RadixDLT claim: consensus with 99% dishonest actors

stumbled across this article from August 10, 2018

can’t say I understood how they do it… but the claims are

  • permissionless consensus system
  • very high scalability
  • asynchronous byzantine fault detection
  • consensus even with 99% dishonest actors (???)

I thought the theoretical limit for ABFT was N/3-1 dishonest nodes maximum?
So is this complete BS what they are claiming? Or if not, could MaidSafe learn sth
from them?

curious about your thoughts

4 Likes

On the surface it sounds BS. How many nodes are in the group doing the consensus?

This looks like a blockchain ledger project. I didn’t read far enough to see if their ledger is a blockchain.

The consensus is across the whole network of nodes which all seem to keep a record of transactions. But then there are subnets too.

I cannot from the reading see how a coordinated group of dishonest actors cannot take over their network if there are enough dishonest (51%) since they are using time stamps as a source of validation against bad actors. Once the coordinated attack has 51% then they can all lie and claim the honest nodes are the liars.

But if it is not coordinated then the timestamps and carrying information from the past can detect incorrect claims of individual nodes. Thus for an uncoordinated attack you could in effect have just one valid node and any inquiries can in theory detect invalid responses due to the crypto history. But a coordinated attack could rewrite history, or just insert their own version of it and claim the others are liars.

2 Likes

Its a new consensus, which they’ve termed ‘Tempo’

2 Likes

Probably bullshit. Just another scam I say.