Over the past couple of days, I’ve been digging through Dusk’s consensus documents—and the more I read, the more I feel a particular detail really hits the mark: in many PoS chains, the validator nodes both propose blocks and verify blocks. In other words, one person serves as both the proposer and the judge. This design, by itself, hides a vulnerability—when someone wants to do harm, the path is especially short.

//
Dusk’s approach is to split these two responsibilities completely.
Its consensus is called SBA (Segregated Byzantine Agreement). The core idea is to cleanly separate the roles of “block production” and “block witnessing” at the design level. One group of nodes specializes in proposing candidate blocks, while another group specializes in validating, voting, and deciding which blocks should be added to the chain.

//
These two groups have different rights and responsibilities, and different identities. One group only proposes, the other only reviews. No one can be both player and referee at the same time.

//
Here’s another engineering detail that’s pretty interesting.
During the verification phase, a large number of nodes each sign and cast votes. If every signature is broadcast and verified separately, the network overhead rises with the number of nodes. The idea behind BLS aggregated signatures is to compress thousands of independent signatures into a single short aggregate result. Other nodes then only need to verify that one aggregate signature—which is equivalent to verifying everyone’s votes at once. As a result, verification cost drops dramatically, and consensus efficiency naturally improves.

//
Where the real value lies in splitting the two stages is that it raises the threshold for collusion.
If block production and block verification are done by the same party, then malicious behavior only requires controlling a single role. After the split, anyone who really wants to manipulate outcomes must simultaneously subvert both the production side and the witnessing side—two groups of people that are completely independent. How those groups are selected, how identities are defined, and what incentives are provided are all designed separately. This directly increases the cost of collusion and the risk of exposure.

The logic of separating responsibilities isn’t new in traditional governance, but Dusk has embedded it into the underlying mechanisms of the consensus layer—not just something people talk about. This is probably what SBA aims to solve.

@Dusk $DUSK #dusk

What is the core design of the SBA consensus?