bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 3쪽 중 1쪽
Axelar Network:
Connecting Applications with Blockchain Ecosystems
Draft 1.0
January 2021
Abstract
Multiple blockchain ecosystems are emerging that provide unique and distinct features attractive
to users and application developers. However, communication across the ecosystems is very sparse
and fragmented. To enable applications to communicate across blockchain ecosystems frictionlessly, we
propose Axelar. Axelar stack provides a decentralized network, protocols, tools, and APIs that allow
simple cross-chain communication. Axelar protocol suite consists of cross-border routing and transfer
protocols. A decentralized open network of validators powers the network; anyone can join, use it, and
participate. Byzantine consensus, cryptography, and incentive mechanisms are designed to achieve high
safety and liveness requirements unique for cross-chain requests.
1 Introduction
Blockchain systems are quickly gaining popularity and attract new use cases for asset tokenization, decen-
tralized finance, and other distributed applications. Several major platforms such as Ethereum, Monero,
EOS, Cardano, Terra, Cosmos, Avalanche, Algorand, Near, Celo, and Polkadot offer distinct features and
development environments that make them attractive for different applications, use-cases, and end-users
[5, 11, 4, 21, 20, 23, 24, 19, 6, 14, 25]. However, the useful features of each new platform are currently
offered to less than 1% of the ecosystem’s users, namely the holders of the native token on that platform.
Can we allow platform developers to plug in their blockchains to other ecosystems easily? Can we enable
application builders to build on the best platform for their needs while still communicating across multiple
blockchain ecosystems? Can we allow users to interact with any application on any blockchain directly from
their wallets?
To bridge the blockchain ecosystems and enable applications to communicate frictionlessly across them,
we propose Axelar network. Validators collectively run a byzantine consensus protocol and run the protocols
facilitating cross-chain requests. Anyone can join the network, participate, and use it. The underlying
network is optimized for high safety and liveness requirements unique for cross-chain requests. Axelar
network also includes a protocol suite and APIs. The core protocols are:
• Cross-Chain Gateway Protocol (CGP). This protocol is analogous to Border Gateway Protocol on
the Internet. This protocol is used to connect multiple autonomous blockchain ecosystems and is
responsible for routing across them. Blockchains do not need to “speak any custom language”, their
platform developers do not need to make any custom changes on their chains, and their chains can be
plugged into the global network easily.
• Cross-Chain Transfer Protocol (CTP). This protocol is analogous to application-level protocols File
Transfer, Hypertext Transfer Protocols on the Internet. It is an application-level protocol stack that
sits on top of routing protocols (such as CGP and other routing technologies). Application developers
can connect their dapps on any chain to perform cross-chain requests. Users can use the CTP protocol
to interact with applications on any chain using simple API calls analogous to HTTP GET/POST
requests. Developers can lock, unlock, and transfer assets between any two addresses across any
blockchain platforms, execute cross-chain application triggers (e.g., an dapps on chain A, can update
1
its state if some other application on chain B satisfies some search criteria (interest rate > X )), and
perform general cross-chain requests between apps across chains (a smart contract on chain A can call
to update a state of a smart contract on chain B). This protocol enables the composability of programs
across blockchain ecosystems.
Axelar network offers the following advantages:
• For blockchain platform builders : Ability to easily plug-in their blockchains to all other blockchain
ecosystems. Only a threshold account needs to be set up on the chain to plug into the network.
• For dapps builders : Application builders can host their dapps anywhere, lock, unlock, transfer assets,
and communicate with applications on any other chain via CTP API.
• For users: Users can interact with all applications across the ecosystem directly from their wallets.
A Platform for Builders. Finally, Axelar network is a platform for developers and a global community.
Its governance model is open to anyone. Developers can propose new integration points, routing, and
application-level protocols, and the users can decide whether to adopt them by voting on the proposals and,
if approved, validators will adopt the changes.
1.1 Existing Interoperability Solutions
Previous attempts to solve interoperability across blockchains fall in one of four categories: centralized
exchanges, interoperable ecosystems, wrapped assets, and token bridges. We briefly summarize these ap-
proaches below.
Centralized Systems. Today centralized systems are the only truly scalable solutions to interoperability
needs for the ecosystem. They can list any asset or onboard any platform relatively easily. However,
centralized systems are known to have various security issues and are not good enough to power the emerging
decentralized financial system that requires robust security, transparency, and open governance. On their
own they cannot power the decentralized applications as they grow.
Interoperability Hubs. Projects such as Cosmos, Polkadot, Ava labs address interoperability between
sidechains native to their ecosystems using custom inter-chain communication protocols [23, 25, 24]. For
example, one can spin-up a sidechain (a Cosmos Zone) that can communicate with the Cosmos Hub. The
sidechain must be based on the Tendermint consensus and speak the protocol natively understood by the
Cosmos Hub. Connections to other blockchains and ecosystems that speak different languages is left to
external technologies.
Pairwise Bridges. Wrapped assets (e.g., wrapped Bitcoins) try to fill-in the missing cross-chain inter-
operability gap in the ecosystem. One example is tBTC [9], which is a custom protocol where a clever
combination of smart contracts and collateral is used to secure the transfers. These solutions require sub-
stantial engineering efforts to build – for each chain pair, developers must build a new smart contract on
destination chain that parses state proofs from origin chain (similar to how each side-chain could, in princi-
ple, parse state of other chains). Only a handful of bridges have been deployed using this approach. These
approaches do not scale when one of the underlying blockchains wants to upgrade its consensus rules or
transaction format. This is because all smart contracts that depend upon the state of these chains would
need to be upgraded. One must also set up validators and require them to lock up different assets in order
to overcollateralize any asset transfer, which limits the economic efficiency of such transfers.
We have also seen a few other single-purpose bridges by platform developers that rewrite state-transition
logic in smart contracts to bridge to other ecosystems [1, 7]. They suffer from multiple scalability issues,
do not allow the ecosystem to scale uniformly, and introduce additional dependencies for applications. For
instance, if one platform changes, then all smart contracts on all bridges will need to be upgraded. This
2
will ultimately put the ecosystem in a gridlock where no-one will be able to upgrade. Finally, if one single-
purpose bridge connects platforms A and B, and a second single-purpose bridge connects B and C, it does
not mean that applications on A will be able to talk to applications on C. One might need to create another
single-purpose bridge or rewire application logic.
Other attempts to tackle interoperability include federated oracles (e.g., Ren [8]), and application specific
interoperable blockchains [10].
To summarize, existing solutions for interoperability require heavy engineering work from both platform
developers and application builders that must understand different communication protocols to communicate
across every pair of ecosystems. And so, interoperability is virtually non-existing in today’s blockchain space.
At the end of the day, platform developers want to focus on building platforms and optimize them for their
use-cases and be able to plug in to other blockchains easily. And application developers want to build dapps
on the best platforms for their needs while still leveraging users, liquidity and communicate with other dapps
on other chains.
2 The Quest for Scalable Cross-Chain Communication
At the core, cross-chain communication requires that heterogeneous networks find the ability to communicate
using the same language. To solve this, we explain the Axelar protocol suite, describe its high-level properties,
and explain how these properties address the core of scalable cross-chain communication.
1. “Plug-and-play” integration. Blockchain platform builders should not be required to perform heavy
engineering or integration work to speak some “custom language” to support cross-chain. The cross-
chain protocol should be able to plug in any existing or new blockchain frictionlessly. New assets should
be added with minimal effort.
2. Cross-chain routing. Functions such as the discovery of network addresses, routing paths, and networks
are at the core of the Internet and facilitated by BGP and other routing protocols. Similarly, to
facilitate communication across blockchain ecosystems, we need to support the discovery of addresses
across them, applications, and routing.
3. Upgradability support. If one of the blockchain ecosystems changes, it should not affect the interop-
erability of other blockchains. The system needs to recognize updates, and minimal effort should be
required to support them (i.e., no “state transition logic” should be rewritten, and applications should
not break).
4. Uniform language for applications. Applications need a simple protocol to lock, unlock, transfer, and
communicate with other applications no matter which chain they reside on. This protocol must be
chain-agnostic and support simple calls, analogous to HTTP/HTTPS protocols that allow users and
browsers to communicate with any web-server. As more networks and assets join the lower-level routing
protocols, applications should be able to use them for communications without rewriting their software
stacks.
Next, we summarize the security requirements that these protocols must meet.
1. Decentralized trust. The network and protocols must be decentralized, open, and allow everyone to
participate fairly.
2. High safety. The system must satisfy high safety guarantees. The system needs to preserve the safety
of assets and the state as the cross-chain network processes it.
3. High liveness. The system must satisfy high liveness guarantees to support applications leveraging its
cross-chain features.
3
Satisfying a subset of these properties is easy. For instance, one can create a federated multisig account
with their friends and lock/unlock assets on corresponding chains. Such systems are inherently vulnerable
to collusion and censorship attacks and lack proper incentives for the validators to protect them. Creating
a decentralized network and protocol suite where anyone can participate while being correctly incentivized
can enable frictionless cross-chain communication, but solving it is a hard problem that requires a careful
combination of consensus, cryptographic, and mechanism design protocols.
3 Axelar Network
Axelar network provides a uniform solution to cross-chain communication that meets the needs of both plat-
form developers – no integration work is required from them, and application builders – one simple protocol
and API to access global liquidity and communicate with the entire ecosystem.
Axelar network consists of a decentralized network which bridges blockchain ecosystems that speak differ-
ent languages and a protocol suite with APIs on top, making it easy for applications to perform cross-chain
requests. The network connects existing stand-alone blockchains such as Bitcoin, Stellar, Terra, Algorand,
and interoperability hubs such as solutions like Cosmos, Avalanche, Ethereum, and Polkadot. Our mission
is to enable application developers to build such apps easier using a universal protocol and API without
rolling out their proprietary cross-chain protocols underneath or rewriting applications as new bridges are
developed. Towards this, we designed a protocol suite that includes Cross-Chain Gateway Protocol (see
Section 6) and Cross-Chain Transfer Protocol (see Section 7).
A core component of the network are the underlying decentralized protocols. Validators collectively
maintain the Axelar network and run the nodes that secure the Axelar blockchain. They are elected through
a delegation process by the users. Validators receive voting power pro-rata according to the stake delegated
to them. The validators reach consensus on the state of multiple blockchains that the platform is connected
to. The blockchain is responsible for maintaining and running the cross-chain routing and transfer protocols.
Governance rules allow network participants to enact protocol decisions such as which blockchains to bridge
and which assets to support.
Axelar blockchain follows a Delegated Proof-of-Stake (DPoS) model similar to Cosmos Hub. Users elect
validators who must bond their stake to participate in the consensus and maintain high-quality service. The
DPoS model allows maintenance of large decentralized validator set and robust incentives to guarantee that
the validators are responsible for maintaining bridges and shares of cryptographic threshold schemes. As
part of consensus, validators run light-client software of other blockchains, allowing them to verify the state
of other blockchains. The validators report these states to the Axelar blockchain, and once enough of them
report, the state of Bitcoin, Ethereum, and other chains is recorded on Axelar.
Subsequently, the Axelar base layer is aware of the state of external blockchains at any point in time,
creating the “incoming bridges” from other blockchains. The validators collectively maintain threshold sig-
nature accounts on other blockchains (e.g., 80% of validators must approve and co-sign any transaction
out of it), which allows them to lock and unlock assets and state across chains and to post state on other
blockchains, the “outgoing bridges.” Altogether, one can view the Axelar network as a decentralized cross-
chain read/write oracle.
The remainder of the document describes preliminaries and building blocks behind the network (Sec-
tion 4), some technical details of the network (Section 5), cross-chain gateway protocol (Section 6), and
cross-chain transfer protocol (Section 7).
4
4 Preliminaries
4.1 Notation and Assumptions
Let V r denote the set of Axelar validators at round R. Each validator has a weight, a number in (0 , 1]
denoting the voting power of that particular validator. The weights of all validators add up to 1. A validator
is correct if she runs a node that is consistent with the rules of the Axelar protocol. To finalize blocks, or
to sign cross-chain requests, Axelar requires correct validators of total weight > F . We call the parameter
F∈ [0.5, 1] the protocol threshold.
Axelar can be based on an instant finality Delegated-Proof-of-Stakeblockchain. The validators run Byzan-
tine Fault Tolerant (BFT) consensus at each round i to finalize the ith block. Once the ith block is finalized,
new BFT consensus is run to finalize the i + 1th block, and so on. The validators are elected through
stake delegation. A user with some stake may elect to run a validator node, or delegate their voting power
(stake) to an existing validator, who then votes on their behalf. The validator set can be updated, validators
join/leave the set, and users delegate/undelegate their voting power.
Different blockchains work under different network assumptions. Synchronous communication means
that there is a fixed upper bound ∆ on the time messages take to be delivered, where ∆ is known and can
be built into the protocol. Asynchronous communication means that messages may take arbitrarily long to
be delivered, and it is known that BFT protocols cannot be built for asynchronous networks even in the
presence of just one malicious validator. A realistic compromise between synchrony and asynchrony is the
assumption of partially synchronous communication . The network may be completely asynchronous until
some unknown global stabilization time (GST), but after GST communication becomes synchronous with a
known upper bound ∆ [17].
Typical blockchains work under the assumption of > F correct validators. For synchronous networks
F = 1 /2 is typically set, but for the weaker assumption of a partially synchronous network F = 2 /3.
Bitcoin, its forks, and the current Proof-of-Work version of Ethereum only work assuming synchrony. Others
like Algorand and Cosmos only require partial synchrony. When connecting chains through Axelar, the
connection works assuming the strongest network assumptions out of these chains, which is synchrony in
the case of connecting Bitcoin and Cosmos, for instance. The Axelar blockchain itself works in a partially
synchronous setting and thus requires F = 2/3, but it is possible to improve the threshold requirement by
assuming that other existing blockchains are secure and leveraging their security.
4.2 Cryptographic Preliminaries
Digital Signatures. A digital signature scheme is a tuple of algorithms (Keygen, Sign, V erif y ). Keygen
outputs a pair of keys ( P K, SK). Only the owner of SK can sign messages, but anyone can verify the
signatures given the public key P K. Most blockchain systems today use one of the standard signature
schemes such as ECDSA, Ed25519, or a few of their variants [2, 3].
Threshold Signatures. A threshold signature scheme enables a group of n parties to split a secret key
for a signature scheme in such a way that any subset of t + 1 or more parties can collaborate to produce a
signature, but no subset of t or fewer parties can produce a signature or even learn any information about
the secret key. The signatures produced by the threshold protocols for ECDSA and EdDSA look identical
to the signatures produced by the stand-alone algorithms.
A threshold signature scheme replaces the Keygen and Sign algorithms for an ordinary signature scheme
with distributed n-party protocols T .Keygen, T .Sign. These protocols typically require both a public
broadcast channel and private pairwise channels among the parties, and they typically involve several rounds
of communication. After successful completion of T .Keygen each user holds a share s i of a secret key SK
and the corresponding public key PK. The T .Sign protocol allows these parties to produce a signature for a
5
given message that is valid under public key PK. This signature can be verified by anyone using the Verify
algorithm of the original signature scheme.
4.3 Properties of Threshold Signatures
There are several properties a threshold scheme might have that are especially desirable for decentralized
networks:
Security against a dishonest majority . Some threshold schemes have the restriction that they are secure
only when a majority of the n parties are honest. Thus, the threshold parameter t must be smaller
than n/2 [15]. This restriction is typically accompanied by the fact that 2 t + 1 honest parties are
needed to sign, even though only t + 1 corrupted parties can collude to recover the secret key. Schemes
that do not suffer from this restriction are said to be secure against a dishonest majority .
As discussed later in Section 5.2, cross-chain platforms must maximize the safety of their networks and
be able to tolerate as many corrupted parties as possible. Thus, schemes that can tolerate dishonest
majorit프로젝트로 돌아가기