Arbitrum’s fraud proof security model: five myths vs facts

Interactive proving and efficiency

Arbitrum uses interactive proving to resolve disputes. If Alice claims a result and Bob disagrees, they engage in a back-and-forth protocol refereed by an L1 contract. Alice divides her N-step claim into two claims of size N/2. Bob then picks one of the N/2-step claims to challenge. This process continues, cutting the dispute in half at each stage, until the disagreement concerns a single step of execution. The L1 referee does not evaluate the merits of the execution until the dispute narrows to a single step. Because interactive proving can resolve disputes that are larger than one transaction, it allows a rollup block to contain only a single claim about the end state of the chain after all of the execution covered by the block. This method is more efficient than re-executing transactions. Re-execution requires the L1 referee to emulate the execution of an entire transaction to see if the outcome matches the claim. This forces a lower gas limit than Ethereum. It also forces lower contract size limits because the instrumented code must fit into a single Ethereum contract. The main principle is that Alice and Bob should do as much off-chain work as possible needed to resolve their dispute, rather than putting that work onto an L1 contract. If you read about a feature of Arbitrum, you can ask how it supports or takes advantage of interactive proving.

Permissionless validation and BoLD

The Arbitrum BoLD upgrade enables permissionless validation. Currently, a permissioned set of validators prevents delay attacks on the current protocol. A delay attack occurs when actors sacrifice stakes to delay confirmations. BoLD mitigates this risk by enforcing a fixed upper time bound on dispute resolution. This ensures any honest party can defend against malicious claims. The assertion confirmation time includes two challenge periods of approximately 6.4 days each, a 2-day grace period, and a small delta for computing challenges. The Security Council can intervene during the 2-day grace period if severe bugs exist in the OneStepProver contracts. The Security Council uses a 9 of 12 multisig for emergencies and a 7 of 12 multisig for routine upgrades. Implementation includes an audit by Trail of Bits and 8 weeks of deployment to public testnets. This upgrade targets Nitro contracts 2.1.0 for Arbitrum One and Nova, while also whitelisting Infura’s validator address for Arbitrum Nova. This transition represents a step toward the technology stack being recognized as a Stage 2 Ethereum rollup. It allows for a new challenger manager contract on Ethereum. Will BoLD transition the network to a fully decentralized state?

Two-dimensional fee structures

Arbitrum fees consist of L1-native resources and L2-native resources. L1 resources involve Ethereum calldata. L2 resources involve computation in the Arbitrum VM. The total fee is the sum of these two components. The system uses a two-dimensional fee model. The gas limit G is the L2 gas used plus the L1 calldata price multiplied by the L1 calldata size divided by the L2 gas price. In this model, P represents the L2 gas price. Because the L2 gas price is in the denominator, an increase in L2 gas price decreases the value of G. Arbitrum Nitro transaction fees are 20-40% cheaper than Optimism for simple transfers.

Action Cost
Send ETH (21,000 gas) $0.00069
Send USDC (41,000 gas) $0.0013
Approve (46,000 gas) $0.0015
Mint NFT (100,000 gas) $0.0033
Bridge out (100,000 gas) $0.0033

Security vulnerabilities and sequencer reliability

The May 2024 Code4rena audit of the Arbitrum BoLD smart contract system identified 4 vulnerabilities. Two vulnerabilities had a high severity rating, while two had a medium severity rating. One issue involves insolvency. If the required stake is updated to a lower amount, an adversary can make an honest party unable to retrieve their assertion stakes. If the initial stake is 1000 ETH and the admin updates it to 700 ETH, Alice can use a friend to make a child assertion with a 700 ETH stake. Alice can then withdraw her original stake. This leaves the contract with only 700 ETH, making it insolvent for the original 1000 ETH requirement. Another issue involves timer inheritance. Malicious actors can frontrun the confirmation of an honest challenge tree by inheriting timers from it. This happens because the function RollupUserLogic::updateTimerCacheByClaim allows timers to be inherited across different challenge trees. The claiming edge must be exactly one level below the inheriting edge and its originId must match the mutualId of the inheriting edge. Without this validation, rivals belonging to dishonest challenge edge trees can also be valid inheriting edges. A third issue concerns the sequencer. A bug in the batch submission mechanism caused a pause in transaction finalization on June 7, 2022. The sequencer is currently a single, centralized entity. The sequencer has halted three times. On January 9, 2022, a bug in the batch submission mechanism lasted 7 hours. On July 6, 2022, a lack of Ethereum gas caused a one-hour pause. On December 15, 2023, a surge in Inscriptions caused the sequencer to stop relaying transactions properly.

Newsletter