On Tue, Jun 26, 2012 at 10:08 PM, William "Chops" Westfield wrote: > > On Jun 26, 2012, at 11:35 AM, V G wrote: > > > I'm considering using client-side AES256 encryption on all first/last > names. The rest of the data is useless if the names are unknown. > > That wouldn't be secure at all. Well, I guess the database might be > secure, but communications with it wouldn't be. See who goes into the > office, see which encrypted stuff comes out. See "Known plaintext attack.= " > And you have key management issues. > > See also "deterministic encryption" - basically, you set up encryption so > that the same plaintext ALWAYS yields the same cyphertext, and the databa= se > can do all the lookups it needs to do without ever having access to the > plaintext. Key management is still an issue :-( > > (I took Stanford's online encryption class, and now I can see more things > that are bad ideas. Prof's first rule: you don't invent encryption syste= ms > yourself. Not algorithms, not key management, not the little details. > LOTS of examples of people taking reasonably secure algorithms and comin= g > up with overall systems that ended up having "obvious" attacks. 802.11 > WEP, for instance.) The medical database problems ought to be a known > problem with a known solution. Find it or buy it. It's probably expensi= ve > :-( ) > > No one's inventing encryption systems. I'm making use of standard AES encryption and choosing what gets encrypted and what doesn't. The data itself doesn't really matter. The important part is WHO the data belongs to - the names. If the names themselves can't be deciphered, then it should be good enough. What data is related to the names doesn't matter. The whole thing can be looked at as a key-value database with the key being names and the value being a bunch of numbers/blob text/etc. The value itself is meaningless. Example data: smokes, has cancer, enlarged left atrium, is an electrical engineer. That data in itself is useless. This isn't some secret government organization thing, so no one really cares about it in the first place, nor is anyone targeting it, nor would anyone spend any time trying to get it. I'm just being over cautious about this whole thing by nature. The data only becomes slightly useful if I know that the person it represents is William Westfield. That might make some people uncomfortable, and that's why I want to encrypt the names. If the names are encrypted as they leave the client (via javascript, and a password entered on the client only), then the server will have no information to decrypt it. Only the client does after the user enters the decryption key locally. But I do really hope there's a better way to do this. --=20 http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .