4 ms·

Enigma: Decentralized Computation Platform with Guaranteed Privacy [pdf]

82 points by ff_ 11y ago

dang 11y ago
Discussed at https://news.ycombinator.com/item?id=9805385 https://news.ycombinator.com/item?id=9805385, but perhaps a link to the paper will receive more substantive discussion.
- ff_ 11y ago
  Agreed. Honestly both the article you linked and the website [1] were unsatisfying to me, so I read the paper to better understand the implications. BTW thank you for linking the previous discussion, I missed it! [1]: enigma.media.mit.edu
EGreg 11y ago
Sometimes I wonder what kind of operations would actually be useful to do on such data in the real world. I mean, the way the web currently works is that I trust some server to host my data. I can have this service auth an external consumer site and display data in an iframe, say, which the consumer site can't get at. This is good enough for displaying people their personalized info (name, friends) on various services (eg directions to their house in an iframe, for a user authenticated with my chosen provider). But to go further, what if I don't want to trust any provider? Then I could simply encrypt the data and store encrypted data with the provider (or providers for redundancy). The authentication could be replaced with visitors holding a key to decrypt the data (because I gave it to them) and I can switch to using some other key and effectively "unfriend" those who don't get my updated key. But all this is good enough for displaying data and files I upload. Now, why would I want to do operations on those files "in the cloud" without trusting a provider? I am already trusting my friends with the data, since they can reshare it once it's displayed to them. So why not trust a provider? One of my friends can run the provider. I guess the only scenario I see it being useful is if all my friends can only have limited access to the data and all manipulations on the data are collaborative, and that's where the homeomorphism comes in. Perhaps no one will be able to see the whole data and it's not really about data at all, but views of some Enigmatic process running on some network (like an autonomous corporation.) is that the use case?
- sjbase 11y ago
  Maybe this could be used for something like clinical quality analytics. There is an industry of data-driven optimization on dimensions like diagnoses, treatments, outcomes, populations/demographics. A big challenge with this is that the useful data is HIPAA-regulated. Something like SETI@home, but for finding better clinical care options instead of finding aliens.
- amelius 11y ago
  Also for efficiently searching through your data, your provider needs access to the data (or I guess something like homeomorphism).
- oznathan 11y ago
  Yes, that is exactly the case. The idea is to let 3rd parties use the data without ever having direct access to it. They can't reconstruct it, and only get the result of the computation.
berzniz 11y ago
Go Oz!
- oznathan 11y ago
  Thanks Tal!
reilly3000 11y ago
Phenomenal stuff. Can't wait to see this play out IRL.

82 points by ff_ 11y ago

dang 11y ago

Discussed at https://news.ycombinator.com/item?id=9805385 https://news.ycombinator.com/item?id=9805385, but perhaps a link to the paper will receive more substantive discussion.

ff_ 11y ago
Agreed. Honestly both the article you linked and the website [1] were unsatisfying to me, so I read the paper to better understand the implications. BTW thank you for linking the previous discussion, I missed it! [1]: enigma.media.mit.edu

EGreg 11y ago

Sometimes I wonder what kind of operations would actually be useful to do on such data in the real world. I mean, the way the web currently works is that I trust some server to host my data. I can have this service auth an external consumer site and display data in an iframe, say, which the consumer site can't get at. This is good enough for displaying people their personalized info (name, friends) on various services (eg directions to their house in an iframe, for a user authenticated with my chosen provider). But to go further, what if I don't want to trust any provider? Then I could simply encrypt the data and store encrypted data with the provider (or providers for redundancy). The authentication could be replaced with visitors holding a key to decrypt the data (because I gave it to them) and I can switch to using some other key and effectively "unfriend" those who don't get my updated key. But all this is good enough for displaying data and files I upload. Now, why would I want to do operations on those files "in the cloud" without trusting a provider? I am already trusting my friends with the data, since they can reshare it once it's displayed to them. So why not trust a provider? One of my friends can run the provider. I guess the only scenario I see it being useful is if all my friends can only have limited access to the data and all manipulations on the data are collaborative, and that's where the homeomorphism comes in. Perhaps no one will be able to see the whole data and it's not really about data at all, but views of some Enigmatic process running on some network (like an autonomous corporation.) is that the use case?

sjbase 11y ago
Maybe this could be used for something like clinical quality analytics. There is an industry of data-driven optimization on dimensions like diagnoses, treatments, outcomes, populations/demographics. A big challenge with this is that the useful data is HIPAA-regulated. Something like SETI@home, but for finding better clinical care options instead of finding aliens.
amelius 11y ago
Also for efficiently searching through your data, your provider needs access to the data (or I guess something like homeomorphism).
oznathan 11y ago
Yes, that is exactly the case. The idea is to let 3rd parties use the data without ever having direct access to it. They can't reconstruct it, and only get the result of the computation.

berzniz 11y ago

Go Oz!

oznathan 11y ago
Thanks Tal!

reilly3000 11y ago

Phenomenal stuff. Can't wait to see this play out IRL.