noticed that a given FP will be slashed only when signing the canonical (i.e. indexed) block, and an alternative (fork) block at the same height.
This implies that a malicious FP could sign and submit any number of non-canonical block signatures for a given height. As long as he doesn’t sign and submit the canonical block signature as well, he won’t be slashed.
This is because slashing currently depends on a FP signing both, an offending / fork block, AND the canonical (i.e. indexed) block.
Perhaps the code / logic can be changed here. If a given FP signs two different blocks at any given height (independently of if one is the canonical or not), he’s slashed.
So, slash when signing both the canonical and other block, and also when signing two other blocks.
This seems easy to fix: If Evidence already exists in the store for this FP at that height, this is a double sign.
Update: Just noticed that the detection of double signing of forks is slightly below in the code
Summary of Bug
Taking a look at the slashing logic in
https://github.com/babylonchain/babylon/blob/fe247dd38a83ccbf940dabed5dafaa69698b9eb0/x/finality/keeper/msg_server.go#L132-L140
noticed that a given FP will be slashed only when signing the canonical (i.e. indexed) block, and an alternative (fork) block at the same height.
This implies that a malicious FP could sign and submit any number of non-canonical block signatures for a given height. As long as he doesn’t sign and submit the canonical block signature as well, he won’t be slashed.
This is because slashing currently depends on a FP signing both, an offending / fork block, AND the canonical (i.e. indexed) block.
Perhaps the code / logic can be changed here. If a given FP signs two different blocks at any given height (independently of if one is the canonical or not), he’s slashed.
So, slash when signing both the canonical and other block, and also when signing two other blocks.
This seems easy to fix: If
Evidencealready exists in the store for this FP at that height, this is a double sign.Update: Just noticed that the detection of double signing of forks is slightly below in the code
https://github.com/babylonchain/babylon/blob/fe247dd38a83ccbf940dabed5dafaa69698b9eb0/x/finality/keeper/msg_server.go#L153-L172
The bug is then that, since neither fork is the canonical block, this code is never reached (the
ifabove always returns without error).Version
Latest.
Steps to Reproduce
N/A.