CN-122001736-A - BFT consensus method based on random sampling committee and cross-view voting accumulation

CN122001736ACN 122001736 ACN122001736 ACN 122001736ACN-122001736-A

Abstract

The invention discloses a BFT consensus method based on random sampling committee and cross-view voting accumulation, and relates to the technical field of block chains and distributed ledgers. The invention organizes the consensus process into successive views, electing a leader in each view and randomly sampling to produce a scale Is a committee of (c). The leader proposes blocks and the committee members check and vote, which propagates and accumulates in a pipelined fashion across multiple views. And the system dynamically counts the effective votes obtained by the block in the continuous K views, generates QC when the accumulated votes reach 2f+1, and realizes the final submission of the block by continuously obtaining a plurality of QC acknowledgements. The number K of successive views required for voting accumulation can be dynamically adjusted according to the network conditions to balance security and efficiency. The invention reduces the communication expenditure through the committee sampling, improves the consensus flexibility and fault tolerance through the cross-view voting accumulation, and is suitable for the Bayesian fault tolerance scene in the high-dynamic and asynchronous network environment.

Inventors

Yan Zhongju
ZHANG CHENYU
XU HAO
LI KEQIU

Assignees

天津大学

Dates

Publication Date: 20260508
Application Date: 20260210

Claims (8)

1. The BFT consensus method based on random sampling committee and cross-view voting accumulation is characterized by being applied to a distributed system comprising N auxiliary nodes, wherein N >3f and f are the number of tolerable Bayesian and busy-family copies preset by the system, and specifically comprises the following steps of: S1, view and role division, wherein the view is taken as a basic unit in the consensus process, each view corresponds to a preset duration, a leader is elected in each view through a preset election algorithm, meanwhile, a committee is selected through a random sampling algorithm, and the committee members participate in block voting and voting result forwarding; S2, block proposal and committee voting, wherein a leader of the current view generates a new block based on the data to be consensus, performs validity pre-verification, and then sends the block to all members of the committee; after receiving the block, the committee member performs validity check, after the check passes, the signature voting is performed on the block, a voting message is generated, and the voting message and the corresponding block information are forwarded to other replica nodes of the unselected committee in the system; S3, a pipeline type voting propagation mechanism, wherein a consensus process advances in a plurality of continuous views in a pipeline manner, and the voting flow of the same block gradually accumulates votes across the plurality of continuous views; S4, cross-view voting accumulation and confirmation certificate generation, namely counting the effective voting quantity of each block in K continuous views in real time, wherein K is a dynamic adjustment variable; when a certain block accumulates and obtains the effective votes of at least 2f+1 non-repeated duplicate nodes in the continuous K views, the system automatically generates a corresponding legal confirmation certificate to finish the staged confirmation of the block; S5, dynamic voting depth control, wherein the K value is dynamically adapted according to the sampling probability of a random sampling algorithm, the current network condition and the voting arrival rate, and the K value is reduced when the network delay is low and the voting arrival rate is high; S6, a block submitting rule is that after a certain block continuously obtains legal confirmation certificates with preset times for confirmation, the block is formally submitted and written into a system state machine to complete final consensus, and a submitting result is synchronized to all copy nodes; S7, designing a random sampling interface, namely providing a universal random sampling algorithm interface, supporting pure random sampling or constrained random sampling, wherein the constrained random sampling controls the member overlapping proportion of adjacent view committees by setting member screening conditions and a committee overlapping degree threshold; and S8, view switching processing, namely triggering view switching when the fault of the leader or the duration of the view is over, re-electing the leader and generating a new committee, simultaneously keeping a voting record of a historical block, and continuously advancing voting accumulation of a corresponding block by the new view to realize cross-view voting multiplexing.
2. The method of BFT consensus based on random sampling committee and cross-view voting accumulation according to claim 1, wherein the committee in S1 is of the scale Where N is the total number of replica nodes currently surviving in the system.
3. The BFT consensus method based on random sampling committee and cross-view voting accumulation as recited in claim 1, wherein the validity check in S2 comprises a data integrity check and a signature validity check.
4. The BFT consensus method based on random sampling committee and cross-view voting accumulation as recited in claim 1, wherein the current network conditions in S5 include network delay and packet loss rate.
5. The BFT consensus method based on random sampling committee and cross-view voting accumulation according to claim 1, wherein in the block submission rule in S6, a preset legal validation certificate validation number threshold is 3 times, and the threshold is configured and adjusted according to system security requirements.
6. The BFT consensus method based on random sampling committee and cross-view voting accumulation as recited in claim 1, wherein the member screening conditions in the constrained random sampling in S7 include one or more of node historical behavior credibility, computational effort level.
7. The BFT consensus method based on random sampling committee and cross-view voting accumulation as claimed in claim 1, wherein the committee overlap threshold in S7 is used to limit the proportion of the number of identical members in neighboring view committees to reduce the risk of malicious node centralization.
8. The BFT consensus method based on random sampling committee and cross-view voting accumulation as recited in claim 1, wherein during the view switch in S8, a history of voting records is maintained continually by the voting accumulation statistics module, and a leader of a new view continues to initiate voting calls for unacknowledged blocks based on the records.

Description

BFT consensus method based on random sampling committee and cross-view voting accumulation Technical Field The invention relates to the technical field of blockchains and distributed ledgers, in particular to a BFT consensus method based on random sampling committee and cross-view voting accumulation, which is suitable for distributed system scenes requiring high security and high consistency, such as finance, data sharing, distributed computing and the like. Background With the wide application of blockchain and distributed ledger technologies in the fields of finance, data sharing, distributed computing and the like, a Bayesian fault tolerance (Byzantine Fault Tolerant, BFT) consensus mechanism is one of core technologies for guaranteeing system safety and consistency, and can ensure that each node in a distributed system agrees on data under the condition that part of nodes have malicious faults (Bayesian faults). Existing BFT consensus protocols (e.g., PBFT, tendermint, hotStuff, etc.) typically require that in each round of consensus or View (View), all or most copies are broadcast by the leader and votes are collected, forming a quorum of acknowledgement messages (Quorum Certificate, QC). However, when the system scale increases, the above scheme has the following problems in general: 1. The communication complexity is high, the voting and broadcasting process involves a large number of copies, so that the network load is linearly increased along with the number of nodes, even in square level, and the problem of network congestion is prominent in a large-scale node deployment scene, so that the consensus efficiency is seriously influenced; 2. The leader usually needs to wait for fixed timeout time to collect enough votes, the fixed timeout mechanism lacks flexibility, is difficult to adapt to dynamic network environments such as network delay fluctuation, temporary offline of nodes and the like, and is easy to cause consensus blocking; 3. the random committee is uncontrollable, a part of schemes are introduced into the random sampling committee to reduce communication overhead, but under a pure random mechanism, the rationality of committee members is difficult to guarantee, so that voting convergence speed and system security are difficult to balance, and the risk of malicious node centralized access to the committee can occur in extreme cases; 4. the prior consensus protocol generally requires QC to be formed in the same view, after the view is switched (such as view change caused by a leader fault), the history voting cannot continue to play a role, and the voting process needs to be restarted, so that the waste of computing resources and network resources is caused, and the consensus efficiency is further reduced. Therefore, a new bayer fault-tolerant consensus scheme is needed, communication overhead is reduced, consensus propulsion efficiency is improved, and adaptability to dynamic network environments is enhanced on the premise of ensuring safety. Disclosure of Invention The invention aims to provide a BFT consensus method based on random sampling committee and cross-view voting accumulation, which solves the problems of high communication complexity, long waiting time, uncontrollable random committee and incapability of multiplexing cross-view voting in the existing BFT consensus protocol, and realizes the following steps on the premise of strictly ensuring the safety of a Bayesian court: 1. The communication complexity of single-round consensus is obviously reduced, and the network load pressure in a large-scale node scene is relieved; 2. the expandability of the system under the large-scale node environment is improved, and the application scene of dynamic increase and decrease of the number of the nodes is adapted; 3. the time for the leader to wait for voting is reduced, the consensus responsiveness is improved, and the network environment with dynamic fluctuation is adapted; 4. The controllability of random committee selection is realized, the security and the consensus performance of the system are balanced, and the centralized risk of malicious nodes is avoided; 5. and cross-view voting multiplexing is realized, resource waste is reduced, and consensus propulsion efficiency is further improved. In order to achieve the above purpose, the present invention adopts the following technical scheme: The BFT consensus method based on random sampling committee and cross-view voting accumulation is applied to a distributed system comprising N auxiliary nodes, wherein N >3f and f are the number of tolerable Bayesian and busy copies preset by the system, and the method specifically comprises the following steps of: S1, view and role division, wherein a common-knowledge process takes views (views) as basic units, each View corresponds to a preset duration, a Leader (Leader) is elected in each View through a preset election algorithm, meanwhile, a committee is selected through a random sampling algo