[Cryptography] Advances in homomorphic encryption
Bear
bear at sonic.net
Sun Jan 12 12:21:43 EST 2014
On Sat, 2014-01-11 at 19:16 -0800, Christian Huitema wrote:
> > cryptdb uses homomorphic encryption (as part of a larger system that leaks
> more information that "pure" HE would, but that is an
> > irrelevant tangent).
>
> Homomorphic encryption enables computations using additions and
> multiplications, but I wonder about the domain of application. Whether you
> consider projection and joint in a SQL database, or map/reduce, we need to
> perform comparisons. For example, we may want to find all the sales that
> happened last week and compute the average profit. That requires accessing
> the date field and verifying that it falls within last week's range. I
> can't see how to do that if the date is encrypted in a way that provides
> semantic security.
>
> -- Christian Huitema
This is true. Proof of "less than" is not simple. It winds up needing
to take the form of a zero-knowlege proof.
Of course there are other possibilities like discrete integral fields
for year, month, day, hour, minute, that could be checked quickly
in succession -- first by stepping through years looking for equality,
then months, then days, then hours, then minutes. You could figure
out whether a particular transaction is within a given date range by
making a relatively quick and simple series of equality checks.
Most other things you'd be interested in making comparisons about can
also be reduced to discrete integral fields.
Bear
More information about the cryptography
mailing list