Xai
Badges
About
Xai is an Ethereum Layer-3 that leverages Arbitrum AnyTrust to enable open trade in the next generation of video games.
Badges
About
Xai is an Ethereum Layer-3 that leverages Arbitrum AnyTrust to enable open trade in the next generation of video games.
Arbitrum OneFunds can be stolen if
- none of the whitelisted verifiers checks the published state. Fraud proofs assume at least one honest and able validator (CRITICAL),
- a contract receives a malicious code upgrade. There is no delay on code upgrades (CRITICAL),
- none of the whitelisted verifiers checks the published state. Fraud proofs assume at least one honest and able validator (CRITICAL).
Funds can be lost if
Users can be censored if
MEV can be extracted if
| SEQUENCER FAILURE | STATE VALIDATION | DATA AVAILABILITY | EXIT WINDOW | PROPOSER FAILURE | |
| Arbitrum One L2 | Self sequence | Fraud proofs (INT) | Onchain | 7d | Self propose |
| Xai L3 • Individual | Self sequence | Fraud proofs (INT) | External (DAC) | None | Self propose |
| Xai L3 • Combined | Self sequence | Fraud proofs (INT) | External (DAC) | None | Self propose |
Sequencer failure
Self sequenceState validation
Fraud proofs (INT)No actor outside of the single Proposer can submit fraud proofs. Interactive proofs (INT) require multiple transactions over time to resolve. The challenge protocol can be subject to delay attacks. There is a 6d 8h challenge period.
Data availability
External (DAC)Proof construction relies fully on data that is NOT published onchain. There exists a Data Availability Committee (DAC) with a threshold of 5/7 that is tasked with protecting and supplying the data.
Exit window
NoneThere is no window for users to exit in case of an unwanted regular upgrade since contracts are instantly upgradable.
Proposer failure
Self proposeAnyone can become a Proposer after 25d 10h of inactivity from the currently whitelisted Proposers.
Arbitrum OneFraud proofs ensure state correctness
After some period of time, the published state root is assumed to be correct. For a certain time period, one of the whitelisted actors can submit a fraud proof that shows that the state was incorrect. The challenge protocol can be subject to delay attacks. After the state root is published, there is also a trusted entity, called Challenger, that signs it and submits the signature to a Referee smart contract. The signatures submitted to the referee are used then verified by sentry nodes. The role of sentry nodes is to verify (assert) the submitted state root after it has been submitted. There is no integrated way to flag an invalid state root, sentry nodes will have to raise the alarm by external means, making them just observation nodes.
Funds can be stolen if none of the whitelisted verifiers checks the published state. Fraud proofs assume at least one honest and able validator (CRITICAL).
- How is fraud proven - Arbitrum documentation FAQ
- Arbitrum Glossary: Challenge Period
- RollupUser.sol - Etherscan source code, onlyValidator modifier
- Referee.sol - Etherscan source code, submitChallenge function
- Referee.sol - Etherscan source code, submitAssertionToChallenge function
- Solutions to Delay Attacks on Rollups
Data is not stored on chain
Users transactions are not published on-chain, but rather sent to external trusted parties, also known as committee members (DAC). Members of the DAC collectively produce a Data Availability Certificate (comprising BLS signatures from a quorum) guaranteeing that the data behind the new transaction batch will be available until the expiry period elapses (currently a minimum of two weeks). This signature is not verified by L1, however external Validators will skip the batch if BLS signature is not valid resulting. This will result in a fraud proof challenge if this batch is included in a consecutive state update. It is assumed that at least one honest DAC member that signed the batch will reveal tx data to the Validators if Sequencer decides to act maliciously and withhold the data. If the Sequencer cannot gather enough signatures from the DAC, it will “fall back to rollup” mode and by posting the full data directly to the L1 chain. The current DAC threshold is 5 out of 7.
Funds can be lost if the external data becomes unavailable (CRITICAL).
Users can be censored if the committee restricts their access to the external data.
Xai DACUpdates to the system state can be proposed and challenged by a set of whitelisted validators. If a state root passes the challenge period, it is optimistically considered correct and made actionable for withdrawals.
Whitelisted validators propose state roots as children of a previous state root. A state root can have multiple conflicting children. This structure forms a graph, and therefore, in the contracts, state roots are referred to as nodes. Each proposal requires a stake, currently set to 0.1 ETH, that can be slashed if the proposal is proven incorrect via a fraud proof. Stakes can be moved from one node to one of its children, either by calling stakeOnExistingNode or stakeOnNewNode. New nodes cannot be created faster than the minimum assertion period by the same validator, currently set to 15m. The oldest unconfirmed node can be confirmed if the challenge period has passed and there are no siblings, and rejected if the parent is not a confirmed node or if the challenge period has passed and no one is staked on it.
Funds can be stolen if none of the whitelisted verifiers checks the published state. Fraud proofs assume at least one honest and able validator (CRITICAL).
A challenge can be started between two siblings, i.e. two different state roots that share the same parent, by calling the startChallenge function. Validators cannot be in more than one challenge at the same time, meaning that the protocol operates with partial concurrency. Since each challenge lasts 6d 8h, this implies that the protocol can be subject to delay attacks, where a malicious actor can delay withdrawals as long as they are willing to pay the cost of losing their stakes. If the protocol is delayed attacked, the new stake requirement increases exponentially for each challenge period of delay. Challenges are played via a bisection game, where asserter and challenger play together to find the first instruction of disagreement. Such instruction is then executed onchain in the WASM OneStepProver contract to determine the winner, who then gets half of the stake of the loser. As said before, a state root is rejected only when no one left is staked on it. The protocol does not enforces valid bisections, meaning that actors can propose correct initial claim and then provide incorrect midpoints.
Arbitrum OneThe system has a centralized sequencer
While forcing transaction is open to anyone the system employs a privileged sequencer that has priority for submitting transaction batches and ordering transactions.
MEV can be extracted if the operator exploits their centralized position and frontruns user transactions.
Users can force any transaction
Because the state of the system is based on transactions submitted on the underlying host chain and anyone can submit their transactions there it allows the users to circumvent censorship by interacting with the smart contract on the host chain directly. After a delay of 1d in which a Sequencer has failed to include a transaction that was directly posted to the smart contract, it can be forcefully included by anyone on the host chain, which finalizes its ordering.
Arbitrum OneRegular exit
The user initiates the withdrawal by submitting a regular transaction on this chain. When the block containing that transaction is finalized the funds become available for withdrawal on L1. The process of block finalization usually takes several days to complete. Finally the user submits an L1 transaction to claim the funds. This transaction requires a merkle proof.
Tradeable Bridge Exit
When a user initiates a regular withdrawal a third party verifying the chain can offer to buy this withdrawal by paying the user on L1. The user will get the funds immediately, however the third party has to wait for the block to be finalized. This is implemented as a first party functionality inside Arbitrum’s token bridge.
Autonomous exit
Users can (eventually) exit the system by pushing the transaction on L1 and providing the corresponding state root. The only way to prevent such withdrawal is via an upgrade.
EVM compatible smart contracts are supported
Arbitrum One uses Nitro technology that allows running fraud proofs by executing EVM code on top of WASM.
Funds can be lost if there are mistakes in the highly complex Nitro and WASM one-step prover implementation.
The system uses the following set of permissioned addresses:
Central actors allowed to submit transaction batches to L1.
They can submit new state roots and challenge state roots. Some of the operators perform their duties through special purpose smart contracts.
This is a Gnosis Safe with 3 / 5 threshold. Multisig that can execute upgrades via the UpgradeExecutor.
Those are the participants of the XaiMultisig.
The Xai Deployer EOA can upgrade all staking v2 related contracts (NodeLicenseRegistry, PoolFactory, SentryReferee, StakingPool) instantly and potentially steal all funds.
The system consists of the following smart contracts on the host chain (Arbitrum One):
Router managing token <–> gateway mapping.
Upgrade delay: No delay
Implementation used in:
The referee contract manages the Xai Sentry protocol. Sentry nodes that are tasked to watch the state transitions on Xai receive esXAI rewards for their service. These watchers participate in a game with a central ‘challenger’ by posting their assertions to make sure they are actually watching. In case of a malicious state transition, sentries are supposed to raise an alarm offchain. The referee contract is also a whitelisted address in the esXAI token contract, which allows it to initiate arbitrary esXAI token transfers. New staking through this contract is disabled in favor of the new v2 staking. V1 Stakers can continue to get staking rewards here or withdraw/migrate their assets.
Upgrade delay: No delay
The PoolFactory allows creating and managing staking pools for V2 staking. Users can stake esXAI (and / or Sentry Keys) in pools. This contract’s address is whitelisted in the esXAI token contract, which allows it to initiate arbitrary esXAI token transfers. V2 staking through this contract is currently enabled.
Upgrade delay: No delay
This is the contract where Xai Sentry Keys to run a node are minted.
Upgrade delay: No delay
Main contract implementing Arbitrum One Rollup. Manages other Rollup components, list of Stakers and Validators. Entry point for Validators creating new Rollup Nodes (state commits) and Challengers submitting fraud proofs.
Implementation used in:
Contract managing Inboxes and Outboxes. It escrows the native token used for gas on the chain. This contract stores the following tokens: XAI.
Implementation used in:
Main entry point for the Sequencer submitting transaction batches.
Implementation used in:
Implementation used in:
Contract allowed to upgrade the system.
Implementation used in:
Contract that allows challenging invalid state roots. Can be called through the RollupProxy.
Implementation used in:
Contract used to perform the last step of a fraud proof.
Implementation used in:
Contract used to perform the last step of a fraud proof.
Implementation used in:
Contract used to perform the last step of a fraud proof.
Implementation used in:
Contract used to perform the last step of a fraud proof.
Implementation used in:
Contract used to perform the last step of a fraud proof.
Implementation used in:
Value Locked is calculated based on these smart contracts and tokens:
Main entry point for users depositing ERC20 tokens. Upon depositing, on L2 a generic, “wrapped” token will be minted.
Implementation used in:
Contract managing Inboxes and Outboxes. It escrows XAI sent to L2.
Upgrade delay: No delay
Implementation used in:
The current deployment carries some associated risks:
Funds can be stolen if a contract receives a malicious code upgrade. There is no delay on code upgrades (CRITICAL).