[Preliminary Research] User Ownership of Channel Funds
Today, the blast radius of a malicious sequencer set is disastrous. A malicious set means sequencers holding more than min(transfer threshold, config threshold). The minimum is what counts, because a config quorum can change the set of sequencer keys, or the transfer threshold and then transfer. With that power, the sequencers can update the configuration and/or withdraw all the bridge funds atomically in a single Mantle transaction, in a finger snap.
Most Ethereum L2s have a similar issue. A committee can change the contract that validates state transitions, including withdrawals. So the committee can change the rules defining a valid withdrawal. What puts them in a better position is a delay before the committee can touch the funds or update the rules. This delay gives users enough time to exit with their funds if they disagree with the new rules.
In our model, this escape is impossible. Inside the zone, funds are held under users' keys. But on the L1, bridge funds move on the sequencers' signature alone, with no verification of the state transition. Users cannot prove to the L1 that they own anything. The L1 only sees a channel note controlled by a key that the sequencers can modify quickly and without notice. Even with a delay, users would have no way to get their funds out.
The full fix is to give users back power over their funds, with the L1 guaranteeing their ownership. This needs three things:
This is essentially a SNARK contract or what we initially defined as Native Zones where the channel creator sets the verification key. The L1 does not need to encode the zone rules. It needs one unique validation that proves funds movements the same way for all zones, plus a verification key to validate the associated zone state transitions. It can also be optimistic: a proof is required only when a transition is challenged. We lose some finality in exchange.
The cost is high. Supporting ZK proofs for any zone state transition is a different world, and creating a channel becomes much harder. It also changes how a zone like LEZ manages notes by forcing the adoption of our (not yet defined) channel transfer circuit. Channel notes lag behind the LEZ ledger, so LEZ proving would have to show the link between their state transition and our channel transfer circuit somehow.
Withdrawals and channel transfers are delayed. Honest sequencers keep the L1 channel notes in line with their internal state on a regular schedule, so each channel note mirrors what its owner holds.
If the sequencers stop maintaining this mapping on schedule, channel users vote (not sure yet how) to dissolve the channel. When enough channel note value votes (2/3 for example it’s not decided or it could be decided in the channel config), the channel stops and the channel notes move back to the main ledger. The L1 already enforces the key on each channel note, so each owner gets their note back.
Any deviation from the schedule counts as misbehaviour. Ledgers must not drift too far apart. The time scale, like for Ethereum layers 2, would be from a few days to a few weeks, and news of misbehaviour need to spreads fast, so users have time to coordinate.
The L1 does not enforce the mapping. Only channel users do, and some of them will be offline.
The threshold cannot be tuned well. If it is high, a theft of a small part of the funds never reaches it, because the scheme assumes that users who are not robbed will vote to stop the theft. If it is low, dissolving becomes cheap, and anyone can grief the channel.