HoneyBadger BFT


#1

Hi,

I recently discovered this BFT protocol called HoneyBadger which is:

  • asynchronous
  • Has a high throughput (I know this isn’t the sole important metric)
    • even if the tests were made in a permissioned mode, they achieved 10K+ TPS with nodes distributed in AWS regions (with bit batches of txs)
  • Use treshold encryption to hide the elected leader
    • Looks similar to the secret-key computation of Uroboros used to not predict the eligibility of a stakeholder

The authors of this protocol sayed that it could be used in a permissionless manner, with a PoS/DPoS election process. I don’t know how this permissioned “mode” could be easily applied.

Being non-synchronous seems to be in favor of a worldwide permissionless network, and could avoid to do some network delay assumption.
Is the big batch of tx too big for small bandwidth ?

I didn’t saw any mention to this async protocol in the Uroboros papers, did you studied it ? If so, do you have any opinion ?

Thank you


#2

here is an actual link to the paper : https://eprint.iacr.org/2016/199.pdf


#3

If we can get the consensus guys in an AMA, it would be good to know how this compares to Praos :slight_smile: I’ll see what I can do.


#4

shamelessly rip off… if this works or provides some value over Ouroboros or complements it… why not? It’s not like the brainpower to integrate anything into Ada isn’t there at IOHK :slight_smile:


#5

Just noticed that the HoneyBadgerBFT was made with Andrew Miller (from University of Illinois) who guess what… also worked on the Non-Interactive-PoW paper (https://eprint.iacr.org/2017/963.pdf) with IOHK.
World is small.