Avoiding Signing Restarts on Witness Unreachability: Is it implemented or evaluated?

18 views
Skip to first unread message

Leander Jehl

unread,
Oct 5, 2018, 5:44:57 AM10/5/18
to Cothority Discussion
Hi,

I'm wondering if the variation of the cosigning protocol, used to tolerate witnesses that become unavailable after commitment is implemented? 
(Section IV D in the Honest or Bust paper)
Did someone evaluate the impact of a small share of unavailable witnesses on the performance, especially on verification latency?
If not, do you have any expectations regarding this?

Thanks,

Leander


Gasser Linus

unread,
Oct 8, 2018, 2:33:42 AM10/8/18
to Leander Jehl, Cothority Discussion
Dear Leander,

Thank you very much for your interest in that paper. All our implementation with regard to collective signing is here:


We did never get around to handle a witness failing _after_ the commit, only before. What we currently do, if that happens, is to restart the protocol. As we use the CoSi protocol with our industrial partners, the most common fail-case is a node that is down, or too busy. So restarting is fine.

We have, however, some work in progress here:


Which is currently completely unsupported and a mere student project. The idea is to replace the schnorr-signatures with BLS signatures, then we don’t need two rounds, and don’t have the problem of failing nodes after the 1st round.

What is your use-case?

Sincerely,

Linus

--
You received this message because you are subscribed to the Google Groups "Cothority Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cothority+...@googlegroups.com.
To post to this group, send email to coth...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cothority/7dfb28c3-425c-43eb-818b-a799b4d877c6%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

signature.asc

Leander Jehl

unread,
Oct 8, 2018, 3:50:14 AM10/8/18
to Cothority Discussion
Thanks Linus,

I will have a look at the bls project.
My interest is mainly academic. I'm working on failure detection in a byzantine environment and want to understand the requirements on failure detection necessary for co-signing. 

Regards,

Leander

 
Reply all
Reply to author
Forward
0 new messages