Average consensus privacy protection method based on random additive sharing

Abstract

The invention provides an average consensus privacy protection method based on random additive sharing, and the method comprises the steps: a first intelligent agent generates a first random number, a second intelligent agent generates a second random number, and the second intelligent agent is any one of adjacent intelligent agents in communication connection with the first intelligent agent; the first random number, the second random number, the virtual state of the first intelligent agent determined in the last round and the virtual state of the second intelligent agent determined in the last round are used as parameters for operation, and secret value parameters of the first intelligent agent and the second intelligent agent are determined; sharing a secret value parameter of the first intelligent agent and a secret value parameter of the second intelligent agent, and determining a shared secret value of the first intelligent agent and the second intelligent agent; the first agent updates the virtual state of the first agent according to the shared secret value; and repeating the steps until the whole system reaches the consensus. The virtual state is generated by using the random number, so that the intelligent agent can share the change of the state and achieve an average consensus under the condition that the privacy of the intelligent agent is not leaked.

Description

Technical Field

[0001] The present invention belongs to the field of information security technology, and specifically relates to an average consensus privacy protection method based on random additive sharing. Background Art

[0002] The consensus algorithm of a multi-agent system refers to each agent in a distributed network using its own information and data collected from neighboring agents as input and adjusting its own state. Under this algorithm, the state information of all agents will eventually reach a consensus over time. Multi-agent consensus has a wide range of applications in various fields, such as drone formation control, sensor networks, distributed energy management, etc. These applications rely on the collaboration between agents to complete the assigned tasks. However, collaboration essentially involves the exchange of information between these agents. If this information contains sensitive data, it will bring vulnerabilities to network security. Therefore, it is necessary to solve the problem of privacy protection in the average consensus of multi-agent systems.

[0003] Among non-encrypted methods, a common approach to addressing the pressing needs of privacy and security is differential privacy schemes. Typically, these methods achieve obfuscation by introducing noise into the state to mask the actual state value. However, in terms of the balance between privacy and convergence accuracy, differential privacy methods often fail to provide accurate average consensus values. This effect may directly hinder convergence to the true value, or limit convergence only in a statistical mean square sense. Another approach is state decomposition, which is to pre-divide the true state of the agent to ensure the privacy of all participating agents. However, the feasibility of this approach depends largely on the topology of the agent network. If the topology changes frequently, this approach may result in a large amount of resource consumption or fail to keep up with the update of the multi-agent state quickly.

[0004] In terms of cryptographic methods, homomorphic encryption has gradually become the main method for privacy protection. Homomorphic encryption has the property of preserving addition and multiplication operations. Specifically, when two encrypted values ​​are homomorphically encrypted, the decryption of their sum and product is equal to the sum and product of their respective decrypted values. Homomorphic encryption schemes can promote exact convergence under the condition that at least one trusted agent is connected. Many researchers have applied homomorphic encryption to various agent operation scenarios. However, homomorphic encryption methods are not conducive to low-power systems. Each algorithm update of an agent requires the use of a homomorphic encryption algorithm for its neighboring agents. This operation process significantly increases the demand for the system's computing resources. In addition, homomorphic encryption schemes have an obvious limitation, that is, they are usually designed for processing integers and cannot be directly applied to scenarios where the agent state involves decimals. Therefore, a scheme that takes into account both the efficiency and security of state sharing is a strongly related problem that needs to be solved when privacy protection technology is applied to multi-agent systems. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide an average consensus privacy protection method and device based on random additive sharing to meet the requirements of protecting the privacy and security of each intelligent agent.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] According to the first aspect, the present invention provides an average consensus privacy protection method based on random additive sharing, including: Step 1: the first agent generates a first random number, and the second agent generates a second random number, and the second agent is any one of the adjacent agents that are communicatively connected to the first agent; Step 2: using the first random number, the second random number, the first agent's own virtual state determined in the previous round, and the second agent's own virtual state determined in the previous round as parameters to perform additive operations to determine the first agent's secret value parameters and the second agent's secret value parameters; Step 3: sharing the first agent's secret value parameters and the second agent's secret value parameters, and determining the first agent and the second agent's shared secret value based on the first agent's secret value parameters and the second agent's secret value parameters; Step 4: the first agent updates the first agent's virtual state based on the shared secret values ​​with multiple second agents; repeat the above steps 1 to 4 until the entire system reaches a consensus.

[0008] Optionally, an additive operation is performed using the first random number, the second random number, the virtual state of the first agent determined in the previous round, and the virtual state of the second agent determined in the previous round as parameters to determine the secret value parameters of the first agent and the second agent, including: the first agent determines the first agent state representation data based on its own virtual state determined in the previous round and the first random number; the second agent determines the second agent state representation data based on its own virtual state determined in the previous round and the second random number; the first agent sends the first random number to the second agent, and the second agent sends the second agent state representation data to the first agent; the first agent determines the first secret value parameter based on the second agent state representation data and the first trusted random number, and determines the second secret value parameter based on the first agent state representation data and the second trusted random number, the first trusted random number and the second trusted random number being pre-issued to the first agent by a third-party trusted institution; the second agent determines the third secret value parameter based on the first random number and the third trusted random number, and determines the fourth secret value parameter based on the second random number and the fourth trusted random number, and the third trusted random number and the fourth trusted random number being pre-issued to the second agent by a third-party trusted institution.

[0009] Optionally, determining a shared secret value between the first agent and the second agent based on the secret value parameters of the first agent and the secret value parameters of the second agent includes: determining a first shared secret value based on the first secret value parameter and the fourth secret value parameter; determining a second shared secret value based on the second secret value parameter and the third secret value parameter.

[0010] Optionally, the first intelligent agent updates the virtual state of the first intelligent agent based on a shared secret value with multiple second intelligent agents, including: the second intelligent agent determines the second intelligent agent secret state value based on the shared secret value and a set of trusted random numbers pre-issued to the second intelligent agent by a third-party trusted agency; the first intelligent agent determines the first intelligent agent secret state value based on the shared secret value and a set of trusted random numbers pre-issued to the first intelligent agent by a third-party trusted agency; the first intelligent agent receives the second intelligent agent secret state value sent by the second intelligent agent, and determines the virtual state difference between the first intelligent agent and the second intelligent agent based on the second intelligent agent secret state value and the first intelligent agent secret state value; the first intelligent agent updates the virtual state of the first intelligent agent based on the virtual state difference between the first intelligent agent and the second intelligent agent.

[0011] Optionally, before the first agent generates a first random number and the second agent generates a second random number, the process includes: the first agent generates a set of random numbers and the second agent generates a set of random numbers; the first agent determines an initial virtual state of the first agent based on the generated set of random numbers and the random number array generated by the second agent.

[0012] Optionally, determining the virtual state difference between the first agent and the second agent according to the secret state value of the second agent and the secret state value of the first agent comprises:

[0013]

[0014] x i '-x' y = log(z i +z j );

[0015] Among them, z j represents the secret state value of the first agent, z i represents the secret state value of the second agent, z i =de+db i +ea i +c i , z j =db j +ea j +c j , d represents the first shared secret value, d = d i +d j , represents the second random number generated by the second agent, represents the state representation data of the second agent, x i ' represents the virtual state of the second agent determined in the previous round, e represents the second shared secret value, e = e i +e j , represents the first random number generated by the first agent, represents the state representation data of the first agent, x' j represents the virtual state of the first agent determined in the previous round, (a i ,b i ,c i ) represents a set of trusted random numbers issued by a third-party trusted institution to the second agent, (a j ,b j ,c j ) represents a set of trusted random numbers issued to the first agent by a third-party trusted institution; x' i -x' j Represents the difference in virtual state between the first agent and the second agent.

[0016] Optionally, the first agent updates the virtual state of the first agent according to the virtual state difference between the first agent and the second agent, including:

[0017]

[0018] Among them, x j' (k+1) represents the virtual state of the first agent in round K+1, x j' (k) represents the virtual state of the first agent in round K, represents a constant step size, Δ is the maximum degree of graph G, N j represents the total number of adjacent second agents that are connected to the first agent for communication, a ij (k) represents the weighted adjacency matrix parameter. If the first agent is not a neighbor of the second agent, then a ij (k) = 0, if the first agent is in communication with the second agent, then there exists a positive number η such that a ij (k)∈[η,1] holds for all k, x i' (k)-x j' (k) represents the virtual state difference between the first agent and the second agent in round K.

[0019] According to the second aspect, an embodiment of the present invention provides an average consensus privacy protection device based on random additive sharing, including: a random number generation module, used for a first agent to generate a first random number, and a second agent to generate a second random number, the second agent being any one of the adjacent agents that are communicatively connected to the first agent; a secret value parameter determination module, used for performing additive operations using the first random number, the second random number, the first agent's own virtual state determined in the previous round, and the second agent's own virtual state determined in the previous round as parameters to determine the first agent's secret value parameters and the second agent's secret value parameters; a secret value sharing module, used for determining a shared secret value between the first agent and the second agent based on the secret value parameters of the first agent and the secret value parameters of the second agent; a virtual state update module, used for the first agent to update the virtual state of the first agent based on the shared secret value with multiple second agents; and a consensus module, used for repeating the above-mentioned random number generation module steps to the virtual state update module steps until the entire system reaches a consensus.

[0020] According to the third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the average consensus privacy protection method based on random additive sharing described in the first aspect or any implementation scheme of the first aspect.

[0021] According to the fourth aspect, an embodiment of the present invention provides a computer storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the average consensus privacy protection method based on random additive sharing as described in the first aspect or any embodiment of the first aspect.

[0022] The present invention provides an average consensus privacy protection method based on random additive sharing, which ensures that agents share changes in their states and reach average consensus without revealing their privacy by using random numbers to generate virtual states. When each agent is connected to at least one trusted agent, it can resist security attacks from honest but curious agents and global eavesdropping agents, and the method only relies on simple mathematical operations such as addition and multiplication of constants, as well as exponential and logarithmic calculations. Despite the simplicity of calculation, the method is not limited to the positive integer states of the agent, but also includes decimals and negative values. A large amount of encryption processing is avoided through simple mathematical operations.

[0023] Other advantages, objectives and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or those skilled in the art may be taught from the practice of the present invention. The objectives and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration:

[0025] Figure 1 This is a specific example flow chart of an average consensus privacy protection method based on random additive sharing of the present invention;

[0026] Figure 2 A simulation diagram of the multi-agent state change process of the non-privacy protection solution of the present invention and the method provided by the embodiment of the present invention;

[0027] Figure 3 A simulation diagram showing the performance comparison between the method provided in the embodiment of the present invention and the homomorphic encryption scheme under conditions of different topological complexities;

[0028] Figure 4 It is a principle block diagram of a specific example of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, it can also be the internal connection of two components, it can be a wireless connection, or it can be a wired connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0031] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0032] For ease of understanding, the following are the professional concepts mentioned in this method and the mathematical foundations it relies on:

[0033] Secret sharing: When sensitive information needs to be distributed and stored across multiple locations, secret sharing, as a cryptographic technique, provides a secure way to split sensitive information into multiple shared fragments and distribute them among different participants. Only when a certain number of fragments are combined together can the original secret be recovered. This technology has a wide range of applications in distributed systems, cryptography, network security and other fields.

[0034] Secret sharing is a less restrictive form of homomorphic encryption. Homomorphic encryption requires significant computational overhead, which may affect system performance in some cases. The ciphertext length of homomorphic schemes is usually much longer than the original plaintext. In contrast, secret sharing exhibits high computational efficiency due to the lack of encryption and decryption processes.

[0035] The method provided by the present invention relies on graph theory and average consensus as mathematical foundations. The following describes the relevant contents of graph theory and average consensus for reference:

[0036] 1. Graph Theory:

[0037] Let G = (V, E, A) be a connected undirected graph. Each edge (i, j) ∈ E represents the connection between two different agents i and j. The set V = {1, 2, ..., N} contains all agents, and N = |V| is the total number of agents. Agents are connected to each other, and the edges have no directionality. The weighted adjacency matrix is ​​denoted by A(k) = [a ij (k)] N×N , where k = 0, 1, 2 ... represents the current time in the discrete time system and represents the time-varying nature of the weight. For the sake of convenience, in some cases a ij Can refer to a ij (k). It is generally assumed that in a weighted adjacency matrix, for all k≥0,i,j=1,2,…,N:

[0038] (1)a ij (k) = a ji (k);

[0039] (2) If j is not a neighbor of i, then a ij (k) = 0;

[0040] (3) If j is connected to i, then there exists a positive number η such that a ij (k)∈[η,1] holds for all k.

[0041] Under this assumption, the influence of an agent can continue to affect its neighbors and is not constrained by additional constraints. Conditions (1) and (2) are derived from the inherent properties of undirected graphs. Condition (3) further ensures that each agent has the ability to distribute constrained influence weights between individual agents and their neighbors.

[0042] In addition, in order to ensure the connectivity of the multi-agent system, it is generally assumed that the underlying graph between the multi-agent systems is connected. In an undirected graph, it is equivalent to the concept of strong connectivity.

[0043] 2. Average consensus:

[0044] In the undirected multi-agent system G, the state update equation can be expressed as:

[0045]

[0046] in represents a constant step size. Δ is the maximum degree of graph G.

[0047] Under the influence of the above update rules, the state of the agent eventually converges to the mean of the initial value. It is called average consensus convergence, as shown below:

[0048]

[0049] The achievement of average consensus in discrete-time multi-agent systems depends on two key conditions: (i) G is strongly connected; (ii) there exists a positive constant η ≥ 0 such that η ≤ a ij (k)≤1 holds for all k≥0.

[0050] The update rule can also be written as:

[0051] x(k+1)=Ρ(k)x(k);

[0052] where x(k)=[x1(k),x2(k),...,x N (k)] T ,P(k)=1-εL(k). L(k) is the Laplace matrix at k.

[0053] In the network, normal neutral agents are defined as agents that faithfully follow the consensus protocol. They will not help other agents infer or conceal information. Before introducing the method provided by the present invention, two common eavesdropping scenarios that may be encountered in multi-agent networks are first introduced.

[0054] (1) Honest but curious attackers. Honest but curious agents are attackers who actively participate in the normal communication and collaboration process. However, they try to collect and record the data they receive during the transmission process. Their purpose is not to destroy the operational integrity of the system or compromise its consistency. Instead, they carefully observe the information flow within the system. Honest and curious agents do not know the topological connections, weights, etc.

[0055] (2) Global eavesdropping attackers. Global eavesdropping agents represent a more threatening group of attackers who have the ability to intercept most or all of the communication data in the network. Using their extensive eavesdropping capabilities, these attackers can access a large amount of sensitive data, which may pose a serious threat to the security and confidentiality of the system.

[0056] It should be noted that neither of the above two attackers can obtain the initial state of the agent.

[0057] The embodiment of the present invention provides an average consensus privacy protection method based on random additive sharing, such as Figure 1 As shown, including:

[0058] Step S1: The first agent generates a first random number, and the second agent generates a second random number, where the second agent is any one of the adjacent agents that are in communication connection with the first agent;

[0059] Exemplarily, the first agent j generates a first random number The second agent i generates a second random number In this embodiment, for different scenes, the first agent j and the second agent i can be objects in the corresponding scenes. For example, in an intelligent transportation system, the first agent j and the second agent i can be vehicles, and in a drone formation, the first agent j and the second agent i can be drones. There can be multiple adjacent agents near the first agent j, referred to as the first agent i.

[0060] Step S2: performing an additive operation using the first random number, the second random number, the virtual state of the first agent determined in the previous round, and the virtual state of the second agent determined in the previous round as parameters to determine the secret value parameter of the first agent and the secret value parameter of the second agent;

[0061] For example, the first agent j determines its virtual state x' according to the last round (the Kth round) j (k)(abbreviated as x' j ) and the first random number Determine the first agent state representation data The specific process includes: In order to convert the privacy-preserving multiplication sharing result into the addition calculation in the state consensus, this embodiment uses the power form to calculate its own virtual state x' j Process and obtain The virtual state s of the power form j and the first random number Do subtraction to get the state representation data of the first agent

[0062] The second agent i determines its own virtual state x according to the previous round (round K) i '(k)(abbreviated as x i ') and the second random number Determine the state representation data of the second agent The specific process includes: In order to convert the privacy-preserving multiplication sharing result into the addition calculation in the state consensus, this embodiment uses the power form to calculate its own virtual state x' i Process and obtain The virtual state s of the power form i and the second random number Do subtraction to get the state representation data of the second agent

[0063] The first agent j sends the first random number To the second agent i, the second agent i sends the second agent state representation data To the first agent, agents i and j exchange Reservations i and j In this step, if an external eavesdropper only knows Unable to deduce s i and j The specific value of .

[0064] The first agent j represents data based on the state of the second agent and the first trusted random number a j Determine the first secret value parameter d j , And represent the data according to the first agent state and the second trusted random number b j Determine the second secret value parameter e j , The first trusted random number a j and the second trusted random number b j It is sent to the first agent j in advance by a third-party trusted organization.

[0065] The second agent i according to the first random number and the third trusted random b i The third secret value parameter e is determined by i , According to the second random number and the fourth trusted random number a i Determine the fourth secret value parameter d i , The third trusted random number b i and the fourth trusted random number a iIt is sent to the second intelligent agent i in advance by a third-party trusted organization.

[0066] Step S3: sharing the secret value parameter of the first agent and the secret value parameter of the second agent, and determining the shared secret value of the first agent and the second agent according to the secret value parameter of the first agent and the secret value parameter of the second agent;

[0067] For example, the second agent i and the first agent j share the secret value parameter d with each other through the channel i ,e i With d j ,e j , and according to the first secret value parameter d j and the fourth secret value parameter d i , determine the first shared secret value d, d = d i +d j ; According to the second secret value parameter e j and the third secret value parameter e i , determine the second shared secret value e, e = e i +e j .

[0068] Step S4: the first agent updates the virtual state of the first agent according to the shared secret value with the plurality of second agents;

[0069] Repeat steps 1 to 4 above until the entire system reaches consensus.

[0070] For example, the second agent i uses the shared secret value d, e and a trusted random number set (a i ,b i ,c i ), determine the second agent secret state value z i , specifically:

[0071] z i =de+db i +ea i +c i ;

[0072] The first intelligent agent j uses the shared secret value d, e and a trusted random number set (a j ,b j ,c j ), determine the secret state value z of the first agent j , specifically:

[0073] z j =db j +ea j +cj ;

[0074] The first agent j receives the second agent secret state value z sent by the second agent i i , and according to the second agent secret state value z i and the first agent's secret state value z j , determine the virtual state difference x' between the first agent and the second agent i -x' j .

[0075] Specifically, firstly, the second agent secret state value z i and the first agent's secret state value z j Add and extract the exponential function that carries the virtual state difference information of the first agent and the second agent Then, the logarithm of the exponential function is taken to obtain the virtual state difference between the first agent and the second agent, including:

[0076]

[0077] x i '-x' j = log(z i +z j );

[0078] Among them, z j represents the secret state value of the first agent, z i represents the secret state value of the second agent, z i =de+db i +ea i +c i , z j =db j +ea j +c j , d represents the first shared secret value, d = d i +d j , represents the second random number generated by the second agent, represents the state representation data of the first agent, x' i represents the virtual state of the second agent determined in the previous round, e represents the second shared secret value, e = e i +e j , represents the first random number generated by the first agent, represents the state representation data of the first agent, x' j represents the virtual state of the first agent determined in the previous round, (a i ,b i ,c i ) represents a set of trusted random numbers issued by a third-party trusted institution to the second agent, (a j ,b j ,c j ) represents a set of trusted random numbers issued to the first agent by a third-party trusted institution; x' i -x' j Represents the difference in virtual state between the first agent and the second agent.

[0079] The first agent updates the virtual state of the first agent according to the difference between the virtual states of the first agent and the second agent.

[0080] Specifically, the following formula is used to complete the virtual state update of the first agent:

[0081]

[0082] Among them, x j' (k+1) represents the virtual state of the first agent in round K+1, x j' (k) represents the virtual state of the first agent in round K, represents a constant step size. Δ is the maximum degree of graph G, N j represents the total number of adjacent second agents that are connected to the first agent for communication, a ij (k) represents the weighted adjacency matrix parameter. If the first agent is not a neighbor of the second agent, then a ij (k) = 0, if the first agent is in communication with the second agent, then there exists a positive number η such that a ij (k)∈[η,1] holds for all k, x i' (k)-x j' (k) represents the virtual state difference between the first agent and the second agent in round K.

[0083] According to the definition:

[0084]

[0085] In an undirected topology, therefore:

[0086]

[0087] The method provided in this embodiment can achieve accurate consensus.

[0088] The present invention provides an average consensus privacy protection method based on random additive sharing, which ensures that agents share changes in their states and reach average consensus without revealing their privacy by using random numbers to generate virtual states. When each agent is connected to at least one trusted agent, it can resist security attacks from honest but curious agents and global eavesdropping agents, and the method only relies on simple mathematical operations such as addition and multiplication of constants, as well as exponential and logarithmic calculations. Despite the simplicity of calculation, the method is not limited to the positive integer states of the agent, but also includes decimals and negative values. A large amount of encryption processing is avoided through simple mathematical operations.

[0089] As an optional implementation, Figure 1 As shown, the process includes an initialization phase before the first agent generates a first random number and the second agent generates a second random number. Specifically, the first agent generates a set of random numbers and the second agent generates a set of random numbers; the first agent determines the initial virtual state of the first agent based on the generated set of random number arrays and the random number array generated by the second agent.

[0090] For example, at the system initialization stage, each agent generates a set of random numbers where i N is the Nth neighbor of agent i, and then collects A trusted third party is required in the preparation stage. Two adjacent first agents j and second agents i try to share information. They are each assigned a set of data (a i ,b i ,c i ) and (a j ,b j ,c j ). These datasets are generated from positive tuples (a,b,c):

[0091] a=a i +a j , b=b i +b j , c=ab=c i +c j

[0092] During the initialization process of k = 0, the first agent j passes Calculate its virtual state, i∈N j Indicates that agent i is a neighboring agent that is connected to the first agent through communication; the second agent i is connected to the first agent through communication Calculate its virtual state, j∈N i It means that agent j is a neighboring agent that is in communication connection with the second agent i.

[0093] In order to further illustrate the effect of this solution, the simulation results are given below for reference:

[0094] In the simulation of the method provided in the embodiment of the present invention, a total of 5 intelligent agents are set, and their initial states are set to [1, 4, 3, 6, 8].

[0095] Figure 2 The multi-agent state change process of the non-privacy protection scheme and the method provided by the embodiment of the present invention is demonstrated. Simulation proves that the invention can achieve accurate consensus without affecting the convergence speed.

[0096] Figure 3 The performance comparison between the method provided by the embodiment of the present invention and the homomorphic encryption scheme under different topological complexities is shown. As the network topology complexity increases, the homomorphic encryption scheme shows a significant increase in latency and bandwidth due to encryption and decryption. In contrast, the scheme proposed by the method provided by the embodiment of the present invention maintains lightweight consumption, with latency and data size within the range of 4 milliseconds and 1000 bytes, respectively. The method provided by the embodiment of the present invention has significant advantages such as lightweight implementation, accurate consensus, and a wide range of applications.

[0097] In summary, the present invention proposes a privacy-preserving average consensus method based on additive secret sharing. The method provided in the embodiment of the present invention replaces the complex homomorphic encryption with a simple arithmetic secret sharing method to protect the state change value of the adjacent intelligent agent in each iteration. In addition, the method provided in the embodiment of the present invention uses a random mechanism to protect the state value of the adjacent intelligent agent. The method provided in the embodiment of the present invention has obvious advantages in computing, communication, storage and signaling overhead, can resist physical attacks and various network attacks at the same time, and has broad application prospects and market value.

[0098] The embodiment of the present invention provides an average consensus privacy protection device based on random additive sharing, comprising:

[0099] The random number generation module is used for the first agent to generate a first random number and the second agent to generate a second random number. The second agent is any one of the adjacent agents that are communicatively connected to the first agent. For details, please refer to the corresponding part of the above method embodiment, which will not be repeated here.

[0100] The secret value parameter determination module is used to perform additive operations using the first random number, the second random number, the first agent's own virtual state determined in the previous round, and the second agent's own virtual state determined in the previous round as parameters to determine the secret value parameters of the first agent and the secret value parameters of the second agent; please refer to the corresponding part of the above method embodiment for details, which will not be repeated here.

[0101] The secret value sharing module is used to determine the shared secret value of the first agent and the second agent based on the secret value parameters of the first agent and the secret value parameters of the second agent; please refer to the corresponding part of the above method embodiment for details, which will not be repeated here.

[0102] The virtual state updating module is used for the first agent to update the virtual state of the first agent according to the shared secret value with multiple second agents; please refer to the corresponding part of the above method embodiment for details, which will not be repeated here.

[0103] The consensus module is used to repeat the steps of the random number generation module to the virtual state update module until the entire system reaches a consensus. For details, please refer to the corresponding part of the above method embodiment, which will not be repeated here.

[0104] The present application also provides an electronic device, such as Figure 4 As shown, a processor 501 and a memory 502, wherein the processor 501 and the memory 502 may be connected via a bus or other means.

[0105] The processor 501 may be a central processing unit (CPU). The processor 501 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0106] The memory 502, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs and modules, such as program instructions / modules corresponding to the average consensus privacy protection method based on random additive sharing in the embodiment of the present invention. The processor executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions and modules stored in the memory.

[0107] The memory 502 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the processor, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 502 may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0108] The one or more modules are stored in the memory 502, and when executed by the processor 501, the following is performed: Figure 1 The average consensus privacy protection method based on random additive sharing in the illustrated embodiment.

[0109] For details of the above electronic equipment, please refer to Figure 1 The corresponding related descriptions and effects in the illustrated embodiments can be understood and will not be repeated here.

[0110] This embodiment also provides a computer storage medium, which stores computer executable instructions, and the computer executable instructions can execute the average consensus privacy protection method based on random additive sharing in any of the above method embodiments. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above types of memory.

[0111] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. An average consensus privacy protection method based on random additive sharing, characterized in that: include: Step 1: The first agent generates a first random number, and the second agent generates a second random number, where the second agent is any one of the adjacent agents that are in communication connection with the first agent; Step 2: Perform an additive operation using the first random number, the second random number, the virtual state of the first agent determined in the previous round, and the virtual state of the second agent determined in the previous round as parameters to determine the secret value parameter of the first agent and the secret value parameter of the second agent; Step 3: Share the secret value parameter of the first agent and the secret value parameter of the second agent, and determine the shared secret value of the first agent and the second agent according to the secret value parameter of the first agent and the secret value parameter of the second agent; Step 4: The first agent updates the virtual state of the first agent according to the shared secret value with the plurality of second agents; Repeat steps 1 to 4 above until the entire system reaches consensus.

2. According to claim 1, the average consensus privacy protection method based on random additive sharing is characterized in that: The first random number, the second random number, the virtual state of the first agent determined in the previous round, and the virtual state of the second agent determined in the previous round are used as parameters to perform an additive operation to determine the secret value parameters of the first agent and the second agent, including: The first agent determines the first agent state representation data according to its own virtual state determined in the previous round and the first random number; The second agent determines the second agent state representation data according to its own virtual state determined in the previous round and the second random number; The first agent sends a first random number to the second agent, and the second agent sends second agent state representation data to the first agent; The first agent determines a first secret value parameter according to the second agent state representation data and a first trusted random number, and determines a second secret value parameter according to the first agent state representation data and a second trusted random number, wherein the first trusted random number and the second trusted random number are sent to the first agent in advance by a third-party trusted institution; The second agent determines a third secret value parameter based on the first random number and the third trusted random number, and determines a fourth secret value parameter based on the second random number and the fourth trusted random number. The third trusted random number and the fourth trusted random number are pre-issued to the second agent by a third-party trusted organization.

3. The average consensus privacy protection method based on random additive sharing according to any one of claim 2, characterized in that: Determining a shared secret value between the first agent and the second agent according to a secret value parameter of the first agent and a secret value parameter of the second agent, comprising: Determining a first shared secret value according to the first secret value parameter and the fourth secret value parameter; A second shared secret value is determined based on the second secret value parameter and the third secret value parameter.

4. According to claim 3, the average consensus privacy protection method based on random additive sharing is characterized in that: The first agent updates the virtual state of the first agent according to the shared secret value with the plurality of second agents, including: The second agent determines the second agent secret state value according to the shared secret value and a set of trusted random numbers issued to the second agent by a third-party trusted institution in advance; The first agent determines the secret state value of the first agent according to the shared secret value and a set of trusted random numbers issued to the first agent by a third-party trusted institution in advance; The first agent receives the second agent secret state value sent by the second agent, and determines the virtual state difference between the first agent and the second agent according to the second agent secret state value and the first agent secret state value; The first agent updates the virtual state of the first agent according to the difference between the virtual states of the first agent and the second agent.

5. According to claim 1, the average consensus privacy protection method based on random additive sharing is characterized in that: The first agent generates a first random number, and the second agent generates a second random number before: The first agent generates a set of random numbers, and the second agent generates a set of random numbers; The first agent determines the initial virtual state of the first agent based on a set of random arrays generated and a random array generated by the second agent.

6. According to claim 4, the average consensus privacy protection method based on random additive sharing is characterized in that: Determining a virtual state difference between the first agent and the second agent according to the second agent secret state value and the first agent secret state value, comprising: x i '-x' y =log(z i +z j ); Among them, z j represents the secret state value of the first agent, z i represents the secret state value of the second agent, z i =de+db i +ea i +c i , z j =db j +ea j +c j , d represents the first shared secret value, d = d i +d j , represents the second random number generated by the second agent, represents the state representation data of the second agent, x′ i represents the virtual state of the second agent determined in the previous round, e represents the second shared secret value, e = e i +e j , represents the first random number generated by the first agent, represents the state representation data of the first agent, x′ j represents the virtual state of the first agent determined in the previous round, (a i ,b i ,c i ) represents a set of trusted random numbers issued by a third-party trusted institution to the second agent, (a j ,b j ,c j ) represents a set of trusted random numbers issued by a third-party trusted institution to the first agent; x′ i -x′ j Represents the difference in virtual state between the first agent and the second agent.

7. The average consensus privacy protection method based on random additive sharing according to claim 5 or 6, characterized in that: The first agent updates the virtual state of the first agent according to the difference between the virtual states of the first agent and the second agent, including: Among them, x j' (k+1) represents the virtual state of the first agent in round K+1, x j' (k) represents the virtual state of the first agent in round K, represents a constant step size, Δ is the maximum degree of graph G, N j represents the total number of adjacent second agents that are connected to the first agent for communication, a ij (k) represents the weighted adjacency matrix parameter. If the first agent is not a neighbor of the second agent, then a ij (k) = 0, if the first agent is in communication with the second agent, then there exists a positive number η such that a ij (k)∈[η,1] holds for all k, x i' (k)-x j' (k) represents the virtual state difference between the first agent and the second agent in round K.

8. An average consensus privacy protection device based on random additive sharing, characterized in that: include: A random number generation module, used for a first agent to generate a first random number, and a second agent to generate a second random number, where the second agent is any one of the adjacent agents in communication connection with the first agent; A secret value parameter determination module, used to perform an additive operation using the first random number, the second random number, the virtual state of the first agent determined in the previous round, and the virtual state of the second agent determined in the previous round as parameters to determine the secret value parameter of the first agent and the secret value parameter of the second agent; A secret value sharing module, used for determining a secret value shared by the first agent and the second agent according to a secret value parameter of the first agent and a secret value parameter of the second agent; A virtual state updating module, used for the first agent to update the virtual state of the first agent according to the shared secret value with the plurality of second agents; The consensus module is used to repeat the above steps from the random number generation module to the virtual state update module until the entire system reaches a consensus.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the steps of the average consensus privacy protection method based on random additive sharing as described in any one of claims 1-7 above.

10. A computer storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by the processor, the steps of the average consensus privacy protection method based on random additive sharing are implemented as described in any one of claims 1-7.

Images