3 ms·

I'm intersted in knowing how the revocation of access to data at a later time could be possible. Also it seems likely that any organization you gave access to c

by retox 9y ago

I'm intersted in knowing how the revocation of access to data at a later time could be possible. Also it seems likely that any organization you gave access to could simple store and sell the data. The biggest problem I can see from the article is that once someone loses their key their identity is lost forever and they must rebuild the data. The first few times that happens the public will be turned off to the idea.

jameslevy 9y ago
It's very difficult, arguably impossible, to revoke access to data that's already been accessed. You could try a DRM-type scheme but there's always ways to get around it. What you could do, pretty easily, is revoke access to the latest data that hasn't already been shared, and to have an audit trail of exactly what data has been accessed and when it was accessed. That would still be a massive improvement on the status quo. I wouldn't rule out the eventual possibility of something like Facebook migrating their Open Graph API to some type of encrypted blockchain. Although I would be completely shocked if it were more than essentially a publicity tactic that wasn't actually using any true decentralization.
- noddy1 9y ago
  For medical records - consider a secure reader tablet-like device. It has encrypted hardware which can be temporarily permissioned to have access to encrypted data based on blockchain timestamps. It is tamperproof - if you open it it has some self-destruct mechanism. Data is accessed in chunks, therefore the owner of the data can see whether data is being accessed in the way a human might use it, or if it is being downloaded en masse. The organisation with data access devices gets audited on an annual basis to see where all their access devices are and to ensure none of them are being tampered with. Suddenly, we get to a situation where the security of the medical file begins to resemble that of the paper-based medical records department. If you want to copy records, you need to physically go to where they are, get access to each individual file, and go page-by-page photographing each.
  - lawpoop 9y ago
    This is actually a genius application. I hope more HN readers read it and this meme spreads.
  - reitanqild 9y ago
    The improvement here is this: a secure reader tablet-like device. It has encrypted hardware which can be temporarily permissioned to have access to encrypted data This part is how it works today and it has been like this for years: Data is accessed in chunks, therefore the owner of the data can see whether data is being accessed in the way a human might use it, or if it is being downloaded en masse. The organisation with data access devices gets audited on an annual basis to see where all their access devices are and to ensure none of them are being tampered with. No blockchain is needed to achieve this.
    - noddy1 9y ago
      it is if you are working across multiple jurisdictions/regimes and you don't want a centralized point of failure.
- dogma1138 9y ago
  There isn’t a single DRM scheme that cannot be technically broken even it its with pen and paper. DRM is only as strong as the legal mechanism behind it. Since the DMCA and other legislation protects DRM the actual strength of the DRM are irrelevant. DVDs are still protected by the same broken key, BR are also broken every DRM system gets broken or is easily circumvented through side channels.
flak48 9y ago
Maybe by ensuring all data access/decryption happens from the client side perhaps, so that the data never flows(unencrpyted) through corporate servers. Like FB implemented as Signal on steroids, when I add a friend I exchange keys to open up a channel with him to enable him to see posts/photos etc from my life, but the org (say FB) itself cannot see any of this data. Not sure how blockchain will fit in here though or enhance this
- noddy1 9y ago
  blockchain would allow: - trustless cloud-based indexed storage of encrypted data using protocols such as IPFS - logging of data access requests to a globally accessible database and ability to give and remove permissions based on timestamps - auditing of data accessors to ensure that their access patterns match their functions - eg. if a hospital suddenly starts trying to download data on a million patients instead of 300 then this would be visible on the blockchain and individuals could then question why this is occurring and prevent access. All of these things can be done with trusted intermediaries, but when you start considering use cases such as - being in china and needing to allow a chinese hospital to access your records back in the USA, suddenly these things become important.
  - acdha 9y ago
    That’s a lot of buzzwords but it fails to answer how any of it would work, why you’d want that, or why it’s better than mature alternatives. For example, how are you defining trustless? Personal data systems needs to be based on trust and reputation (otherwise you’ll have a bunch of 4chaners spamming claims about you) so are you simply thinking tamper-evidence and non-repudiation? That’d raise the questions of why this would be better than PKI, Merkel trees, etc. All of the other claims hit similar problems: e.g. you can’t store the data in a blockchain and make any of those promises, and it’s not clear how useful a shared index or log would be when you can’t trust all of the participants. Another fatal flaw is the lost of control and inability to correct mistakes. If there’s ever a vulnerability all of that data is irrecoverably public. What organization is going to turn control of something they get sued for over to an entity they don’t control? “They told us the software was perfect” doesn’t seem like a good HIPAA defense tactic to me.

by retox 9y ago

jameslevy 9y ago

It's very difficult, arguably impossible, to revoke access to data that's already been accessed. You could try a DRM-type scheme but there's always ways to get around it. What you could do, pretty easily, is revoke access to the latest data that hasn't already been shared, and to have an audit trail of exactly what data has been accessed and when it was accessed. That would still be a massive improvement on the status quo. I wouldn't rule out the eventual possibility of something like Facebook migrating their Open Graph API to some type of encrypted blockchain. Although I would be completely shocked if it were more than essentially a publicity tactic that wasn't actually using any true decentralization.

noddy1 9y ago
For medical records - consider a secure reader tablet-like device. It has encrypted hardware which can be temporarily permissioned to have access to encrypted data based on blockchain timestamps. It is tamperproof - if you open it it has some self-destruct mechanism. Data is accessed in chunks, therefore the owner of the data can see whether data is being accessed in the way a human might use it, or if it is being downloaded en masse. The organisation with data access devices gets audited on an annual basis to see where all their access devices are and to ensure none of them are being tampered with. Suddenly, we get to a situation where the security of the medical file begins to resemble that of the paper-based medical records department. If you want to copy records, you need to physically go to where they are, get access to each individual file, and go page-by-page photographing each.
- lawpoop 9y ago
  This is actually a genius application. I hope more HN readers read it and this meme spreads.
- reitanqild 9y ago
  The improvement here is this: a secure reader tablet-like device. It has encrypted hardware which can be temporarily permissioned to have access to encrypted data This part is how it works today and it has been like this for years: Data is accessed in chunks, therefore the owner of the data can see whether data is being accessed in the way a human might use it, or if it is being downloaded en masse. The organisation with data access devices gets audited on an annual basis to see where all their access devices are and to ensure none of them are being tampered with. No blockchain is needed to achieve this.
  - noddy1 9y ago
    it is if you are working across multiple jurisdictions/regimes and you don't want a centralized point of failure.
dogma1138 9y ago
There isn’t a single DRM scheme that cannot be technically broken even it its with pen and paper. DRM is only as strong as the legal mechanism behind it. Since the DMCA and other legislation protects DRM the actual strength of the DRM are irrelevant. DVDs are still protected by the same broken key, BR are also broken every DRM system gets broken or is easily circumvented through side channels.

flak48 9y ago

Maybe by ensuring all data access/decryption happens from the client side perhaps, so that the data never flows(unencrpyted) through corporate servers. Like FB implemented as Signal on steroids, when I add a friend I exchange keys to open up a channel with him to enable him to see posts/photos etc from my life, but the org (say FB) itself cannot see any of this data. Not sure how blockchain will fit in here though or enhance this

noddy1 9y ago
blockchain would allow: - trustless cloud-based indexed storage of encrypted data using protocols such as IPFS - logging of data access requests to a globally accessible database and ability to give and remove permissions based on timestamps - auditing of data accessors to ensure that their access patterns match their functions - eg. if a hospital suddenly starts trying to download data on a million patients instead of 300 then this would be visible on the blockchain and individuals could then question why this is occurring and prevent access. All of these things can be done with trusted intermediaries, but when you start considering use cases such as - being in china and needing to allow a chinese hospital to access your records back in the USA, suddenly these things become important.
- acdha 9y ago
  That’s a lot of buzzwords but it fails to answer how any of it would work, why you’d want that, or why it’s better than mature alternatives. For example, how are you defining trustless? Personal data systems needs to be based on trust and reputation (otherwise you’ll have a bunch of 4chaners spamming claims about you) so are you simply thinking tamper-evidence and non-repudiation? That’d raise the questions of why this would be better than PKI, Merkel trees, etc. All of the other claims hit similar problems: e.g. you can’t store the data in a blockchain and make any of those promises, and it’s not clear how useful a shared index or log would be when you can’t trust all of the participants. Another fatal flaw is the lost of control and inability to correct mistakes. If there’s ever a vulnerability all of that data is irrecoverably public. What organization is going to turn control of something they get sued for over to an entity they don’t control? “They told us the software was perfect” doesn’t seem like a good HIPAA defense tactic to me.