Re: [viff-devel] A java implementation of VIFF
>> The program still has one drawback, in that the different players >> communicate over Java queues, and these queues cannot communicate through >> sockets. Therefore it is not a complete MPC implementation. > > This part sounds a bit odd to be honest. I think what Tord means is that the players currently exchange messages directly via a ConcurrentLinkedQueue from the java.util.concurrent package. This of course only works when all players live in the same Java runtime on the same machine. Best regards, Thomas ___ viff-devel mailing list (http://viff.dk/) viff-devel@viff.dk http://lists.viff.dk/listinfo.cgi/viff-devel-viff.dk
Re: [viff-devel] A java implementation of VIFF
Citat af Tord Ingolf Reistad : Hello, As some of you might know I have not programmed that much in VIFF because I am more familiar with C++ and Java than Python as a programming language. Therefore I have now created a Java rival to VIFF. Taking all the good design principles from VIFF and adding some things that VIFF cannot offer such as multi-threading. I have taken a look at the code. It's a bit hard to follow without further explanation, but I think I understand the basics. At least it seems like I wouldn't have a hard time writing small MPC programs using your code (MPCTestAlgorithm certainly looks quite simple). As Thomas has suggested, yes it would be fun to see benchmark comparisons with VIFF. When VIFF was designed, I don't think it was the idea to skip multithreading with the knowledge that it would be worse or slower. A thorough comparison of both schemes could help show if there are any clear advantages to either of them. I hereby volunteer someone else to do all the work :) The program still has one drawback, in that the different players communicate over Java queues, and these queues cannot communicate through sockets. Therefore it is not a complete MPC implementation. This part sounds a bit odd to be honest. The whole program was programmed and tested in just 34 hours (including 6 hours sleep). It contains 2115 lines of code and 32 classes Ouch :) ___ viff-devel mailing list (http://viff.dk/) viff-devel@viff.dk http://lists.viff.dk/listinfo.cgi/viff-devel-viff.dk
Re: [viff-devel] A java implementation of VIFF
Nice work! It would be interesting to see how it performs compared to VIFF. Best regards, Thomas ___ viff-devel mailing list (http://viff.dk/) viff-devel@viff.dk http://lists.viff.dk/listinfo.cgi/viff-devel-viff.dk
[viff-devel] A java implementation of VIFF
Hello, As some of you might know I have not programmed that much in VIFF because I am more familiar with C++ and Java than Python as a programming language. Therefore I have now created a Java rival to VIFF. Taking all the good design principles from VIFF and adding some things that VIFF cannot offer such as multi-threading. The program still has one drawback, in that the different players communicate over Java queues, and these queues cannot communicate through sockets. Therefore it is not a complete MPC implementation. The following things have been implemented: Sharing, Negation and Revealing of Shares. Adding and Multiplication between Shares and/or Field elements. The shares are computed using Shamir's secret sharing scheme. The algorithm, sharing scheme and communication parts have all been separated so it can be modified easily. The whole program was programmed and tested in just 34 hours (including 6 hours sleep). It contains 2115 lines of code and 32 classes, some parts where heavily influenced by previous programming attempts. Programmed in NetBeans. For anyone interested in the code I have dropped of a zip of the code at http://drop.io/tordpaper/ Tord ___ viff-devel mailing list (http://viff.dk/) viff-devel@viff.dk http://lists.viff.dk/listinfo.cgi/viff-devel-viff.dk