24 KiB
Ethereum 2.0 Phase 0 -- Honest Validator
NOTICE: This document is a work-in-progress for researchers and implementers. This is an accompanying document to Ethereum 2.0 Phase 0 -- The Beacon Chain that describes the expected actions of a "validator" participating in the Ethereum 2.0 protocol.
Table of Contents
- Ethereum 2.0 Phase 0 -- Honest Validator
Introduction
This document represents the expected behavior of an "honest validator" with respect to Phase 0 of the Ethereum 2.0 protocol. This document does not distinguish between a "node" and a "validator client". The separation of concerns between these (potentially) two pieces of software is left as a design decision that is outside of scope.
A validator is an entity that participates in the consensus of the Ethereum 2.0 protocol. This is an optional role for users in which they can post ETH as collateral to seek financial returns in exchange for building and securing the protocol. This is similar to proof of work networks in which a miner provides collateral in the form of hardware/hash-power to seek returns in exchange for building and securing the protocol.
Prerequisites
All terminology, constants, functions, and protocol mechanics defined in the Phase 0 -- The Beacon Chain doc are requisite for this document and used throughout. Please see the Phase 0 doc before continuing and use as a reference throughout.
Constants
Misc
Name | Value | Unit | Duration |
---|---|---|---|
ETH1_FOLLOW_DISTANCE |
2**10 (= 1,024) |
blocks | ~4 hours |
Becoming a validator
Initialization
A validator must initialize many parameters locally before submitting a deposit and joining the validator registry.
BLS public key
Validator public keys are G1 points on the BLS12-381 curve. A private key, privkey
, must be securely generated along with the resultant pubkey
. This privkey
must be "hot", that is, constantly available to sign data throughout the lifetime of the validator.
BLS withdrawal key
A secondary withdrawal private key, withdrawal_privkey
, must also be securely generated along with the resultant withdrawal_pubkey
. This withdrawal_privkey
does not have to be available for signing during the normal lifetime of a validator and can live in "cold storage".
The validator constructs their withdrawal_credentials
through the following:
- Set
withdrawal_credentials[:1] == BLS_WITHDRAWAL_PREFIX_BYTE
. - Set
withdrawal_credentials[1:] == hash(withdrawal_pubkey)[1:]
.
RANDAO commitment
A validator's RANDAO commitment is the outermost layer of a 32-byte hash-onion. To create this commitment, perform the following steps:
- Randomly generate a 32-byte
randao_seed
. - Store this
randao_seed
in a secure location. - Calculate
randao_commitment = repeat_hash(randao_seed, n)
wheren
is large enough such that within the lifetime of the validator, the validator will not propose more thann
beacon chain blocks.
Assuming >= 100k validators
, on average a validator will have an opportunity to reveal once every >= 600k seconds
, so <= 50 times per year
. At this estimate, n == 5000
would last a century. If this value is poorly configured and a validator runs out of layers of to reveal, the validator can no longer propose beacon blocks and should exit.
Note: A validator must be able to reveal the next layer deep from their current commitment at any time. There are many strategies that trade off space and computation to be able to provide this reveal. At one end of this trade-off, a validator might only store their randao_seed
and repeat the repeat_hash
calculation on the fly to re-calculate the layer n-1
for the reveal. On the other end of this trade-off, a validator might store all layers of the hash-onion and not have to perform any calculations to retrieve the layer n-1
. A more sensible strategy might be to store every m
th layer as cached references to recalculate the intermittent layers as needed.
Custody commitment
A validator's custody commitment is the outermost layer of a 32-byte hash-onion. To create this commitment, perform the following steps:
- Randomly generate a 32-byte
custody_seed
. - Store this
custody_seed
in a secure location. - Calculate
custody_commitment = repeat_hash(custody_seed, n)
wheren
is large enough such that within the lifetime of the validator, the validator will not attest to more thann
beacon chain blocks.
Assuming a validator changes their custody_seed
with frequency >= 1 week
, the validator changes their seed approximately <= 50 times per year
. At this estimate, n == 5000
would last a century. If this value is poorly configured and a validator runs out of layers of to reveal, the validator can no longer update their custody_commitment
and should exit.
See above note on hash-onion caching strategies in RANDAO commitment.
Note: although this commitment is being committed to and stored in phase 0, it will not be used until phase 1.
Submit deposit
In phase 0, all incoming validator deposits originate from the Ethereum 1.0 PoW chain. Deposits are made to the deposit contract located at DEPOSIT_CONTRACT_ADDRESS
.
To submit a deposit:
- Pack the validator's initialization parameters into
deposit_input
, aDepositInput
object. - Set
deposit_input.proof_of_possession = EMPTY_SIGNATURE
. - Let
proof_of_possession
be the result ofbls_sign
of thehash_tree_root(deposit_input)
withdomain=DOMAIN_DEPOSIT
. - Set
deposit_input.proof_of_possession = proof_of_possession
. - Let
amount
be the amount in Gwei to be deposited by the validator whereMIN_DEPOSIT_AMOUNT <= amount <= MAX_DEPOSIT_AMOUNT
. - Send a transaction on the Ethereum 1.0 chain to
DEPOSIT_CONTRACT_ADDRESS
executingdeposit
along withdeposit_input
as the singularbytes
input along with a depositamount
in Gwei.
Note: Multiple deposits can be made by the same validator (same initialization params). A singular Validator
will be added to state.validator_registry
with each deposit amount being added to the validator's balance. A validator can only be activated when total deposits meet or exceed
Process deposit
Deposits cannot be processed into the beacon chain until the eth1.0 block in which they were deposited or any of its ancestors is added to the beacon chain state.eth1_data
. This takes a minimum of ETH1_FOLLOW_DISTANCE
eth1.0 blocks (~4 hours) plus ETH1_DATA_VOTING_PERIOD
slots (~1.7 hours). Once the necessary eth1.0 data is added, the deposit will normally be added to a beacon chain block and processed into the state.validator_registry
within an epoch or two. The validator is then in a queue to be activated.
Validator index
Once a validator has been processed and added to the state's validator_registry
, the validator's validator_index
is defined by the index into the registry at which the ValidatorRecord
contains the pubkey
specified in the validator's deposit. This validator_index
is used throughout the specification to dictate validator roles and responsibilities at any point and should be stored locally.
Activation
In normal operation, the validator is quickly activated at which point the validator is added to the shuffling and begins validation after an additional ENTRY_EXIT_DELAY
slots.
The function is_active_validator
can be used to check if a validator is active at a given slot. Usage is as follows:
validator = state.validator_registry[validator_index]
is_active = is_active_validator(validator, slot)
Once a validator is active, the validator is assigned responsibilities until exited.
Note: There is a maximum validator churn per finalized epoch so the delay until activation is variable depending upon finality, total active validator balance, and the number of validators in the queue to be activated.
Beacon chain responsibilities
A validator has two primary responsibilities to the beacon chain -- proposing blocks and creating attestations. Proposals happen infrequently, whereas attestations should be created once per epoch.
Block proposal
A validator is expected to propose a BeaconBlock
at the beginning of any slot during which get_beacon_proposer_index(state, slot)
returns the validator's validator_index
. To propose, the validator selects the BeaconBlock
, parent
, that in their view of the fork choice is the head of the chain during slot
. The validator is to create, sign, and broadcast a block
that is a child of parent
that creates a valid beacon chain state transition.
Block header
Slot
Set block.slot = slot
where slot
is the current slot at which the validator has been selected to propose. The parent
selected must satisfy that parent.slot < block.slot
.
Note: there might be "skipped" slots between the parent
and block
. These skipped slots are processed in the state transition function without per-block processing.
Parent root
Set block.parent_root = hash_tree_root(parent)
.
State root
Set block.state_root = hash_tree_root(state)
of the resulting state
of the parent -> block
state transition.
Note: To calculate state_root
, the validator should first run the state transition function on an unsigned block
containing a stub for the state_root
. It is useful to be able to run a state transition function that does not validate signatures for this purpose.
Randao reveal
Set block.randao_reveal
to the n
th layer deep reveal from the validator's current randao_commitment
where n = validator.randao_layers + 1
. block.randao_reveal
should satisfy repeat_hash(block.randao_reveal, validator.randao_layers + 1) == validator.randao_commitment
.
Eth1 Data
block.eth1_data
is a mechanism used by block proposers vote on a recent Ethereum 1.0 block hash and an associated deposit root found in the Ethereum 1.0 deposit contract. When consensus is formed, state.latest_eth1_data
is updated, and validator deposits up to this root can be processed.
- Let
D
be the set ofEth1DataVote
objectsvote
instate.eth1_data_votes
where:vote.eth1_data.block_hash
is the hash of an eth1.0 block that is (i) part of the canonical chain, (ii) >=ETH1_FOLLOW_DISTANCE
blocks behind the head, and (iii) newer thanstate.latest_eth1_data.block_data
.vote.eth1_data.deposit_root
is the deposit root of the eth1.0 deposit contract at the block defined byvote.eth1_data.block_hash
.
- If
D
is empty:- Let
block_hash
be the block hash of theETH1_FOLLOW_DISTANCE
th ancestor of the head of the canonical eth1.0 chain. - Let
deposit_root
be the deposit root of the eth1.0 deposit contract at the block defined byblock_hash
.
- Let
- If
D
is nonempty:- Let
best_vote
be the member ofD
that has the highestvote.eth1_data.vote_count
, breaking ties by favoring block hashes with higher associated block height. - Let
block_hash = best_vote.eth1_data.block_hash
. - Let
deposit_root = best_vote.eth1_data.deposit_root
.
- Let
- Set
block.eth1_data = Eth1Data(deposit_root=deposit_root, block_hash=block_hash)
.
Signature
Set block.signature = signed_proposal_data
where signed_proposal_data
is defined as:
proposal_data = ProposalSignedData(
slot=slot,
shard=BEACON_CHAIN_SHARD_NUMBER,
block_root=hash_tree_root(block), # where `block.sigature == EMPTY_SIGNATURE
)
proposal_root = hash_tree_root(proposal_data)
signed_proposal_data = bls_sign(
privkey=validator.privkey, # privkey store locally, not in state
message=proposal_root,
domain=get_domain(
state.fork_data, # `state` is the resulting state of `block` transition
state.slot,
DOMAIN_PROPOSAL,
)
)
Block body
Proposer slashings
Up to MAX_PROPOSER_SLASHINGS
ProposerSlashing
objects can be included in the block
. The proposer slashings must satisfy the verification conditions found in proposer slashings processing. The validator receives a small "whistleblower" reward for each proposer slashing found and included.
Casper slashings
Up to MAX_CASPER_SLASHINGS
CasperSlashing
objects can be included in the block
. The Casper slashings must satisfy the verification conditions found in Casper slashings processing. The validator receives a small "whistleblower" reward for each Casper slashing found and included.
Attestations
Up to MAX_ATTESTATIONS
aggregate attestations can be included in the block
. The attestations added must satisfy the verification conditions found in attestation processing. To maximize profit, the validator should attempt to add aggregate attestations that include the most available that have not previously been added on chain.
Deposits
Up to MAX_DEPOSITS
Deposit
objects can be included in the block
. These deposits are constructed from the Deposit
logs from the Eth1.0 deposit contract and must be processed in sequential order. The deposits included in the block
must satisfy the verification conditions found in deposits processing.
Exits
Up to MAX_EXITS
Exit
objects can be included in the block
. The exits must satisfy the verification conditions found in exits processing.
Attestations
A validator is expected to create, sign, and broadcast an attestation during each epoch. The slot during which the validator performs this role is any slot at which get_shard_committees_at_slot(state, slot)
contains a committee that contains validator_index
.
A validator should create and broadcast the attestation halfway through the slot
during which the validator is assigned -- that is SLOT_DURATION * 0.5
seconds after the start of slot
.
Attestation data
First the validator should construct attestation_data
, an AttestationData
object based upon the state at the assigned slot.
Slot
Set attestation_data.slot = slot
where slot
is the current slot of which the validator is a member of a committee.
Shard
Set attestation_data.shard = shard
where shard
is the shard associated with the validator's committee defined by get_shard_committees_at_slot
.
Beacon block root
Set attestation_data.beacon_block_root = hash_tree_root(head)
where head
is the validator's view of the head
block of the beacon chain during slot
.
Epoch boundary root
Set attestation_data.epoch_boundary_root = hash_tree_root(epoch_boundary)
where epoch_boundary
is the block at the most recent epoch boundary in the chain defined by head
-- i.e. the BeaconBlock
with slot == head.slot - head.slot % EPOCH_LENGTH
.
Note: This can be looked up in the state using get_block_root(state, head.slot - head.slot % EPOCH_LENGTH)
.
Shard block root
Set attestation_data.shard_block_root = ZERO_HASH
.
Note: This is a stub for phase 0.
Latest crosslink root
Set attestation_data.latest_crosslink_root = state.latest_crosslinks[shard].shard_block_root
where state
is the beacon state at head
and shard
is the validator's assigned shard.
Justified slot
Set attestation_data.justified_slot = state.justified_slot
where state
is the beacon state at head
.
Justified block root
Set attestation_data.justified_block_root = hash_tree_root(justified_block)
where justified_block
is the block at state.justified_slot
in the chain defined by head
.
Note: This can be looked up in the state using get_block_root(state, justified_slot)
.
Construct attestation
Next the validator creates attestation
, an Attestation
object.
Data
Set attestation.data = attestation_data
where attestation_data
is the AttestationData
object defined in the previous section, attestation data.
Participation bitfield
- Let
aggregation_bitfield
be a byte array filled with zeros of length(len(committee) + 7) // 8
. - Let
index_into_committee
be the index into the validator'scommittee
at whichvalidator_index
is located. - Set
aggregation_bitfield[index_into_committee // 8] |= 2 ** (index_into_committee % 8)
. - Set
attestation.aggregation_bitfield = aggregation_bitfield
.
Note: Calling get_attestation_participants(state, attestation.data, attestation.aggregation_bitfield)
should return a list of length equal to 1, containing validator_index
.
Custody bitfield
- Let
custody_bitfield
be a byte array filled with zeros of length(len(committee) + 7) // 8
. - Set
attestation.custody_bitfield = custody_bitfield
.
Note: This is a stub for phase 0.
Aggregate signature
Set attestation.aggregate_signature = signed_attestation_data
where signed_attestation_data
is defined as:
attestation_data_and_custody_bit = AttestationDataAndCustodyBit(
attestation.data,
False,
)
attestation_message_to_sign = hash_tree_root(attestation_data_and_custody_bit)
signed_attestation_data = bls_sign(
privkey=validator.privkey, # privkey store locally, not in state
message=attestation_message_to_sign,
domain=get_domain(
state.fork_data, # `state` is the state at `head`
state.slot,
DOMAIN_ATTESTATION,
)
)
How to avoid slashing
"Slashing" is the burning of some amount of validator funds and immediate ejection from the active validator set. In Phase 0, there are two ways in which funds can be slashed -- proposal slashing and attestation slashing. Although being slashed has serious repercussions, it is simple enough to avoid being slashed all together by remaining consistent with respect to the messages you have previously signed.
Note: Signed data must be within the same Fork
context to conflict. Messages cannot be slashed across forks.
Proposal slashing
To avoid "proposal slashings", a validator must not sign two conflicting ProposalSignedData
(suggest renaming ProposalData
) where conflicting is defined as having the same slot
and shard
but a different block_root
.
The following helper can be run to check if two proposal messages conflict:
def proposal_data_is_slashable(proposal_data_1: ProposalSignedData,
proposal_data_2: ProposalSignedData) -> bool:
if (proposal_data_1.slot != proposal_data_2.slot):
return False
if (proposal_data_1.shard != proposal_data_2.shard):
return False
return proposal_data_1.block_root != proposal_data_2.block_root
Specifically, when signing an BeaconBlock
, a validator should perform the following steps in the following order:
- Save a record to hard disk that an beacon block has been signed for the
slot
andshard
. - Generate and broadcast the block.
If the software crashes at some point within this routine, then when the validator comes back online the hard disk has the record of the potentially signed/broadcast block and can effectively avoid slashing.
Casper slashing
To avoid "Casper slashings", a validator must not sign two conflicting AttestationData
objects where conflicting is defined as a set of two attestations that satisfy either is_double_vote
or is_surround_vote
.
Specifically, when signing an Attestation
, a validator should perform the following steps in the following order:
- Save a record to hard disk that an attestation has been signed for source --
attestation_data.justified_slot // EPOCH_LENGTH
-- and target --attestation_data.slot // EPOCH_LENGTH
. - Generate and broadcast attestation.
If the software crashes at some point within this routine, then when the validator comes back online the hard disk has the record of the potentially signed/broadcast attestation and can effectively avoid slashing.