Detailed Explanation of Offchain Labs' New Product AnyTrust Chains: L2 Chains with Minimal Trust Assumptions that are Cheaper and Faster

Translation: Jimmy

Author: Offchain Labs

Background of AnyTrust Chains

We are big fans of Optimistic Rollups. They inherit the security of the underlying L1 chain (using Ethereum as an example) while offering lower costs and higher throughput than L1. They can achieve all of this in an untrusted environment—anyone can push the correct updates to the chain state. This is why we chose to build Arbitrum One using Optimistic Rollup.

However, in some blockchain applications with special requirements, particularly in the gaming sector, some applications need to further reduce costs or withdraw NFTs more quickly. For these applications, we will introduce AnyTrust Chains, which allow for lower costs and faster withdrawals in exchange for minimal additional trust assumptions.

The main advantage of AnyTrust over sidechains is that, because it is built on Ethereum, it requires much less trust. (We will explain the details below.)

Before diving deeper, we want to emphasize that Arbitrum One will continue to maintain a trustless rollup solution as always. We will keep investing resources to improve the Arbitrum One protocol and ecosystem and will not stop our work in this area.

For example, before launching AnyTrust Chains, we will release Arbitrum Nitro and upgrade Arbitrum One to Arbitrum Nitro. We will not abandon the trustless rollup solution—we will simply introduce another option for those who need it.

Overview of AnyTrust

Here are the key points on how AnyTrust works. The network is operated by a committee of nodes, requiring only minimal assumptions about how many members of the committee are honest. For example, the committee may have 20 members, and it is assumed that at least two of them are honest.

This is a much easier trust assumption compared to traditional BFT sidechains, which require more than 2/3 of the nodes to be honest—this would mean 14 out of 20 members. Thanks to the "fallback to rollups" feature built on Ethereum, we can reduce the trust requirement from 14 to 2, as described below.

Assuming the trust holds and committee members participate, users will gain two major advantages. First, there is no need to record L2 transaction data on the L1 chain, as nodes can rely on the committee to provide the data if needed.

Moreover, in the case where the committee commits to providing data, it is safe to simply record the hash of the transaction batch on L1, thus saving the maximum cost of running rollups. Second, once the committee provides guarantees to L1, withdrawals can be executed immediately.

As long as 19 out of the 20 committee members commit (by signing) that the transaction is fine, both actions can be executed safely. The logic is that if there are at least 2 honest members, and 19 out of 20 have signed the commitment, then at least one of the 19 must be honest.

Fallback to Rollups

What happens if the committee does not sign? What if a group of committee members crashes or refuses to cooperate? In that case, the chain can still operate by falling back to the standard rollups protocol. Data will be published on the L1 Ethereum chain, and withdrawals will have a delay period, just like in standard rollups—until the committee is back up and running, at which point the chain will seamlessly switch back to the cheaper and faster mode.

Why It Is Safe (Given Trust Assumptions)

With 20 committee members, at least 2 of whom are honest, anything signed by a tribunal composed of 19 committee members must be correct, because there are at least two honest members, and only one member can be outside the tribunal, so the tribunal must include an honest member. (In general, if there are N members and K honest members, the tribunal can be any N+1-K members.)

Thus, when the tribunal signs a commitment regarding the availability of a batch of transaction data, we can guarantee the availability of that transaction data—so we can only publish the transaction root hash on L1 instead of the full transaction hash.

Similarly, if the tribunal signs a statement about a certain state, then that transaction state will be accepted directly without waiting for its lengthy challenge period on L1.

If there is no active tribunal willing to sign statements, then none of these things will happen. But that’s okay; the network can still be updated through Arbitrum’s rollups protocol, with transaction data published on Ethereum, and the new rollups state confirmed after the challenge period. Once the tribunal is back up and running, the chain will seamlessly switch back to a more efficient and faster operating mode.

In summary, in our example, if at least two members are honest, the network will operate normally. If there are also 19 members available and cooperating, it will run at the lowest cost. During the tribunal's offline period, the network will continue to operate based on Ethereum's Optimistic Rollups, collecting relevant fees, while the offline members will be used as evidence for eventual replacement.

Back to Where We Started

Are AnyTrust Chains a new idea? No, they are essentially the original Arbitrum design from our 2018 academic paper. That paper described a committee-based blockchain design that could switch to the current Optimistic Rollups in emergencies. Later, we adopted a fallback mode without the committee component and improved it to create our current Arbitrum Rollup product.

We believe it is now time to reintroduce AnyTrust Chains alongside Arbitrum One for use cases willing to make minimal trust assumptions to reduce costs and accelerate non-fungible withdrawals in common scenarios.

Why It Is Safe (Given Trust Assumptions)

We will be releasing more technical details and timelines in the future, including bridges between AnyTrust and Arbitrum One, but we want our community to know that AnyTrust is on our roadmap. If you have any questions, please feel free to reach out to us. We are always looking to expand our team, so please apply here (https://offchainlabs.com/careers/).