[Cryptography] Super-computer project wanted

[Peter Fairbrother]

On 13/07/15 18:00, Dave Horsfall wrote:
> I'm doing an online course in programming super-computers (think:
> massively parallel teraflops!) and we are invited to submit a proposal for
> a sort of post-grad project (there is no guarantee that it will be
> accepted, as machine time is scarce).

> So, is there anything that could benefit from a few parallel teraflops
> here and there?
>

I have a dream ... for a single-database PIR-based anonymous messaging 
service, based on either a variant of Cachin, Micali and Stadler's [1] 
or Gentry and Ramazan's [2] PIR schemes.

Both schemes however require a huge amount of computation, in the form 
of modular exponentiations, by the (distributed) servers.

This would not be a once-only effort, it would have to be a continuous 
service, so it might not be in accord with your project requirements.

However, any work designs for both software and inexpensive hardware 
systems which can do vast numbers of modexps, would be interesting; and 
might lead to a practical computationally-untraceable messaging service, 
for at least tweet-sized messages.



[1] www.zurich.ibm.com/~cca/papers/cpir.pdf . The variant is 
unpublished, but I presented it in a rump session at PET2003.

[2] www.cs.umd.edu/~gasarch/TOPICS/pir/logn.pdf

-- Peter Fairbrother