"SBFT: 面向区块链的可扩展去中心化信任基础架构"

需积分: 0 155 浏览量更新于2024-01-20 收藏 443KB PDF 举报

SBFT是一种面向区块链的可扩展去中心化信任基础架构。该架构由Guy Golan Gueta (VMware R)、Ittai Abraham (VMware R)、Shelly Grossman (TAU)、Dahlia Malkhi (VMware R)、Benny Pinkas (BIU)、Michael K. Reiter (UNC-Chapel Hill)、Dragos-Adrian Seredinschi (EPFL)、Orr Tamir (TAU)和Alin Tomescu (MIT)共同开发。该架构实现了一种新的拜占庭容错算法，解决了可扩展性和去中心化的挑战。 SBFT的目标是在分布式系统中实现高性能和去中心化。在现有的区块链系统中，由于交易的广播和验证需要得到大多数节点的确认，因此在节点数量增加时，系统吞吐量会降低，无法满足高并发的需求。而SBFT通过引入快速的异步复制和轻量级的共识算法，实现了高吞吐量和低延迟。 SBFT的核心思想是将区块链系统分为不同的组，并使每个组内的节点执行拜占庭容错算法来达成共识。这样，系统的吞吐量可以随着组的数量而线性增加，而不受节点数量的限制。同时，SBFT还通过使用异步传输机制和选择性复制机制来减少网络通信的开销，进一步提高性能。为了保证系统的去中心化，SBFT采用了多权益交叉选举机制。这意味着每个组中的节点不仅可以投票选择领导者，同时也可以投票选择其他组的领导者。这样，系统中没有单一的控制节点，所有节点都有平等的参与权，并且没有任何一组可以垄断整个系统。为了验证SBFT的可扩展性和性能，研究团队进行了一系列的实验。实验结果表明，与传统的区块链共识算法相比，SBFT在高负载情况下具有更高的吞吐量和更低的延迟。此外，SBFT还可以在网络异常的情况下继续工作，保证了系统的稳定性和可靠性。总之，SBFT是一种面向区块链的可扩展去中心化信任基础架构。通过引入新的拜占庭容错算法和多权益交叉选举机制，SBFT实现了高吞吐量、低延迟和去中心化的特性。研究团队的实验结果表明，SBFT在性能和可靠性方面具有优势，有望成为未来区块链系统的重要基础架构之一。

(see Section 4.7); and π, with threshold of f + 1 for the execution protocol (see Section 4.4).

We assume a computationally bounded adversary that cannot do better than known attacks as of 2017 on the

cryptographic hash function SHA256 and on BLS BN-P254 [BGDM

10] based signatures. We use a PKI setup

between replicas for authentication. Clients need to know just one public key, which is needed for veriﬁcation of the

threshold signature π, unlike PBFT, where every client needs to maintain membership and know the public keys of all

replicas.

3 Service Properties

Our algorithm provides a scalable fault tolerant implementation of a generic replicated service. We implement an

authenticated key-value store that uses a Merkle tree interface [Mer88] for data authentication. We begin by deﬁning

the generic service interface and the authenticated data structure framework needed to leverage SBFT’s beneﬁts. We

then detail the safety, liveness and scalability guarantees of our system.

3.1 Service interface

As a generic replication library, SBFT requires an implementation of the following service interface to be received

as an initialization parameter. The interface implements any deterministic replicated service with state, deterministic

operations and read-only queries. An execution val = execute(D, o) modiﬁes state D according to the operation o

and returns an output val. A query val = query(D, q) returns the value of the query q given state D (but does not

change state D). These operations and queries can perform arbitrary deterministic computations on the state.

The state of the service moves in discrete blocks. Each block contains a series of requests. We denote by D

the state of the service at the end of sequence number j. We denote by req

the series of operations of block j, that

changed the state from state D

j−1

to state D

An authenticated key-value store. For our blockchain implementation we use a key-value store. In order to sup-

port efﬁcient client acknowledgement from one replica we augment our key-value store with a data authentication

interface. As in public permissionless blockchains, we use a Merkle trees interface [Mer88] to authenticate data. To

provide data authentication we require an implementation of the following interface:

(1) d = digest(D) returns the Merkle hash root of D as digest.

(2) P = proof (o, val, s, D, l) returns a proof that operation o was executed as the lth operation in the the series

of requests in the decision block whose sequence number is s, whose state is D and the output of this operation was

val. For a key-value store, proof for a put operation is a Merkle tree proof that the put operation was conducted as

the lth operation in the requests of sequence number s. For a read only-query q, we write P = proof (q, val, s, D) and

assume all queries are executed with respect to D

(the state D after completing sequence number s). For a key-value

store, proof for a get operation is a Merkle tree proof that at the state with sequence number s the required variable

has the desired value;

(3) verify(d, o, val, s, l, P ) returns true iff P is a valid proof that o was executed as the lth operation in sequence

number s and the resulting state after this decision block was executed has a digest of d and val is the return value for

operation o (and similarly verify(d, q, val, s, P ) when q is a query). For a key-value store and a put operation above,

the veriﬁcation is the Merkle proof veriﬁcation [Mer88] rooted at the digest d (Merkle hash root).

3.2 Replication service guarantees

For n = 3f + 2c + 1 replicas SBFT obtains the following properties:

(1) Safety in the asynchronous mode (adversary controlling at most f Byzantine nodes and all network delays).

This means that any two replicas that execute a decision block for a given sequence number, execute the same decision

block.

(2) Liveness in the synchronous mode (adversary controlling at most f Byzantine nodes). Roughly speaking,

liveness means that client requests return a response.

剩余22页未读，继续阅读

萌新小白爱学习

粉丝: 25
资源: 311

"SBFT: 面向区块链的可扩展去中心化信任基础架构"

安卓Java实现QQ邮件发送与Hyperledger Fabric 0.6代码解析

区块链农业应用科技行动倡议.pdf

区块链共识算法：PoW、PoS和其他变种的比较分析

简化的拜占庭容错算法（SBFT）

SBFT算法在各种应用场合都优于PBFT算法

【中国房地产业协会-2024研报】2024年第三季度房地产开发企业信用状况报告.pdf

【中国银行-2024研报】美国大选结果对我国芯片产业发展的影响和应对建议.pdf

RM1135开卡工具B17A

毕业设计&课设_宿舍管理系统：计算机毕业设计项目.zip

毕业设计&课设_画手交易管理系统：Java 毕设项目.zip

最新资源