send email to sunlightl...@googlegroups.com. Visit this group at http://groups.google.com/group/sunlightlabs. For more options, visit https://groups.google.com/groups/opt_out. -- You received this message because