Electric power data fingerprint evaluation system based on Bayesian network

Abstract

The invention discloses an electric power data fingerprint evaluation system based on a Bayesian network, relates to an electric power data fingerprint evaluation device, and aims to solve the problem that an electric power data fingerprint evaluation result is inaccurate due to the fact that an existing evaluation method cannot objectively reflect information of electric power data fingerprints. A fingerprint analysis module is used for performing performance analysis on electric power data and selecting an alternative index set; the screening and evaluating module is used for screening the alternative index set, screening out evaluation indexes influencing the fingerprint security of the power data, performing analytic hierarchy process on the evaluation indexes, and determining the relationship strength among the node indexes; the index quantification module is used for quantitatively giving the prior probability of each node index; the model construction module is used for constructing a structure of a Bayesian network model; the parameter learning module is used for giving Bayesian network parameters; the cluster tree conversion module is used for generating a cluster tree; and the result output module is used for outputting a data fingerprint evaluation result. The beneficial effect is that the accuracy of the evaluation result is improved.

Description

Technical Field

[0001] The invention relates to a power data fingerprint evaluation device. Background Art

[0002] With the increasing demand for power data encryption, power data fingerprint, as a new and effective data encryption and traceability technology, has been increasingly widely used in real life. The performance of power data fingerprint represents the degree of protection of power data, so the evaluation of power data fingerprint has strong practical significance. Objective and efficient evaluation methods can effectively guide decision makers to select the data fingerprint scheme with the best performance to achieve encryption, traceability and protection of power information; in order to ensure the security of power data, power data fingerprint technology is applied to control data distribution in power data; power data containing power data fingerprint faces an uncertain environment during distribution, so it is necessary to evaluate the performance indicators of power data fingerprint, so as to obtain useful information and select appropriate power data fingerprint technology.

[0003] Combined with the current research status of domestic and foreign scholars, power data fingerprints have the following conventional indicators: capacity, robustness, imperceptibility, reliability, and anti-exploitation. In real scenarios, it is also necessary to consider the impact of the embedded position on the data itself, the encryption and decryption calculation efficiency of the data fingerprint, the versatility of the data fingerprint scheme, expert opinions, etc. For the above multiple indicators, traditional single evaluation methods such as the analytic hierarchy process and the entropy weight method cannot accurately and objectively complete the evaluation task; the TOPSIS method proposes the concept of ideal solution closeness to rank the evaluation objects. However, since the ideal solution of the evaluation object is difficult to give, it is not suitable for data fingerprint evaluation; the gray analytic hierarchy process combines gray theory and analytic hierarchy process to quantify fuzzy indicators. The determination of its weight is determined by the classical analytic hierarchy process, which belongs to the category of subjective evaluation and cannot objectively reflect the information of the data fingerprint itself.

[0004] In view of this, it is necessary to find a suitable system to evaluate the power data fingerprint. Summary of the invention

[0005] The purpose of the present invention is to solve the problem that the existing evaluation method cannot objectively reflect the information of power data fingerprint itself, resulting in inaccurate power data fingerprint evaluation results, and proposes a power data fingerprint evaluation system based on Bayesian network.

[0006] The power data fingerprint evaluation system based on Bayesian network described in the present invention comprises a fingerprint analysis module, a screening evaluation module, an indicator quantification module, a model construction module, a parameter learning module, a clique tree conversion module and a result output module;

[0007] Fingerprint analysis module, used to perform performance analysis on power data, obtain evaluation data in the security evaluation results, and select candidate indicator sets based on the evaluation data and daily usage of power data;

[0008] The screening and evaluation module is used to screen the candidate indicator set selected by the fingerprint analysis module using the entropy weight method, screen out the evaluation indicators that affect the security of the power data fingerprint, and perform hierarchical analysis on each evaluation indicator to determine the relationship strength between each node indicator;

[0009] The indicator quantification module is used to quantitatively give the prior probability of each node indicator based on the evaluation indicators screened by the screening and evaluation module;

[0010] A model building module is used to build the structure of the Bayesian network model according to the relationship strength between the node indicators determined by the screening and evaluation module and the prior probability of each node indicator given by the indicator quantification module;

[0011] A parameter learning module is used to perform Bayesian network parameter learning on the structure of the Bayesian network model constructed by the model building module by utilizing the distribution characteristics of the integrated power data, and to provide the Bayesian network parameters;

[0012] The clique tree conversion module is used to convert the Bayesian network parameters given by the parameter learning module to generate a clique tree;

[0013] The result output module is used to use the cluster tree propagation algorithm to infer the cluster tree generated by the cluster tree conversion module and output the power data fingerprint evaluation result.

[0014] Furthermore, the specific method for the screening and evaluation module to determine the strength of the relationship between the node indicators is:

[0015] The selected evaluation indicators that affect the security of power data fingerprint are subjected to hierarchical analysis, and these factors are divided into several groups according to their attributes and dominant relationships to form different levels; then the strength of the relationship between each node is determined.

[0016] Furthermore, the specific method for the indicator quantification module to quantitatively give the prior probability of each node indicator is:

[0017] Combined with the performance of power data on various indicators, numerical representation is given, and expert survey method is used for each node to determine the value range and weight jump value, and a weight coefficient selection table and selection instructions are compiled. Combined with the results given by power experts, the prior probability of each node is weighted.

[0018] Furthermore, the method in which the model building module constructs the structure of the Bayesian network model is:

[0019] According to the characteristics of power data, the normal distribution model is used to represent the possibility of parameter values ​​given by expert knowledge; the target optimization model is constructed, and the Beta distribution is used to approximate the normal distribution of prior knowledge, and the parameter values ​​of the Beta distribution, namely the virtual sample statistics, are calculated; the virtual sample statistics are combined with the existing small sample data set, and the maximum a posteriori probability method of Bayesian estimation is used to calculate the estimated values ​​of the parameters, and the structure of the Bayesian network model is constructed.

[0020] Furthermore, the specific method of generating a cluster tree by the cluster tree conversion module is:

[0021] Establish a Bayesian network moral graph, triangulate the Bayesian network moral graph, determine the clique nodes, and generate the clique tree.

[0022] Furthermore, the specific method for the result output module to output the power data fingerprint evaluation result is:

[0023] Initialize algorithm parameters, absorb messages, calculate edge probabilities, and give output power data fingerprint evaluation results.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The present invention analyzes the basic process of data fingerprint structure through the fingerprint analysis module, and the screening and evaluation module uses the entropy weight method to screen the candidate indicator set, so as to be used for the model construction module to construct the power data fingerprint evaluation indicator system, so that the indicator system is simpler and more reasonable, and the efficiency is greatly improved; the parameter learning module uses the Bayesian network parameter learning method integrating expert prior knowledge to perform parameter learning of the Bayesian network, which greatly improves the usability; the clique tree conversion module and the result output module use the clique tree propagation algorithm to perform accurate reasoning of the Bayesian network, which saves the reasoning time and improves the accuracy of the power data fingerprint evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a structural block diagram of a power data fingerprint evaluation system based on a Bayesian network as described in a specific implementation method 1. DETAILED DESCRIPTION

[0027] Specific implementation method 1. Combination Figure 1 The present embodiment is described. The power data fingerprint evaluation system based on Bayesian network described in the present embodiment is characterized in that the power data fingerprint evaluation system includes a fingerprint analysis module 1, a screening evaluation module 2, an indicator quantification module 3, a model construction module 4, a parameter learning module 5, a clique tree conversion module 6 and a result output module 7;

[0028] Fingerprint analysis module 1 is used to perform performance analysis on power data, obtain evaluation data in the security evaluation results, and select a candidate indicator set based on the evaluation data and daily usage of power data;

[0029] The screening and evaluation module 2 is used to screen the candidate indicator set selected by the fingerprint analysis module 1 using the entropy weight method, screen out the evaluation indicators that affect the security of the power data fingerprint, and perform hierarchical analysis on each evaluation indicator to determine the relationship strength between each node indicator;

[0030] The indicator quantification module 3 is used to quantitatively give the prior probability of each node indicator according to each evaluation indicator screened by the screening and evaluation module 2;

[0031] The model building module 4 is used to build the structure of the Bayesian network model according to the relationship strength between the node indicators determined by the screening and evaluation module 2 and the prior probability of each node indicator given by the indicator quantification module 3;

[0032] The parameter learning module 5 is used to perform Bayesian network parameter learning on the structure of the Bayesian network model constructed by the model building module 4 by utilizing the distribution characteristics of the integrated power data, and to provide the Bayesian network parameters;

[0033] A clique tree conversion module 6 is used to convert the Bayesian network parameters given by the parameter learning module 5 to generate a clique tree;

[0034] The result output module 7 is used to use the cluster tree propagation algorithm to infer the cluster tree generated by the cluster tree conversion module 6 and output the power data fingerprint evaluation result.

[0035] In this embodiment, a power data fingerprint evaluation system based on Bayesian network includes the following three parts: fingerprint analysis module 1 establishes power data fingerprint evaluation index system, model construction module 4 establishes Bayesian network evaluation model and result output module 7 performs Bayesian network reasoning; by analyzing the specific process of power data fingerprint, the evaluation index is screened from each index using entropy weight method to construct a simple and reasonable power data fingerprint evaluation index system; the dependency relationship of each index node is distinguished, and according to the characteristics of power data, the maximum a posteriori probability method of Bayesian estimation is used to calculate the estimated value of the parameter after the Beta distribution is used to approximate the normal distribution, and the Bayesian network parameter learning is completed, and the Bayesian network structure is given. A reasonable Bayesian network structure is conducive to accurate reasoning and evaluation; power data has high confidentiality. In order to solve the problem of only small sample data, the expert scoring method is used to quantitatively give the prior probability of each node index for parameter learning; and then the clique tree propagation algorithm is used to perform precise reasoning of the Bayesian network, which saves the reasoning time and improves the accuracy of the power data fingerprint evaluation results.

[0036] Specific implementation method 2: This implementation method further limits the power data fingerprint evaluation system based on Bayesian network described in specific implementation method 1. In this implementation method, the screening evaluation module 2 determines the specific method of the relationship strength between each node indicator as follows:

[0037] The selected evaluation indicators that affect the security of power data fingerprint are subjected to hierarchical analysis, and these factors are divided into several groups according to their attributes and dominant relationships to form different levels; then the strength of the relationship between each node is determined.

[0038] In this implementation, the basic performance of power data fingerprint includes fidelity, effectiveness, security, robustness, capacity, efficiency, and anti-exploitation. Among them, fidelity includes imperceptibility; robustness is based on attacks, and the attacks faced include individual attacks and multi-person attacks. Single-person attacks include removal attacks, synchronization attacks, and protocol attacks. Synchronous attacks include geometric distortion attacks, mosaic attacks, jitter attacks, and Oracle attacks. Protocol attacks include interpretation attacks and copy attacks; reliability includes false detection rate, missed detection rate, traceability, and provability; anti-exploitation includes communication confidentiality, encryption algorithm confidentiality, and embedded security.

[0039] Specific implementation method 3: This implementation method further limits the power data fingerprint evaluation system based on Bayesian network described in specific implementation method 2. In this implementation method, the specific method for the index quantification module 3 to quantitatively give the prior probability of each node index is:

[0040] Combined with the performance of power data on various indicators, numerical representation is given, and expert survey method is used for each node to determine the value range and weight jump value, and a weight coefficient selection table and selection instructions are compiled. Combined with the results given by power experts, the prior probability of each node is weighted.

[0041] Specific implementation method 4: This implementation method further limits the power data fingerprint evaluation system based on Bayesian network described in specific implementation method 3. In this implementation method, the method for the model building module 4 to construct the structure of the Bayesian network model is:

[0042] According to the characteristics of power data, the normal distribution model is used to represent the possibility of parameter values ​​given by expert knowledge; the target optimization model is constructed, and the Beta distribution is used to approximate the normal distribution of prior knowledge, and the parameter values ​​of the Beta distribution, namely the virtual sample statistics, are calculated; the virtual sample statistics are combined with the existing small sample data set, and the maximum a posteriori probability method of Bayesian estimation is used to calculate the estimated values ​​of the parameters, and the structure of the Bayesian network model is constructed.

[0043] Specific implementation mode 5: This implementation mode further limits the power data fingerprint evaluation system based on Bayesian network described in specific implementation mode 4. In this implementation mode, the specific method for generating a cluster tree by the cluster tree conversion module 6 is:

[0044] Establish a Bayesian network moral graph, triangulate the Bayesian network moral graph, determine the clique nodes, and generate the clique tree.

[0045] In this embodiment, all nodes and edges in the original Bayesian network are retained. If a node in the original Bayesian network has more than two parent nodes, all parent nodes of the node are connected in pairs in the moral graph to establish a Bayesian network moral graph. When triangulating the Bayesian network moral graph, if there are rings with more than 3 nodes in the moral graph, an undirected edge is added to connect two non-adjacent nodes in the ring. If there are still rings with more than 3 nodes, the triangulation is continued until a triangulated graph is formed. The clique node is determined, and the clique node is a maximal complete subgraph in the triangulated graph. A clique tree is generated, and each node in the clique tree corresponds to a clique node, and the intersection of two clique nodes is used as a separator node.

[0046] Specific implementation method 6: This implementation method further limits the power data fingerprint evaluation system based on Bayesian network described in specific implementation method 5. In this implementation method, the specific method of the result output module 7 outputting the power data fingerprint evaluation result is:

[0047] Initialize algorithm parameters, absorb messages, calculate edge probabilities, and give output power data fingerprint evaluation results.

[0048] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A power data fingerprint evaluation system based on Bayesian network, characterized in that: The power data fingerprint evaluation system comprises a fingerprint analysis module (1), a screening evaluation module (2), an indicator quantification module (3), a model construction module (4), a parameter learning module (5), a clique tree conversion module (6) and a result output module (7); A fingerprint analysis module (1) is used to perform performance analysis on power data, obtain evaluation data in the security evaluation results, and select a candidate indicator set based on the evaluation data and daily usage of power data; A screening and evaluation module (2) is used to screen the candidate indicator set selected by the fingerprint analysis module (1) using an entropy weight method, screen out the evaluation indicators that affect the security of the power data fingerprint, and perform a hierarchical analysis on the evaluation indicators to determine the strength of the relationship between the indicators at each node; The indicator quantification module (3) is used to quantitatively give the prior probability of each node indicator according to each evaluation indicator selected by the screening evaluation module (2); A model building module (4) is used to build a structure of a Bayesian network model according to the relationship strength between the node indicators determined by the screening and evaluation module (2) and the prior probability of each node indicator given by the indicator quantification module (3); A parameter learning module (5) is used to perform Bayesian network parameter learning on the structure of the Bayesian network model constructed by the model construction module (4) by utilizing the distribution characteristics of the integrated power data, and to provide Bayesian network parameters; A cluster tree conversion module (6) is used to convert the Bayesian network parameters given by the parameter learning module (5) to generate a cluster tree; The result output module (7) is used to use the cluster tree propagation algorithm to infer the cluster tree generated by the cluster tree conversion module (6) and output the power data fingerprint evaluation result.

2. According to the Bayesian network-based power data fingerprint evaluation system of claim 1, it is characterized in that: The specific method for the screening and evaluation module (2) to determine the relationship strength between each node indicator is: The selected evaluation indicators that affect the security of power data fingerprint are subjected to hierarchical analysis, and these factors are divided into several groups according to their attributes and dominant relationships to form different levels; then the strength of the relationship between each node is determined.

3. The power data fingerprint evaluation system based on Bayesian network according to claim 2 is characterized in that: The specific method of the indicator quantification module (3) to quantitatively give the prior probability of each node indicator is: Combined with the performance of power data on various indicators, numerical representation is given, and expert survey method is used for each node to determine the value range and weight jump value, and a weight coefficient selection table and selection instructions are compiled. Combined with the results given by power experts, the prior probability of each node is weighted.

4. The power data fingerprint evaluation system based on Bayesian network according to claim 3 is characterized in that: The method for constructing the structure of the Bayesian network model in the model construction module (4) is as follows: According to the characteristics of power data, the normal distribution model is used to represent the possibility of parameter values ​​given by expert knowledge; the target optimization model is constructed, the Beta distribution is used to approximate the normal distribution of prior knowledge, and the parameter value of the Beta distribution, that is, the virtual sample statistical value, is calculated; The virtual sample statistics are combined with the existing small sample data set, and the maximum a posteriori probability method of Bayesian estimation is used to calculate the estimated values ​​of the parameters to construct the structure of the Bayesian network model.

5. The power data fingerprint evaluation system based on Bayesian network according to claim 4 is characterized in that: The specific method of generating a cluster tree by the cluster tree conversion module (6) is as follows: Establish a Bayesian network moral graph, triangulate the Bayesian network moral graph, determine the clique nodes, and generate the clique tree.

6. The power data fingerprint evaluation system based on Bayesian network according to claim 5 is characterized in that: The specific method of the result output module (7) outputting the power data fingerprint evaluation result is: Initialize algorithm parameters, absorb messages, calculate edge probabilities, and give output power data fingerprint evaluation results.

Images