EIP-7441 Switching proposer selection to Whisk (Upgrade block proposer election to Whisk)
No call decision · Consensus & staking · EIP status Stagnant · created 2023-09-01
A proposal to change proposer selection to the Whisk protocol, which keeps the validator who will produce the next block secret until the block is actually revealed.
Why it is needed
Currently, the beacon chain selects the block proposers for the next 32 slots at the start of every epoch and publishes the result to everyone. Once it is known who will produce a block in which slot, an attacker can target each proposer's network in sequence with denial-of-service (DoS) attacks. If a proposer is knocked offline by such an attack, the corresponding slot goes empty or gets filled late, raising the risk that Ethereum as a whole slows down or stalls. This EIP aims to prevent this kind of pre-emptive targeted attack by hiding the proposer's identity until the block is actually produced.
What changes
Today
- At the start of every epoch, the identities of the next 32 block proposers are revealed to everyone
- An attacker can see the published order and launch DoS attacks against each proposer in sequence
- Proposer selection is determined by get_beacon_proposer_index, and the result is exposed as is
After EIP-7441
- Validators register a tracker and a commitment at their first proposal after the fork
- The next proposer list is determined in shuffled form through RANDAO randomness and private shuffling
- When producing a block, the proposer proves their selection solely with an opening proof, and their identity remains hidden until then
Who is affected
- Validator: A new tracker and commitment must be registered at the first block proposal after the fork.
- Block-building participants such as PBS builders: Unless the proposer discloses it themselves, blocks cannot be prepared ahead of time targeting a specific fee recipient address.
- Block explorer and tooling developers: Even when a slot is missed, it can no longer be directly attributed to a specific validator index.
- Ordinary users: The risk of block production being delayed by DoS attacks targeting proposers is also indirectly reduced.
In depth
Replacing the selection mechanism
Fields are added to BeaconState to track validator trackers, commitments, and candidate selection across two rounds. select_whisk_candidate_trackers computes the next candidate vector from the validator set, and select_whisk_proposer_trackers computes the next proposer vector from the current candidates. process_whisk_updates is added to the epoch processing logic, and process_whisk_opening_proof verifies that the block proposer knows the tracker selected for that slot. process_block_header is changed to no longer confirm the proposer via get_beacon_proposer_index, checking only whether the opening proof is valid. Fields are also added to BeaconBlockBody to hold the opening proof, the shuffled trackers and their proofs, and the tracker registration proof.
Fields per validator: tracker and commitment
Whisk requires one tracker (rG, krG) and one unique commitment kG per validator, and both are updated only once, at that validator's first proposal after the fork. The tracker is registered with a randomized base (rG, krG), a device intended to make it difficult for an adversary to trace a specific tracker as it passes through the shuffling gates. However, it is stated that this security can weaken if the number of honest shufflers is small.
Identity binding method
Each tracker must be bound to a single validator's identity; otherwise multiple participants could claim the same proposer slot, allowing a proposer to sell their slot, or creating fork-choice problems when competing blocks appear. Whisk binds identity by storing a commitment to the tracker's secret kG in the validator record, and this storage approach also guarantees the uniqueness of k. An alternative of matching a hash prefix of hash(kG) to the validator index was considered, but it requires brute-forcing k, which creates a barrier to entry for participants with limited computing resources. Setting k = hash(nonce + pubkey) was also considered, but it has the problem that a proposer must reveal k when proposing repeatedly in adjacent shuffling phases, which de-anonymizes them.
Alternatives considered
Non-cryptographic secret non-single leader election was also considered, but it was judged unsuitable at the time of writing because it complicates fork choice and could be exposed to MEV time-buying attacks. Network anonymization protocols such as Dandelion or Dandelion++ increase participant privacy but were deemed insufficient for Ethereum's purposes. SASSAFRAS is a simpler SSLE alternative on the consensus side, but because it depends on a network anonymity layer, its tradeoffs were assessed as not fitting Ethereum's overall threat model better than Whisk does.
Anonymity set size
Whisk's anonymity set is the 8,192 candidate trackers not selected as proposer. This number corresponds to a smaller number of p2p nodes; applying the Pareto assumption that '20% of nodes run 80% of validators' puts the average anonymity set at around 2,108 nodes. It is noted that enlarging the candidate pool further could make the shuffling strategy itself unsafe, while conversely shuffling more trackers per round increases the cost of the zero-knowledge (ZK) proof.
RANDAO biasing risk
Because Whisk uses RANDAO for both candidate selection and proposer selection, it can be exposed to attacks where a malicious proposer biases the RANDAO value. Spreading selection events across the entire shuffling period could keep security at its current level, but doing so would make the protocol more complex, and it is noted that even the current level of security is not ideal to begin with.
Scope of application and status
There are no changes to the Execution Layer; all changes are backwards-incompatible changes to the Consensus Layer's block validation rules, so a hard fork is required. Builders participating in PBS will not know the next proposer's validator index unless the proposer discloses it themselves in advance, so they cannot target a specific fee recipient ahead of time, and block explorers and tooling likewise cannot attribute a missed slot to a specific validator index. This EIP's status is Stagnant, meaning it is no longer actively progressing, and it is uncertain whether it will be adopted in its current form.