Re: [Belenios-discuss] Helios-C and the Majority Judgment

[Julien Moutinho <julm+inria AT autogeree.net>]

Hi Stéphane,

Le jeu. 02 mai 2019 à 08:42:24 +0200, Stéphane Glondu a écrit :
> Did you see the Belenios specification?
> I believe it is the most practical guide to implementation.
>   http://www.belenios.org/specification.pdf
Sure, the other papers I'm reading are:
- « A generic construction for voting correctness at minimum cost - 
Application to Helios »
  https://eprint.iacr.org/2013/177.pdf
- « Implementation of the Belenios Receipt-Free+ Online Voting Protocol »
  (whose « Background » section was enlightening to me)
  http://www.dgalindo.es/mscprojects/thomas.pdf
- « Belenios: a simple private and verifiable electronic voting system »
  https://hal.inria.fr/hal-02066930/document
And above all the source code of Belenios.

> About Majority Judgement: it can be encoded with the homomorphic tally
> of (current) Belenios by encoding each "choice" as a "question" and each
> "preference" as a "choice. This is reasonnable as long as there are not
> too many "choices x preferences" (let's say, < 50). Indeed, the size of
> the ballot is proportional to this product.
Yes, thanks for mentioning this ballot size drawback I didn't know,
The scale of grades (what you called "preferences",
but preferences suggests a comparison between the choices,
while grades are meant to be absolute judgments of the choices)
usually spans 5, 6 or 7 grades.
Therefore a 10 candidates election would overflow this order of magnitude.

So, because it's a concern for me to make everything possible to encourage 
pluralism,
I'd prefer an implementation working well with that order of magnitude.
Correct me if I'm wrong but I don't think that implementing
the Majority Judgment properly would reduce the ballot size that much,
maybe passing from (#choices * (2 * #grades + 1)) proofs
to (#choices * #grades) proofs.

So far I was just unsatisfied by this encoding because it's a bit contrived.
I was thinking about replacing directly the actual individual proofs (0 or 1)
by interval proofs for each grade given to each choice.
The overall proof already being a model of how to do such an interval proof.
By doing so I would also remove the overall proof,
and the blank proof which is complex and useless for the Majority Judgment.

> However, we are also implementing mixnets (in progress), which allow to
> recover each individual ballot after tally. The main motivation is to
> support Condorcet-style voting, but they also allow a simpler encoding
> of Majority Judgement without the size issue mentioned above. I was
> considering adding support of Majority Judgement to Belenios this way,
> what do you think?
Ok. I don't know mixnets (I just have some vague idea of how Tor works)
but AFAICS switching individual proofs to grade proofs is all what is needed
to implement the Majority Judgment properly.

A cryptosystem like Helios-C, though allegedly simpler than others
(I've never studied any other cryptosystem so I don't have an opinion on that)
still is a very intimidating piece of math and software,
thus I'd very like to keep it to the simplest.

And regarding that concern, supporting only the Majority Judgment simplifies
things quite a bit, which I agree since, as a Mieux Voter's member,
I'm convinced that the Majority Judgment is superior
to any known method of voting, in theory and in practice,
therefore I don't feel the need to support other methods of voting.

By the way, for those who don't know the Majority Judgment,
a full scientific textbook on it is there:
« Majority Judgment: Measuring, Ranking, and Electing »
http://booksdescr.org/item/index.php?md5=BF67AA4298C1CE7633187546AA53E01D
And a french comicbook vulgarizing it is there:
https://www.lechoixcommun.fr/articles/Vous_reprendrez_bien_un_peu_de_democratie-2.html
Other papers I've used myself are listed here:
https://hackage.haskell.org/package/hjugement

> Glad to hear you took the free software path.
Well it wouldn't be fair otherwise :)
First, I'm hugely inspired by Belenios which is free software under copyleft,
second, a voting software must be verifiable by anyone more than any other 
software,
and last but not least, the socialized salary I have from the Social security 
(the CAF actually)
is also here to enable that kind of work for the commons.

> Do you plan some kind of interoperability with Belenios?
> An alternative, public and independent, verifier for Belenios would be 
> great.
It would be great indeed. I'm trying to stick to the current specs everywhere 
I can,
but AFAICS I won't be able to if I want to support the Majority Judgment 
properly
and only the Majority Judgment. This said, I think most of the data 
structures and functions
required to implement a tool interoperable with Belenios would still be 
implemented,
and thus reusable for who has the energy to maintain the full 
interoperability.

Regards,
Julien

Attachment: signature.asc
Description: PGP signature