168 reputation
8
bio website kanjibox.net
location Kyoto
age 93
visits member for 8 months
seen Dec 19 '13 at 10:21

Bioinformatics, Machine-learning and (a very little bit of) Crypto.


Dec
18
awarded  Supporter
Dec
18
accepted Logical OR operation in a homomorphic additive cryptosystem
Dec
18
comment Logical OR operation in a homomorphic additive cryptosystem
Indeed, you're right. Completely forgot about NAND. That would indeed negate the non-FHE condition. At least now I know there's no point searching!
Dec
17
asked Logical OR operation in a homomorphic additive cryptosystem
Nov
17
awarded  Scholar
Nov
17
accepted Homomorphic (encrypted) comparison to an integer
Nov
16
revised Homomorphic (encrypted) comparison to an integer
added 1 characters in body
Nov
16
comment Homomorphic (encrypted) comparison to an integer
No: thanks for your patience. Secure computing is far from my specialty and I'm learning as I go (that being said, it's interesting that the first paper I went with, a refereed published paper, had a much different version of what seems to be the right conditions)...
Nov
16
awarded  Teacher
Nov
16
revised Homomorphic (encrypted) comparison to an integer
added 8 characters in body
Nov
16
awarded  Commentator
Nov
16
revised Homomorphic (encrypted) comparison to an integer
added 596 characters in body
Nov
16
revised Homomorphic (encrypted) comparison to an integer
added 41 characters in body
Nov
16
revised Homomorphic (encrypted) comparison to an integer
added 456 characters in body
Nov
16
comment Homomorphic (encrypted) comparison to an integer
let us continue this discussion in chat
Nov
16
comment Homomorphic (encrypted) comparison to an integer
@D.W.: I do, but while finding the exact limit for which the statistical test would work (assuming we can agree on a significance level) might be difficult (now looking at it...), finding a conservative range (represented by the difference between $r$ and $max(y-x)$) for which it is trivially verified, is not particularly difficult. For a start, checking a binary value ($(y-x) \in {0,1}$), with $r$ such that $log_2(r) < log_2(n) - 3$ and $r' < r$, trivially fits the requirements (for any $n$ large enough to provide a secure encryption).
Nov
16
revised Homomorphic (encrypted) comparison to an integer
added 401 characters in body
Nov
16
comment Homomorphic (encrypted) comparison to an integer
@D.W.: sorry, I should have mentioned: for the sake of clarity, I have added the detailed explanations in an edit to the answer itself. As you can see, $l()$ is simply the bit length ($log_2()$). This does define very precisely the range of $r$ (and therefore $r'$) to avoid modulo overflow. The requirement of (statistical) uniformity of the distribution of $r(y-x)-r'$ can be trivially achieved for small enough values of $(y-x)$ relative to $r$ and $r'$, which is a indeed a big limitation, but still something...
Nov
16
revised Homomorphic (encrypted) comparison to an integer
added 40 characters in body
Nov
15
revised Homomorphic (encrypted) comparison to an integer
added 1 characters in body