Privacy protection average consensus method and system based on rr mechanism under directed network

Abstract

The application discloses a privacy protection average consistency method and system based on a Round-Robin mechanism in a directed network, and is based on a push-sum algorithm; in the iteration process of the algorithm, the information exchanged between users is the real state value of nodes, and the total amount of information exchanged in a single iteration is too much, so that there is a privacy leakage risk; the application adopts a state decomposition mechanism to reduce the accuracy of transmitted information, and decomposes the user state into two sub-states; a part is private to the user and is used for updating the state value of the user; the other part is used for information exchange; meanwhile, a Round-Robin transmission mechanism is adopted to reduce the amount of information transmitted in a single iteration, and by adjusting the weight, each user only transmits effective information to one neighbor in the single iteration process. The application can ensure that the privacy of each user is protected on the basis of guaranteeing that each user reaches an average consistency result, and does not produce a too high calculation cost.

Description

Technical Field

[0001] This invention designs a privacy-preserving average consistency method and system based on the RR (Round-Robin) mechanism in directed networks, which is applicable to the fields of information transmission and privacy protection. Background Technology

[0002] Over the past few decades, cooperative control and distributed optimization have been widely applied in fields such as signal processing, robotics, and social networks, resulting in a wealth of related research. As a fundamental component of distributed control and optimization, the average consensus problem is currently a hot topic. Average consensus refers to the process by which each user in a network, through information exchange and continuous updating of their state value, eventually reaches the average of the initial state values ​​of all users. To achieve average consensus, a distributed algorithm with local information is needed. Under this algorithm, each user can adjust their behavior by exchanging information with their neighbors, rather than relying on a single "central figure." These methods often rely on explicit state value exchanges between neighboring users to achieve consensus in distributed computation. However, directly using real explicit state values ​​for information exchange infringes on user privacy, as these users may not want to disclose state values ​​containing sensitive and private information. This raises the privacy protection issue of average consensus methods.

[0003] In practical applications, protecting user privacy is crucial. For example, a team using average consensus to calculate a consensus might want to keep the opinions of each individual member confidential. Similarly, in the power sector, multiple power generators want to reach an agreement on costs while maintaining the confidentiality of their respective power generation information, as this information is sensitive in bidding. Therefore, in real-world scenarios, there is a need to develop distributed average consensus methods that can guarantee privacy.

[0004] The push-sum algorithm guarantees average consistency for each user, but it carries a risk of privacy breaches due to several factors: 1. During iteration, users exchange actual node state values; 2. Each user sends state values ​​to all neighbors in each iteration, resulting in excessive information exchanged in a single iteration. Therefore, further improvements are needed to ensure average consistency while protecting each user's privacy without incurring excessive computational costs. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to propose a privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks. This method can provide privacy protection for users while ensuring that each user achieves an average consistency result, without incurring a large amount of computational cost.

[0006] Technical solution: To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks includes user state decomposition and information transmission processes, with the specific operations as follows:

[0008] S1: Set the initial state value of each user k to I. k I k For the privacy information that users need to protect; read the state variable of each user k and represent it as u. k Assign each user k an artificial state variable w k , used for internal iterative calculations;

[0009] S2: Transfer the two state variables u respectively k w k Decomposed into two sub-state variables, represented as and in, It is private to the user, does not participate in information transmission, and is only used for updating the user's own status; These are sub-state variables transmitted to neighbors, and are also used for updating the state of both the variable itself and its neighbors.

[0010] S3: Based on the given network communication topology, each user obtains the neighbors that can transmit information and sorts these neighbors without repetition. In each iteration, each user selects a neighbor in order based on the Round-Robin mechanism to transmit information. If user k can transmit its own state information to m different neighbors, then every m times, user k selects the same neighbor. This process continues, with a selection operation of period m, until the iteration stops.

[0011] S4: In each iteration, user k uses the information from the previous step... We perform a weighted summation to obtain the updated state variables, and simultaneously calculate the state ratio.

[0012] S5: When the state ratio of all users All gathered to Each user achieves average consistency, where N is the total number of users.

[0013] Furthermore, in step S1: each user k has a state variable u k initial value u k [0] = I k Artificial state variable w k initial value w k [0] = 1;

[0014] Furthermore, in step S2: at the initial time, for user k, a real number is randomly generated. according to get The value; randomly generate a real number. according to get The value of .

[0015] Furthermore, in step S2: in the v-th iteration, the state variables of user k follow the following iteration rules:

[0016]

[0017]

[0018]

[0019]

[0020] in p represents all neighbors that can transmit information to user k. ki p represents the weight of the information transmitted from user i to user k. kk The weight representing the state update of user k is given by the information transmitted by user i to user k in the νth iteration. In each iteration, user k's private substate It is the weighted sum of the state values ​​used for exchange by all its neighbors and its own private state value at the previous time step; the sub-state used for exchange by user k It is a weighted sum of its two substates; during iteration, represent The weights assigned during weighted summation, and the private sub-state variables. Update Similarly, it represents The weights assigned during weighted summation, but the sub-state variables used for exchange. Update.

[0021] Furthermore, the specific rule for weight selection in step S2 is as follows: at the initial time, and Let be a randomly generated positive real number that satisfies In any v-th iteration, it is possible to specify in, and A constant greater than 0 and In each iteration, it is required that

[0022] Furthermore, step S3 specifically involves the following operations:

[0023] C1: Represent all outgoing neighbors of user k as Where m represents the total number of neighbors;

[0024] C2: Randomly sort all outgoing neighbors of user k without repetition. Once sorted, this order remains unchanged throughout the process. Represent the sequence as follows: T k This refers to the transmission sequence of the Round-Robin mechanism executed by user k;

[0025] C3: Initial weight settings and and Let p be a randomly generated positive real number; assuming user k selects l as the outgoing neighbor for transmitting information, then set the weight p. kk [0]+p lk [0] = 1 and p kk [0]>0,p lk [0]>0, the remaining unselected neighbor i, the corresponding weight is set to p. ik [0] = 0;

[0026] C4: In any νth iteration, set in, and A constant greater than 0 and Assuming user k selects l as the outgoing neighbor for transmitting information, then set p kk [v]>0,p lk [ν]>0 and p kk [ν]+p lk [ν] = 1, the remaining unselected neighbor i, and the corresponding weight is set to p. ik [ν] = 0.

[0027] Furthermore, steps S4 and S5 specifically involve each user calculating the ratio of the two state variables during each iteration. Perform calculations separately, when and All stable at At that time, that is:

[0028]

[0029] This means the method has converged and achieved an average consistent result. v is the number of iterations, and M is a pre-set, sufficiently large positive integer.

[0030] Based on the same inventive concept, this invention provides a privacy-preserving average consistency system for directed networks based on the Round-Robin mechanism, comprising:

[0031] The initialization and state decomposition module is used to set the initial state value I for each user k. k I k For the privacy information that users need to protect; read the state variable of each user k and represent it as u. k Assign each user k an artificial state variable w k Used for internal iterative calculations; the two state variables u are respectively... k w k Decomposed into two sub-state variables, represented as and in, It is private to the user, does not participate in information transmission, and is only used for updating the user's own status; These are sub-state variables transmitted to neighbors, and are also used for updating the state of both the variable itself and its neighbors.

[0032] The information iterative transmission module is used to obtain neighbors that can transmit information for each user based on a given network communication topology, and to sort these neighbors without repetition. In each iteration, each user selects one neighbor sequentially for information transmission based on a Round-Robin mechanism. If user k can transmit its own state information to m different neighbors, then every m iterations, user k selects the same neighbor. This process continues, performing selection operations with a period of m, until the iteration stops. In each iteration, user k... We perform a weighted summation to obtain the updated state variables, and simultaneously calculate the state ratio. When the state ratio of all users All gathered to Each user achieves average consistency.

[0033] Beneficial Effects: This invention proposes a privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks, aiming to preserve the initial state value of each user without incurring significant computational costs. This invention employs the Round-Robin transmission mechanism and establishes an improved push-sum algorithm based on state decomposition. Specifically, each user's state is divided into two parts: one part is private, and the other part is transmitted to one of their neighbors under the Round-Robin transmission mechanism. Compared to many methods, the Round-Robin transmission mechanism reduces the amount of information transmitted on the network at each time step, not only alleviating the communication burden but also making it difficult for attackers to obtain sufficient valuable information to estimate each user's initial value, thus demonstrating high privacy protection potential. Furthermore, although introducing the Round-Robin transmission mechanism disrupts the connectivity of the communication network, theoretical and experimental analyses confirm that this method can guarantee accurate average consistency and maintain the privacy of all users. Attached Figure Description

[0034] Figure 1 This is a flowchart of a method according to an embodiment of the present invention.

[0035] Figure 2 This is a schematic diagram of the communication network in an embodiment of the present invention.

[0036] Figure 3 This is a diagram illustrating the convergence process of the simulation experiment in this embodiment of the invention.

[0037] Figure 4 This is a diagram showing the predicted results of the simulation experiment in this embodiment of the invention. Detailed Implementation

[0038] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0039] This invention provides a privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks. For ease of understanding, the terminology involved in this method is explained below:

[0040] Directed network: A directed network consists of a set of nodes and a set of directed edges. Each edge connects two nodes and has a direction. In other words, an edge has a starting point and an ending point, indicating the directed relationship between the nodes.

[0041] Average consistency: Average consistency refers to the process in a distributed system where nodes communicate and cooperate with their neighbors by exchanging information, ultimately causing the values ​​or states of all nodes in the system to tend towards an average value or consistency. It is a goal that uses distributed algorithms to achieve coordination and cooperation among nodes to share information and obtain consistent results.

[0042] Round-Robin Mechanism: This is a scheduling algorithm used to allocate tasks or resources in a round-robin fashion. It selects each task or resource in a fixed order to ensure a fair distribution of processing time or resource utilization. Once a task or resource is completed, the next task or resource is selected in sequence for processing. This mechanism is commonly used in task scheduling, process scheduling, and load balancing to achieve fairness and balance.

[0043] Referring to the above explanation, the present invention provides a privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks, such as... Figure 1 As shown, it mainly includes user state decomposition and information transmission process improved based on the Round-Robin mechanism, specifically including the following steps:

[0044] S1: Set the initial state value of each user k to I. k Read the state variables of each user k and represent them as u. k In addition, user k is assigned an artificial state variable w. k , used for internal iterative calculations;

[0045] Wherein, the initial state value I of user k k Information considered as user privacy that needs protection must be ensured to achieve average consistency. k It cannot be deciphered by its neighbors. For each user k, u k Let u be a state variable, with an initial value of u. k [0] = I k ;w k This is an artificial state variable, with an initial value w. k [0] = 1;

[0046] S2: Transfer the two state variables u respectively k w k Decomposed into two sub-state variables, represented as and in, It is private to the user, does not participate in information transmission, and is only used for updating the user's own status; The sub-state variables to be transmitted to neighbors can be used to update both the state of the variable itself and the state of its neighbors. The state decomposition process specifically includes A1-A2.

[0047] A1: At the initial moment, for user k, a real number is randomly generated. according to get The value of . Similarly, a real number is randomly generated. according to get The value of .

[0048] A2: In the v-th iteration, the state variables of user k follow the following iteration rules:

[0049]

[0050]

[0051]

[0052]

[0053] in p represents all neighbors that can transmit information to user k. ki p represents the weight of the information transmitted from user i to user k. kk The weight representing the state update of user k is given by the information transmitted by user i to user k in the νth iteration. In each iteration, user k's private substate It is the weighted sum of the state values ​​used for exchange by all its neighbors and its own private state value at the previous time step; the sub-state used for exchange by user k It is a weighted sum of its two substates; during iteration, represent The weights assigned during weighted summation, and the private sub-state variables. Update Similarly, it represents The weights assigned during weighted summation, but the sub-state variables used for exchange. Update.

[0054] In the above process, the rules for selecting weights include B1-B3:

[0055] B1: At the initial moment, and Given randomly generated positive real numbers, the following conditions must be met:

[0056] B2: In any v-th iteration, it can be specified in, and A constant greater than 0 and

[0057] B3: In each iteration, it is required to substitute...

[0058] S3: Based on the given network communication topology, each user obtains neighbors that can transmit information and sorts these neighbors without repetition. In each iteration, each user selects one neighbor in sequence for information transmission. If user k can transmit its own state information to m different neighbors, then every m iterations, user k selects the same neighbor. This process continues, performing selection operations with a period of m until the iteration stops. The specific steps include C1-C4.

[0059] C1: Represent all outgoing neighbors of user k (all users who can receive information transmitted by k) as Where m represents the total number of neighbors;

[0060] C2: Randomly sort all outgoing neighbors of user k without repetition. Once sorted, this order remains unchanged throughout the process. Represent this sequence as... T k This refers to the transmission sequence of user k executing the Round-Robin mechanism. More specifically, in the first iteration, user k only transmits information to its neighbors. And so on, in the l-th iteration, only the neighbor is transmitted.

[0061] C3: The initial weight settings are consistent with those described above. and and Let p be a randomly generated positive real number. Assume that user k selects neighbor l as the outgoing neighbor for transmitting information, then set the weight p... kk [0]+p lk [0] = 1 and p kk [0]>0,p lk [0]>0. The remaining unselected neighbor i has its corresponding weight set to p. ik [0] = 0;

[0062] C4: In any νth iteration, set Greater than 0 and Assuming user k selects l as the outgoing neighbor for transmitting information, then set p kk [v]>0,p lk [v]>0 and p kk [v]+p lk [v] = 1, the remaining unselected neighbor i, and the corresponding weight is set to p. ik [v] = 0.

[0063] S4: In each iteration, user k uses the information from the previous step... Perform a weighted summation to obtain the updated state variables. Simultaneously calculate the state ratio. in

[0064] S5: When the state ratio of all users All gathered to The method converges when the average of the initial state values ​​of all users is reached, meaning that each user achieves average consistency.

[0065] The following theoretical analysis demonstrates the final average consistency result achieved by this invention:

[0066] set up in

[0067]

[0068] p ik The values ​​of [v] are given in the steps above. It is easy to see that the matrix P[v] is column-random.

[0069] At this point, let N be the total number of users, and let The above iterative process can be written in a compact form:

[0070]

[0071] Let the overall network communication diagram be denoted as G, and the communication diagram at step v be denoted as G... τ [v]. G τ [v] is a subgraph of G, reflecting the communication between users at step v. Assumption: The communication graph G is strongly connected, meaning that every two users can exchange information directly or indirectly.

[0072] If we assume in This represents the number of neighbors that can receive information transmitted by user k. This is the least common multiple of the number of these neighbors. It can be guaranteed that in a finite number of steps... The union of each step of the communication graph, that is... (σ can be any natural number) is strongly connected. By the relevant lemma, The corresponding matrix It is an original random matrix, which can guarantee All gathered to

[0073] The following is an example Figure 2Taking a cluster of 5 users as an example, a simulation experiment will be conducted. The 5 users want to know the average salary of the 5 users without revealing their own salary levels. Let's denote these 5 users as 1, 2, 3, 4, and 5. Figure 2 Each user transmits their personal salary information to their neighbors along the arrows. The initial collected salary levels are 4080, 6809, 5555, 7532, and 4988, in yuan / month. Figure 3 As shown, after a period of information transmission, each user achieved an average consistency result. The simulation system includes 5 user nodes, which form a transmission network, constituting a user cluster; each user node includes private salary information and salary information that can be used for communication. The specific process is as follows:

[0074] S1: Set the initial state value for each user k, i.e., their actual personal salary level is I. k In this example, I1 = 4080, I2 = 6809, I3 = 5555, I4 = 7532, and I5 = 4988. Read the state variables for each user k and represent them as u. k In addition, user k is assigned an artificial state variable w. k , used for internal iterative calculations;

[0075] Among them, user k's real salary level I k This is considered privacy information that users need to protect. For each user k, u k Let u be a state variable, with an initial value of u. k [0] = I k ;w k This is an artificial state variable, with an initial value w. k [0] = 1;

[0076] S2: Transfer the two state variables u respectively k w k Decomposed into two sub-state variables, represented as and in, It is private to the user, does not participate in information transmission, and is only used for updating the user's own status; The sub-state variable is transmitted to the neighbor and can be used to update both the state of itself and the state of the neighbor.

[0077] S3: Based on the given network communication topology, each user obtains neighbors that can transmit information and sorts these neighbors without repetition. In each iteration, each user selects one neighbor in sequence for information transmission. If user k can transmit its own state information to m different neighbors, then every m iterations, user k selects the same neighbor. This process continues, performing selection operations with a period of m, until the iteration stops.

[0078] In this example, we only need to set a weighting matrix that conforms to the weighting rules as required. For ease of calculation, let the initial weighting matrix be... as follows:

[0079]

[0080] S4: In each iteration, user k uses the information from the previous step... Perform a weighted summation to obtain the updated state variables. Simultaneously calculate the state ratio. in

[0081] S5: When the state ratio of all users All gathered to The method converges when the average wage level of all users is reached, meaning that each user reaches an average consensus.

[0082] Under the assumptions of this example, it can be verified that all assumptions are satisfied, and the average consistency result can be calculated using theoretical methods as 5792.8 yuan / month.

[0083] To verify the above results, this embodiment uses MATLAB simulation, which shows that the algorithm proposed in this embodiment can converge to the average consistency result. Figure 3 The convergence process is shown, revealing that the state ratio converges after approximately 25 iterations, reaching the precise average wage value of 5792.8 yuan / month; simultaneously, Figure 4 The study showed that the salary levels of users 2 and 1 were estimated 20 times, with extremely poor prediction results, thus achieving the effect of privacy protection.

[0084] Based on the same inventive concept, this invention discloses a privacy-preserving average consistency system for directed networks based on the Round-Robin mechanism, comprising: an initialization and state decomposition module, used to set the initial state value of each user k to I. k I k For the privacy information that users need to protect; read the state variable of each user k and represent it as u. k Assign each user k an artificial state variable w k Used for internal iterative calculations; the two state variables u are respectively... k w k Decomposed into two sub-state variables, represented as and in, It is private to the user, does not participate in information transmission, and is only used for updating the user's own status; These are sub-state variables transmitted to neighbors, and are also used for updating the state of both the variable itself and its neighbors.

[0085] The information iterative transmission module is used to obtain neighbors that can transmit information for each user based on a given network communication topology, and to sort these neighbors without repetition. In each iteration, each user selects one neighbor sequentially for information transmission based on a Round-Robin mechanism. If user k can transmit its own state information to m different neighbors, then every m iterations, user k selects the same neighbor. This process continues, performing selection operations with a period of m, until the iteration stops. In each iteration, user k... We perform a weighted summation to obtain the updated state variables, and simultaneously calculate the state ratio. When the state ratio of all users All gathered to Each user achieves average consistency. Specific implementation details are provided in the above-described method implementation examples and will not be repeated here.

Claims

1. A privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks, characterized in that, This includes the user state decomposition and information transmission process, with the following specific steps: S1: Set the initial state value of each user k to I. k I k For the privacy information that users need to protect; read the state variable of each user k and represent it as u. k Assign each user k an artificial state variable w k , used for internal iterative calculations; where each user k has a state variable u k initial value u k [0] = I k Artificial state variable w k initial value w k [0] = 1; S2: Transfer the two state variables u respectively k w k Decomposed into two sub-state variables, represented as and in, It is private to the user, does not participate in information transmission, and is only used for updating the user's own status; This is a sub-state variable transmitted to neighbors, and is also used for updating the state of both the user and its neighbors; at the initial moment, for user k, a real number is randomly generated. according to get The value; randomly generate a real number. according to get The value of ; in the v-th iteration, the state variables of user k follow the following iteration rules: in p represents all neighbors that can transmit information to user k. ki p represents the weight of the information transmitted from user i to user k. kk The weight representing the state update of user k is, in the v-th iteration, the information transmitted by user i to user k is... In each iteration, user k's private substate It is the weighted sum of the state values ​​used for exchange by all its neighbors and its own private state value at the previous time step; the sub-state used for exchange by user k It is a weighted sum of its two substates; during iteration, represent The weights assigned during weighted summation, and the private sub-state variables. Update Similarly, it represents The weights assigned during weighted summation, but the sub-state variables used for exchange. Update; S3: Based on the given network communication topology, each user obtains the neighbors that can transmit information and sorts these neighbors without repetition. In each iteration, each user selects a neighbor in order based on the Round-Robin mechanism to transmit information. If user k can transmit its own state information to m different neighbors, then every m times, user k selects the same neighbor. This process continues, with a selection operation of period m, until the iteration stops. S4: In each iteration, user k uses the information from the previous step... We perform a weighted summation to obtain the updated state variables, and simultaneously calculate the state ratio. S5: When the state ratio of all users All gathered to Each user achieves average consistency, where N is the total number of users.

2. The privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks according to claim 1, characterized in that, The specific rule for weight selection in step S2 is as follows: at the initial time, and Let be a randomly generated positive real number that satisfies In any v-th iteration, it is possible to specify in, and A constant greater than 0 and In each iteration, it is required that 3. The privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks according to claim 1, characterized in that, The specific operation of step S3 is as follows: C1: Represent all outgoing neighbors of user k as Where m represents the total number of neighbors; C2: Randomly sort all outgoing neighbors of user k without repetition. Once sorted, this sorting remains unchanged throughout the process. Represent the sequence as follows: T k This refers to the transmission sequence of the Round-Robin mechanism executed by user k; C3: Initial weight settings and and Let p be a randomly generated positive real number; assuming user k selects l as the outgoing neighbor for transmitting information, then set the weight p. kk [0]+p lk [0] = 1 and p kk [0]>0,p lk [0]>0, the remaining unselected neighbor i, the corresponding weight is set to p. ik [0] = 0; C4: In any νth iteration, set in, and A constant greater than 0 and Assuming user k selects l as the outgoing neighbor for transmitting information, then set p kk [v]>0,p lk [v]>0 and p kk [v]+p lk [v] = 1, the remaining unselected neighbor i, and the corresponding weight is set to p. ik [v] = 0.

4. The privacy-preserving average consistency method based on the Round-Robin mechanism in directed networks according to claim 1, characterized in that, Steps S4 and S5 specifically involve each user adjusting the ratio of the two state variables during each iteration. Perform calculations separately, when and All stable at At that time, that is: This means the method has converged and achieved an average consistent result. v is the number of iterations, and M is a preset positive integer.

5. A privacy-preserving average consistency system based on the Round-Robin mechanism in directed networks, characterized in that, include: The initialization and state decomposition module is used to set the initial state value I for each user k. k I k For the privacy information that users need to protect; read the state variable of each user k and represent it as u. k Assign each user k an artificial state variable w k , used for internal iterative calculations; where each user k has a state variable u k initial value u k [0] = I k Artificial state variable w k initial value w k [0] = 1; set the two state variables u respectively. k w k Decomposed into two sub-state variables, represented as and in, It is private to the user, does not participate in information transmission, and is only used for updating the user's own status; This is a sub-state variable transmitted to neighbors, and is also used for updating the state of both the user and its neighbors; at the initial moment, for user k, a real number is randomly generated. according to get The value; randomly generate a real number. according to get The value of ; in the v-th iteration, the state variables of user k follow the following iteration rules: in p represents all neighbors that can transmit information to user k. ki p represents the weight of the information transmitted from user i to user k. kk The weight representing the state update of user k is, in the v-th iteration, the information transmitted by user i to user k is... In each iteration, user k's private substate It is the weighted sum of the state values ​​used for exchange by all its neighbors and its own private state value at the previous time step; the sub-state used for exchange by user k It is a weighted sum of its two substates; during iteration, represent The weights assigned during weighted summation, and the private sub-state variables. Update Similarly, it represents The weights assigned during weighted summation, but the sub-state variables used for exchange. Update; The information iterative transmission module is used to obtain neighbors that can transmit information for each user based on a given network communication topology, and to sort these neighbors without repetition. In each iteration, each user selects one neighbor sequentially for information transmission based on a Round-Robin mechanism. If user k can transmit its own state information to m different neighbors, then every m iterations, user k selects the same neighbor. This process continues, performing selection operations with a period of m, until the iteration stops. In each iteration, user k... We perform a weighted summation to obtain the updated state variables, and simultaneously calculate the state ratio. When the state ratio of all users All gathered to Each user achieves average consistency, where N is the total number of users.

6. A privacy-preserving average consistency system based on the Round-Robin mechanism in a directed network according to claim 5, characterized in that, The weights are set in the Round-Robin mechanism as follows: Initial weight settings, and and Let p be a randomly generated positive real number; assuming user k selects l as the outgoing neighbor for transmitting information, then set the weight p. kk [0]+p lk [0] = 1 and p kk [0]>0,p lk [0]>0, the remaining unselected neighbor i, the corresponding weight is set to p. ik [0] = 0; In any νth iteration, set in, and A constant greater than 0 and Assuming user k selects l as the outgoing neighbor for transmitting information, then set p kk [v]>0,p lk [v]>0 and p kk [v]+p lk [v] = 1, the remaining unselected neighbor i, and the corresponding weight is set to p. ik [v] = 0.

Images