Trustless bridging needs to handle a special kind of equivocation: fork equivocations.
Meaning the validators of one chain finalize one fork in the public consensus protocol of that chain, but collude to privately finalize a different fork and submit that private fork to a trustless bridge. The light-client of the trustless bridge will accept it as valid since it is "correctly" signed by a supermajority of validators.
This is highly unlikely since it requires a super-majority of honest validators to collude to cheat the bridge. But we still need some economic disincentive to doing this, to be able to keep that "highly unlikely" rating.
In the case of BEEFY bridge consensus, that disincentive is validator slashing if proven they engaged in such behaviors:
So Polkadot supports the mechanism to slash in case of misbehavior.
Required
But actual equivocation detection can only be detected by bridge relayers (or some other off-chain component watching the bridge). On Polkadot, the validators only publicly gossip the "good" commitments, and the "bad"/equivocated commitments are built in private, in secret and directly submitted to the bridge.
The only way to detect this is after the fact, when noticing a discrepancy between the finalized hash on Polkadot vs the finalized hash imported by the bridge BEEFY light client.
Once detecting this, an honest relayer can build an equivocation proof and submit it to Polkadot for a reward. This would result in the validators involved getting their stake slashed and thus provides good economic security against this type of attack.
Snowbridge relayers should implement this detection and reporting mechanism.
Description
Trustless bridging needs to handle a special kind of equivocation: fork equivocations.
Meaning the validators of one chain finalize one fork in the public consensus protocol of that chain, but collude to privately finalize a different fork and submit that private fork to a trustless bridge. The light-client of the trustless bridge will accept it as valid since it is "correctly" signed by a supermajority of validators.
This is highly unlikely since it requires a super-majority of honest validators to collude to cheat the bridge. But we still need some economic disincentive to doing this, to be able to keep that "highly unlikely" rating.
In the case of BEEFY bridge consensus, that disincentive is validator slashing if proven they engaged in such behaviors:
So Polkadot supports the mechanism to slash in case of misbehavior.
Required
But actual equivocation detection can only be detected by bridge relayers (or some other off-chain component watching the bridge). On Polkadot, the validators only publicly gossip the "good" commitments, and the "bad"/equivocated commitments are built in private, in secret and directly submitted to the bridge.
The only way to detect this is after the fact, when noticing a discrepancy between the finalized hash on Polkadot vs the finalized hash imported by the bridge BEEFY light client.
Once detecting this, an honest relayer can build an equivocation proof and submit it to Polkadot for a reward. This would result in the validators involved getting their stake slashed and thus provides good economic security against this type of attack.
Snowbridge relayers should implement this detection and reporting mechanism.