CN115065468A - PBFT consensus optimization method based on grouping reputation value - Google Patents

Abstract

Aiming at the problems of the most used PBFT consensus algorithm in the alliance chain, when the number of nodes increases, the algorithm delay becomes longer, the throughput decreases and the selection of master nodes is arbitrary, a PBFT consensus optimization method based on group reputation value is proposed. The nodes of the entire network are grouped to facilitate the selection of nodes and to control the number of nodes participating in the consensus algorithm; each node group converts the point-to-point voting value of the Vague set into the specific value of the fuzzy set as the node reputation value , select the node with the highest reputation value in each group as the representative node to participate in the consensus; use the verifiable random function to randomly select the master node in the consensus node, increase the unpredictability of the master node; finally, in order to reduce the communication frequency between nodes, simplify the consensus process. Experimental tests show that the GV‑PBFT algorithm can effectively reduce the consensus delay and improve throughput.

Description

A PBFT consensus optimization method based on group reputation value

technical field

The invention belongs to the field of blockchain consensus mechanism, and specifically relates to a PBFT consensus optimization method based on group reputation value.

Background technique

The consensus algorithm plays a very important role in the blockchain system, which ensures the correctness and consistency of the data on the nodes. At present, there are many consensus algorithms. Proof of Work (POW), Proof of Stake (POS) and replication and fault tolerance (raft) are commonly used consensus algorithms, but they all have certain problems. POW has the problem of consuming computing power and wasting resources; POS exists Centralization and insufficient consensus efficiency; Raft has the problem of poor performance in high concurrency scenarios. One of the most used consensus algorithms in consortium chains is Practical Byzantine Fault Tolerance (PBFT), which has very high transaction throughput. However, its algorithm itself has some problems. When the number of nodes increases, PBFT needs to communicate point-to-point, and its performance drops sharply. Secondly, PBFT relies on the master node, but its master node is selected arbitrarily, and it is easy to select malicious nodes as the most dominant node. nodes, which will also affect the system efficiency.

In this regard, scholars at home and abroad have also studied and improved some existing consensus algorithms. Chen Zihao et al. proposed to divide the nodes into layers through a clustering algorithm, and first perform the consensus within the group and then perform the consensus outside the group to reduce the number of consensus communications. Woo Yusen et al. proposed to use the improved Raft consensus algorithm to first conduct consensus within the group, and then outside the group, each group leader will conduct PBFT consensus according to the consensus result within the group. Zheng X et al. proposed to use C4.5 to classify nodes, select nodes with high trust level as the main consensus group, and other nodes as the secondary consensus group, first perform PBFT consensus in the secondary consensus group and then perform the second PBFT in the main consensus group Consensus confirmation, and a point voting mechanism is introduced to determine the leader node for consensus. Li et al. proposed to use a random function in the nodes to select some nodes to participate in the consensus, thereby reducing the number of PBFT communications and improving the efficiency of the algorithm. However, the above research only considers the positive impact on the node and does not consider the negative impact of the node in the process of evaluating the reputation value of the node, and the judgment is not accurate enough.

SUMMARY OF THE INVENTION

In view of the deficiencies of the prior art, the present invention provides a PBFT consensus optimization method based on group reputation value.

A PBFT consensus optimization method based on group reputation value, the specific steps are as follows:

Step 1: Group the nodes of the entire network to facilitate the selection of nodes and control the number of nodes participating in the consensus algorithm;

Step 1.1: All nodes in the network will be stored in a list. According to the address identification of the nodes on the list, the nodes are divided into several node groups, and the number of nodes in each group has an upper limit. The limit is L;

Step 1.2: When a new node a appears in the network, it needs to send its own public key P to the surrounding, and then node a enters the waiting state,

Step 1.3: After the other node b in the network receives the request information from a, it first confirms that the number of nodes in its own group does not reach the upper limit L, and then puts the information of accepting a to join the group in the event and transmits it to the surrounding nodes , indicating that the node has joined the group;

Step 1.4: After receiving the acceptance information, node a confirms to join the group, and the receipt information occurs again, indicating that the acceptance is successful;

Step 1.5: After receiving the information, node b broadcasts it to the nodes of the entire network, updates the group node list, and synchronizes the latest list to node a, and then a synchronizes all data information of the entire network;

Step 2: Each node group converts the point-to-point voting value of the Vague set into the specific value of the fuzzy set as the node reputation value, and selects the node with the highest reputation value in each group as the representative node to participate in the consensus, that is, the consensus node;

The specific value of converting the Vague set into the fuzzy set is used as the node reputation value, and the specific formula is as follows:

Among them, [t _P (x), 1-f _P (x)] generally represents the fuzzy value of x; the voting model is used to explain the concept of the Vague set, which means that when the number of supporting votes exceeds the number of non-supporting votes, abstain from voting The person who votes may be more inclined to the side that supports, and vice versa is more inclined to the side that does not support; λ is a constant, λ>0 usually takes 1;

Step 3: Use a verifiable random function in the consensus node to randomly select the master node to increase the unpredictability of the master node;

In the consensus node, the master node is selected according to the verifiable random function. The specific steps are as follows:

Step 3.1: Each consensus node takes private key Sk and message X as input to generate random number y and proof function Proof;

Step 3.2: Determine whether the consensus node is the master node, if it meets the conditions of the following formula, it will return YES and become the master node;

When a node is verified as the master node, other consensus nodes are slave nodes;

Step 4: Consensus is achieved among consensus nodes through GV-PBFT algorithm to achieve global consensus;

The GV-PBFT algorithm calculates the reputation value based on the PBFT algorithm, and randomly selects the representative nodes with high reputation as the master node, which increases the probability of the master node being an honest node. The three-stage algorithm optimization is changed to two stages, and the submission stage is cancelled. thus obtained;

Step 4.1: Client c sends a request message <REQUEST, o, t, c, X> to the consensus node; where o represents the specific operation of the request, t represents the timestamp of the request, and X is a random value selected by the client;

Step 4.2: When the consensus node verifies that it is the master node, it assigns a sequence number n to the request in view v, and broadcasts a pre-prepare message to other slave nodes <<PRE-PREPARE, v, n, d>< Verify<Pk, y, Proof>, m>, where v is the view number, m is the client request message, d is the digest of message m, Pk is the public key of the master node, y is the random number generated by the master node, Proof The proof function generated by the master node;

Step 4.3: The slave node receives the message sent by the master node, and the slave node verifies the master node's identity and the validity of the message; if the message verification passes, each slave node sends a prepare message to other consensus nodes except itself < PREPARE, v, n, d, i>, where i is the node number;

Step 4.4: Each consensus node verifies the received message, and when it receives 2f valid messages sent by different consensus nodes except itself, it will send a response message to the client; the client receives the same response message of f+1< REPLY, v, t, c, i, r>, the consensus is complete.

Beneficial technical effects of the present invention:

The present invention introduces the specific value of the Vague set converted into the fuzzy set as the node reputation value as the basis for selecting the representative node, and better reduces the probability of the malicious node being selected as the consensus node. The introduction of a verifiable random function to select the master node increases the unpredictability of the master node and improves system security. The consensus process of the original PBFT algorithm is simplified, the three-stage algorithm optimization is changed to two stages, and the submission stage is cancelled, thereby reducing the communication frequency between nodes and improving the algorithm performance.

Description of drawings

Fig. 1 is a flow chart of a method for optimizing PBFT consensus based on group reputation value of the present invention.

FIG. 2 is an algorithm flow chart of GV-PBFT in a PBFT consensus optimization method based on group reputation value of the present invention.

FIG. 3 is a flowchart of a PBFT algorithm according to an embodiment of the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and embodiments;

The invention proposes a PBFT consensus optimization algorithm based on group reputation value to solve the problems of time extension, high throughput and random selection of main nodes in the original PBFT consensus algorithm in the alliance chain. The algorithm first dynamically groups nodes, and selects representative nodes in the group according to the score of the Vague set as the reputation value of each node. There are representative nodes for global consensus, because the three situations in which the Vague set is integrated into the voting are more realistic and can be more realistic. Accurately evaluate nodes and reduce the participation of abnormal nodes. Use a verifiable random function in the representative node to randomly determine the master node, making the master node unpredictable and protecting the security of the system. Since reputation evaluation increases the probability that the master node is an honest node, the consensus process can be simplified into two stages to reduce the number of communications.

The invention provides a PBFT consensus optimization method based on group reputation value, which solves the problems of the original PBFT consensus mechanism that the number of nodes increases, the consensus throughput is low, the time is prolonged, and the main node is selected arbitrarily. As shown in Figure 1, the specific steps are as follows:

Step 1: Group the nodes of the entire network to facilitate the selection of nodes and control the number of nodes participating in the consensus algorithm;

Step 1.1: All nodes in the network will be stored in a list. According to the address identification of the nodes on the list, the nodes are divided into several node groups, and the number of nodes in each group has an upper limit. The limit is L;

Step 1.2: When a new node a appears in the network, it needs to send its own public key P to the surrounding, and then node a enters the waiting state,

Step 1.3: After the other node b in the network receives the request information from a, it first confirms that the number of nodes in its own group does not reach the upper limit L, and then puts the information of accepting a to join the group in the event and transmits it to the surrounding nodes , indicating that the node has joined the group;

Step 1.4: After receiving the acceptance information, node a confirms to join the group, and the receipt information occurs again, indicating that the acceptance is successful;

Step 1.5: After receiving the information, node b broadcasts it to the nodes of the entire network, updates the group node list, and synchronizes the latest list to node a, and then a synchronizes all data information of the entire network;

Step 2: Each node group converts the point-to-point voting value of the Vague set into the specific value of the fuzzy set as the node reputation value, and selects the node with the highest reputation value in each group as the representative node to participate in the consensus, that is, the consensus node;

The specific value of converting the Vague set into the fuzzy set is used as the node reputation value, and the specific formula is as follows:

Among them, [t _P (x), 1-f _P (x)] generally represents the fuzzy value of x; the voting model is used to explain the concept of the Vague set, which means that when the number of supporting votes exceeds the number of non-supporting votes, abstain from voting The person who votes may be more inclined to the side that supports, and vice versa is more inclined to the side that does not support; λ is a constant, λ>0 usually takes 1;

For example, there are 10 people voting, assuming that P is a Vague set in U, x ∈ U and the value of Vague is [0.5, 0.9], so t _p (x) = 0.5, f _p (x) = 1-0.9 = 0.1, then the value of Vague is [0.5, 0.9], indicating that there are 5 people who voted for it, 1 person who voted not, and 4 people abstained;

Step 3: Use a verifiable random function in the consensus node to randomly select the master node to increase the unpredictability of the master node;

In the consensus node, the master node is selected according to the verifiable random function. The specific steps are as follows:

Step 3.1: Each consensus node takes private key Sk and message X as input to generate random number y and proof function Proof;

Step 3.2: Determine whether the consensus node is the master node, if it meets the conditions of the following formula, it will return YES and become the master node;

When a node is verified as the master node, other consensus nodes are slave nodes;

Step 4: Consensus is achieved among consensus nodes through the GV-PBFT algorithm to achieve global consensus; as shown in Figure 2;

The GV-PBFT algorithm calculates the reputation value based on the PBFT algorithm, and randomly selects the representative nodes with high reputation as the master node, which increases the probability of the master node being an honest node. The three-stage algorithm optimization is changed to two stages, and the submission stage is cancelled. thus obtained;

Step 4.1: Client c sends a request message <REQUEST, o, t, c, X> to the consensus node; where o represents the specific operation of the request, t represents the timestamp of the request, and X is a random value selected by the client;

Step 4.2: When the consensus node verifies that it is the master node, it assigns a sequence number n to the request in view v, and broadcasts a pre-prepare message to other slave nodes <<PRE-PREPARE, v, n, d>< Verify<Pk, y, Proof>, m>, where v is the view number, m is the client request message, d is the digest of message m, Pk is the public key of the master node, y is the random number generated by the master node, Proof The proof function generated by the master node;

Step 4.3: The slave node receives the message sent by the master node, and the slave node verifies the master node's identity and the validity of the message; if the message verification passes, each slave node sends a prepare message to other consensus nodes except itself < PREPARE, v, n, d, i>, where i is the node number;

Step 4.4: Each consensus node verifies the received message, and when it receives 2f valid messages sent by different consensus nodes except itself, it will send a response message to the client; the client receives the same response message of f+1< REPLY, v, t, c, i, r>, the consensus is complete.

As shown in Figure 3, it can be seen that the three-stage process of PBFT includes pre-preparation, preparation and submission. The specific steps for reaching a consensus in the PBFT algorithm are:

In the preparation phase, the client first sends a request to the master node. The master node checks and processes the client's request, and then broadcasts a prepare message to the slave nodes. Each slave node receives and verifies the validity of the prepared message. Once the message is verified correctly, the slave node will accept the request and begin the preparation phase.

In the preparation phase, each slave node transmits messages to other nodes, and at the same time, they also receive messages from other slave nodes and check the validity of the messages. The preparation phase ends when the slave node obtains 2f valid prepare messages from different slave nodes, where f is the number of malicious nodes.

In the commit phase, each node broadcasts a commit message to other nodes for verification, and once a node in the preparation phase receives a number of messages including other itself equal to or greater than 2f+1, it will send a reply message to the client. When the client receives a reply message and the client receives f+1 of the same reply message, a consensus is reached.

The original PBFT consensus algorithm has three stages: pre-preparation, preparation and submission. The submission stage is to ensure that there are enough trusted nodes to complete the verification of the proposal. When the master node is an honest node, consensus can be reached in two phases, and no commit phase is required. The GV-PBFT algorithm calculates the reputation value and randomly selects the representative nodes with high reputation as the master node, which increases the probability that the master node is an honest node. Therefore, the GV-PBFT algorithm optimizes the three-stage algorithm to two stages, and cancels the submission. stage. GV-PBFT has improved both throughput and delay than the original PBFT. The consensus process of the original PBFT algorithm is simplified, the three-stage algorithm optimization is changed to two stages, and the submission stage is cancelled, thereby reducing the communication frequency between nodes and improving the algorithm performance.

Claims (5)

1. A PBFT consensus optimization method based on a grouping reputation value is characterized by comprising the following specific steps:

step 1: grouping nodes of the whole network, conveniently performing corresponding screening on the selection of the nodes and controlling the number of the nodes participating in the consensus algorithm;

step 2: each node group converts the voting value of the figure set point-to-point into a specific value of a fuzzy set as a node reputation value, and selects the node with the highest reputation value in each group as a representative node to participate in consensus, namely a consensus node;

and step 3: randomly selecting a main node by using a verifiable random function in the consensus node, and increasing the unpredictability of the main node;

and 4, step 4: carrying out consensus among the consensus nodes through a GV-PBFT algorithm to realize global consensus;

the GV-PBFT algorithm is obtained by calculating through a credit value based on a PBFT algorithm, randomly selecting a representative node with high credit as a main node, increasing the probability that the main node is an honest node, changing three-stage algorithm optimization into two stages and canceling a submission stage.

2. The PBFT consensus optimization method based on the packet reputation value according to claim 1, wherein step 1 specifically comprises:

step 1.1: all nodes in the network are stored in a list, the nodes are divided into a plurality of node groups according to the address identification of the nodes on the list, and the number of the nodes in each group has an upper limit which is set as L;

step 1.2: when a new node a appears in the network, it needs to send its own public key P to the periphery, and then node a enters a wait state,

step 1.3: after receiving the request information of a, other nodes b in the network firstly confirm that the node number of the group in which the node b is positioned does not reach the upper limit L, and then put the information of receiving the addition of a into the group into an event and transmit the information to the surrounding nodes to indicate that the node is added into the group;

step 1.4: after receiving the joining acceptance information, the node a confirms to join the group, and generates the receipt information again to represent that the node a accepts successfully;

step 1.5: after receiving the information, the node b broadcasts the information to the nodes in the whole network, updates the small group node list, synchronizes the latest list to the node a, and then synchronizes all data information in the whole network.

3. The PBFT consensus optimization method based on grouping reputation values according to claim 1, wherein step 2 converts the Vague set into specific values of the fuzzy set as node reputation values, and the specific formula is as follows:

wherein, [ t ] _P (x),1-f _P (x)]Fuzzy values that generally represent x; the concept of a Vague set is explained using a voting model, which means that when the number of votes supporting a vote is redundant with the number of votes not supporting a vote, the right to surrender isThe person of the ticket may prefer the party that is supported, and vice versa; λ is a constant, λ>0 is usually taken to be 1.

4. The PBFT consensus optimization method based on the packet reputation value according to claim 1, wherein step 3 selects a master node in the consensus node according to a verifiable random function, and comprises the following specific steps:

step 3.1: each common identification node takes a private key Sk and a message X as input to generate a random number y and a Proof function Proof;

step 3.2: judging whether the consensus node is a main node or not, if the consensus node meets the conditions of the following formula, returning to YES to become the main node;

when one node is verified to be the master node, the other consensus nodes are slave nodes.

5. The PBFT consensus optimization method based on the packet reputation value according to claim 1, wherein step 4 specifically comprises:

step 4.1: the client c sends a REQUEST message < REQUEST, o, t, c, X > to the consensus node; wherein o represents a specific operation of the request, t represents a timestamp of the request, and X is a random value selected by the client;

and 4.2: when the consensus node verifies that the consensus node is the master node, a sequence number n is distributed for the request in a view v, and a PRE-prefix message < < PRE-PREPARE, v, n, d > < Verify < Pk, y, Proof >, m > is broadcasted to other slave nodes, wherein v is a view number, m is a client request message, d is a summary of a message m, Pk is a public key of the master node, y is a random number generated by the master node, and Proof is a Proof function generated by the master node;

step 4.3: the slave node receives the message sent by the master node, and verifies the identity of the master node and the validity of the message; each slave node sends a prefix message < prefix, v, n, d, i > to other common nodes except the slave node after the message verification is passed, wherein i is a node number;

step 4.4: each consensus node verifies the received message, and sends a response message to the client when receiving 2f effective messages sent by different consensus nodes except the consensus node; the client receives the REPLY message < REPLY, v, t, c, i, r > with the same f +1, and the consensus is completed.

Images