Watermarking method based on mutual information screening and IGWO optimization

Abstract

The invention relates to a watermarking method based on mutual information screening and IGWO optimization, and the method comprises the following steps: 1, carrying out the attribute screening through a mutual information method, obtaining an attribute set which is highly related to the classification accuracy, determining an unimportant attribute set, and forming a complete attribute set after the mutual information screening; step 2, based on the attribute set screened by the mutual information attributes in the step 1, carrying out watermark embedding operation for an unimportant attribute set to obtain a watermark data set containing a watermark; and step 3, based on the watermark data set generated in the step 2, watermark data are classified through a decision tree model, and the watermark embedding strength and position are optimized by applying an IGWO algorithm. The method can effectively solve the problem of traceability responsibility of matched power grid service data.

Description

Technical Field

[0001] The invention belongs to the technical field of digital watermarking, and relates to a watermarking method, in particular to a watermarking method based on mutual information screening and IGWO optimization. Background Art

[0002] With the integration of global networks, the generation, extraction and flow of data have become more convenient and faster, and the security control of data has become increasingly difficult. Especially in the power system, power stations, substations, power dispatching centers, etc. are transmitting information all the time. In addition to some authorized public information, it also contains some unauthorized private information. If this kind of data information is leaked, it is very likely that intruders will steal user personal information and cause a large-scale accident.

[0003] In order to better carry out power system management and achieve the goal of "pre-prevention, in-process control, and post-event tracing" in the process of power system data transmission, it is extremely necessary to hide and protect this type of information.

[0004] At present, the data traceability model can solve the accountability problem of data leakage to a certain extent. Among the existing data traceability technologies, digital watermarking technology has high security, strong resistance to repeated additions, and can still ensure partial integrity after signal processing. Based on the authenticity, transparency and undetectability of digital watermarks, it is widely used in data traceability issues in power systems.

[0005] The power system is different from the application scenarios of remote sensing images and three-dimensional networks. It has higher requirements for data security. Once privacy data is leaked, the losses to the government and enterprises will be greater. Its uniqueness further exacerbates the application challenges of watermark technology.

[0006] Existing digital watermarking methods have multiple defects, mainly including the risk of data leakage, insufficient consideration of data availability when enhancing watermark security, failure to consider the availability impact caused by data distribution characteristics in key management, and the complexity of the watermark extraction process. In addition, although some algorithms are robust to specific attacks, they still need to be improved overall to cope with a variety of attacks, especially in high-voltage environments. At the same time, some algorithms fail to effectively improve the efficiency of traceability information query while improving security and robustness, resulting in performance bottlenecks. These problems show that the existing digital watermarking technology applied to power systems still needs to be further optimized and improved in terms of availability, security, complexity and efficiency.

[0007] In summary, although the existing digital watermarking algorithms have analyzed the data security in power application scenarios well, they lack the importance analysis of the attribute characteristics of the watermark data, and it is difficult to effectively match the traceability and accountability issues of power grid business data; secondly, there is a lack of optimization of classification accuracy for the classification problem of watermark data, and most of them use clustering algorithms to mine the availability of data during data analysis, and in actual applications, there is a lack of analysis of the traceability process when data is leaked.

[0008] In order to solve the above problems, the present invention proposes a watermarking method based on mutual information screening and IGWO optimization.

[0009] After searching, no public documents of the prior art that are identical or similar to the present invention were found. Summary of the invention

[0010] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a watermarking method based on mutual information screening and IGWO optimization, which can effectively solve the traceability and accountability problem of matching power grid business data.

[0011] The present invention solves the practical problem by adopting the following technical solutions:

[0012] A watermarking method based on mutual information screening and IGWO optimization comprises the following steps:

[0013] Step 1: Use the mutual information method to screen attributes, obtain the attribute set that is highly correlated with classification accuracy, and determine the unimportant attribute set to form a complete attribute set after mutual information screening;

[0014] Step 2: Based on the attribute set selected by the mutual information attribute in step 1, a watermark embedding operation is performed on the unimportant attribute set to obtain a watermark data set containing the watermark;

[0015] Step 3: Based on the watermark data set generated in step 2, the watermark data is classified through the decision tree model, and the IGWO algorithm is applied to optimize the watermark embedding strength and position.

[0016] Moreover, the specific steps of step 1 include:

[0017] (1) The concept of mutual information is used to measure the relationship between attributes and classification labels. First, the input continuous value attribute data is read and then divided into several intervals. The equal frequency or equal width method can be selected according to the number of samples.

[0018] (2) Number each interval and convert the continuous attribute into the corresponding discrete interval value;

[0019] (3) Calculate the frequency distribution of each interval to ensure that the interval division is reasonable, and normalize the discretized attribute data so that all attribute values ​​are in the same range. The processed attribute data will be stored to generate a new discrete attribute data set, thereby obtaining an attribute set that is highly correlated with classification accuracy; at the same time, filter out the attribute set with lower importance.

[0020] Moreover, the specific steps of step 2 include:

[0021] (1) Sort the mutual information attribute set selected based on step 1. The sorted data set will be divided into several intervals. The size and number of intervals are determined according to the sorting result;

[0022] (2) Determine the mapping interval corresponding to the data points in each attribute set through the sorting results, maintain the distribution consistency of the data before and after the watermark is embedded, and ensure that adding the watermark does not significantly change the statistical characteristics of the data;

[0023] (3) In the process of embedding watermarks into data points, the relative order of the data points is kept unchanged through interval mapping, so that the data distribution after watermark embedding is consistent with the original data distribution, and finally a watermarked data set containing the watermark is obtained.

[0024] Moreover, the specific steps of step 3 include:

[0025] (1) Based on the watermark data set generated in step 2, the watermark data is classified using a decision tree model;

[0026] (2) For the classified watermark data set, the IGWO algorithm is introduced to optimize the embedding strength and position of the watermark: the IGWO algorithm takes the digital watermark vector as the optimization object, calculates the fitness of each watermark vector, and helps determine the optimal watermark embedding scheme.

[0027] Moreover, the calculation formula of step (2) of step 3 is:

[0028] The digital watermark vector wm is used as the optimization object of the IGWO algorithm. According to different watermark vectors, the original data is converted into a watermark data set D according to the watermark embedding method proposed in step 2. w , and use D w The classification accuracy of is used as the fitness function;

[0029] max:accuracy(f(D w ))

[0030]

[0031] Among them, f(D w) is the classification result of the watermark data based on the decision tree model, and WM is the watermark set in the existing established watermark index table.

[0032] Advantages and beneficial effects of the present invention:

[0033] 1. In order to effectively match the traceability and accountability problem of power grid business data, this paper proposes a watermark algorithm based on mutual information screening and IGWO optimization (Watermark based on Mutual Information and Improved GreyWolf Optimization, WMIIGWO). The mutual information method is used to screen attributes, and the attribute set after mutual information screening is obtained. By comparing the importance of the attributes, the watermark is embedded in the unimportant attributes, which reduces the attribute dimension of the watermark data, thereby improving the data classification accuracy and avoiding the accuracy loss caused by watermark embedding in the traditional method.

[0034] 2. The present invention combines the idea of ​​interval mapping. Interval mapping sorts the screened attribute set and divides the data into several intervals to ensure the consistency of data distribution before and after watermark embedding, avoiding significant impact on the statistical characteristics of the data; maintaining the relative order of data points, reducing distortion, ensuring the accuracy of watermark extraction, and supporting data traceability.

[0035] 3. The present invention uses a decision tree model for classification, utilizes the IGWO algorithm to determine the best watermark selection result, and uses the digital watermark vector as the key optimization object in the optimization model to improve the concealment and robustness of the watermark. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a schematic diagram of attribute mutual information values ​​in a specific implementation manner of the present invention;

[0037] Figure 2 It is a watermark vector solution fitness curve diagram in a specific implementation manner of the present invention;

[0038] Figure 3 It is a comparison diagram of watermark data classification availability maintenance in a specific implementation manner of the present invention. DETAILED DESCRIPTION

[0039] The following further describes the embodiments of the present invention:

[0040] A watermarking method based on mutual information screening and IGWO optimization comprises the following steps:

[0041] Step 1: Use the mutual information method to screen attributes, obtain a set of attributes that are highly correlated with classification accuracy, and determine unimportant attribute sets to form a complete attribute set after mutual information screening.

[0042] The specific steps of step 1 include:

[0043] (1) The concept of mutual information is used to measure the relationship between attributes and classification labels. First, the input continuous value attribute data is read and then divided into several intervals. The equal frequency or equal width method can be selected according to the number of samples.

[0044] (2) Number each interval and convert the continuous attribute into the corresponding discrete interval value;

[0045] (3) Calculate the frequency distribution of each interval to ensure that the interval division is reasonable, and normalize the discretized attribute data so that all attribute values ​​are in the same range. The processed attribute data will be stored to generate a new discrete attribute data set, thereby obtaining an attribute set that is highly correlated with classification accuracy; at the same time, filter out the attribute set with lower importance.

[0046] The working principle of step 1 of the present invention is:

[0047] 1. Attribute screening based on mutual information

[0048] Mutual Information (MI) is an information measurement indicator used to measure the correlation between two random variables. In information theory, mutual information describes the degree to which the uncertainty of another random variable is reduced when the value of one random variable is known. As an effective correlation measure, it can reveal the degree of correlation between random variables and provides an important theoretical basis for data analysis and information processing. The mutual information between different attribute column data and classification label data is calculated to measure the correlation between the two. The specific formula is shown in (1):

[0049]

[0050] Among them, X j is the attribute column j data, Y is the real classification label data, P(X j ,Y)=P(X j |Y)×P(Y) is X j , the probability that Y appears simultaneously in the sample data, * represents the multiplication operation, P(x|y) represents the probability of occurrence of random variable x under random variable y, represents the entropy of variable x, Represents the joint entropy of variables x and y. The specific attribute screening based on mutual information is shown in Algorithm 1. It mainly uses the concept of mutual information to measure the relationship between attributes and classification labels to select attributes that contribute more to the accuracy of data classification and reduce the complexity of watermark calculation.

[0051]

[0052]

[0053] Step 2: Based on the attribute set screened by the mutual information attribute in step 1, a watermark embedding operation is performed on the unimportant attribute set to obtain a watermark data set containing the watermark.

[0054] The specific steps of step 2 include:

[0055] (1) Sort the mutual information attribute set selected based on step 1. The sorted data set will be divided into several intervals. The size and number of intervals are determined according to the sorting result;

[0056] (2) Determine the mapping interval corresponding to the data points in each attribute set through the sorting results, maintain the distribution consistency of the data before and after the watermark is embedded, and ensure that adding the watermark does not significantly change the statistical characteristics of the data;

[0057] (3) In the process of embedding watermarks into data points, the relative order of the data points is kept unchanged through interval mapping, so that the data distribution after watermark embedding is consistent with the original data distribution, and finally a watermarked data set containing the watermark is obtained.

[0058] The working principle of step 2 is:

[0059] When the original data D is forwarded, the watermark vector wm can uniquely identify the receiver of the data D according to the data receiver and the watermark index table. The forwarded original data D is only in the attribute set J. A Embed the watermark on the attribute column to obtain the watermark data D w , the watermark embedding method is shown in Algorithm 2:

[0060]

[0061]

[0062] Algorithm 2 determines the set J based on the data sorting results A Data point d in the attribute column ij The mapping interval (a ij ,b ij ), keep the data distribution consistent before and after adding watermark, and wm=(wmf 1 ,wmf 2 ,...wmf j ,...wmf F ), wmf j ∈(0,1), F=|J A |.

[0063] The WMIIGWO algorithm proposed in this invention requires that the original data sharing party, i.e., the original data source or the verification party, always holds the complete original data when tracing the data source, and can calculate the original data point d based on the original data. ij The mapping interval (a ij ,b ij ), based on formula (2), extract the data table D to be traced by the watermark x Each continuous value in the attribute watermark.

[0064]

[0065] Among them, n x Table D is the data table to be traced x The number of records in the table is calculated, and the digital watermark vector wm of the data table to be traced is constructed by each attribute watermark. x , wm x The data recipient can be retrieved by matching with the digital watermark index table.

[0066] Step 3: Based on the watermark data set generated in step 2, the watermark data is classified through the decision tree model, and the IGWO algorithm is applied to optimize the watermark embedding strength and position to improve the watermark concealment and data availability.

[0067] The specific steps of step 3 include:

[0068] (1) Based on the watermark data set generated in step 2, the watermark data is classified using a decision tree model;

[0069] (2) For the classified watermark data set, the IGWO algorithm is introduced to optimize the embedding strength and position of the watermark: the IGWO algorithm takes the digital watermark vector as the optimization object, calculates the fitness of each watermark vector, and helps determine the optimal watermark embedding scheme.

[0070] After optimization, the selected watermark vector can effectively control the embedding strength and position of the watermark, ensure the consistency of the data distribution after embedding, improve the concealment and robustness of the watermark, and provide a strong traceability and accountability basis for data leakage and illegal use in distribution network business, thereby ensuring the integrity and security of user data and enhancing data protection and compliance in business operations.

[0071] The working principle of step 3 is:

[0072] The traditional irreversible watermark embedding method is relatively fixed, which may cause the position and strength of the watermark to be easily detected and removed, reducing the concealment and robustness of the watermark, and the modification of the original data may cause the data availability to decrease. In the power marketing business scenario, the integrity and availability of user data are crucial. Based on the consideration of improving data availability and watermark concealment, the present invention optimizes the watermark embedding strength, determines the best watermark selection result based on the IGWO algorithm, and introduces a uniqueness constraint considering the feasibility of traceability. The present invention uses the digital watermark vector wm as the optimization object of the IGWO algorithm, and converts the original data into a watermark data set D according to the watermark embedding method proposed in step 2 according to different watermark vectors. w , and use D w The classification accuracy is used as the fitness function.

[0073] max:accuracy(f(D w ))

[0074]

[0075] Among them, f(D w ) is the classification result of the watermark data based on the decision tree model, and WM is the watermark set in the existing established watermark index table.

[0076] The working principle of the IGWO optimization algorithm in step 3 of the present invention is:

[0077] The GWO algorithm includes three stages: encirclement, pursuit, and attack. Considering that the algorithm is prone to fall into the local optimal problem during the optimization process, there are many variants of the GWO algorithm. In order to better improve the diversity of the uniform distribution of the population, the chaotic mapping mechanism is adopted to improve the diversity of the population; in order to speed up the optimization speed and improve the optimization accuracy, the position update mechanism of the population during the search process has also been partially improved. The specific improvements are as follows.

[0078] Improvement 1: Population initialization

[0079] In order to maintain the diversity of the population, the present invention uses the Tent chaotic map to initialize the individual positions of the gray wolves. This process can effectively avoid the centralization generated by the algorithm in the initial stage, thereby providing a wider search space for watermark embedding. The expression of the Tent chaotic map can be used to randomly generate the position of the gray wolf, laying the foundation for the subsequent optimization of the strength and position of the watermark. The expression of the Tent chaotic map is:

[0080]

[0081] Among them, x i,r represents the position of the gray wolf at the i-th iteration, and u represents the mapping parameter, which is 0.6 here.

[0082] Improvement 2: Nonlinear parameter control

[0083] In the process of watermark embedding, a nonlinear model is used to control the parameter a to ensure that the position change of the population in the optimization process is more flexible and effective. This nonlinear change helps to prevent the gray wolf from falling into the local optimum during the optimization process, thereby enhancing the concealment and robustness of the watermark.

[0084] a=2-2(t / T) 3 (5)

[0085] A=2(2r 1 -1)[1-(t / T) 3 ] (6)

[0086] The variable a represents the parameter that decreases from 2 to 0 during the iteration process, t represents the current number of iterations, T represents the maximum number of iterations, and r represents the maximum number of iterations. 1 ∈[0,1] represents a random factor.

[0087] Improvement 3: Weight-based gray wolf position update mechanism

[0088] In the improved IGWO algorithm, different weight coefficients are set to affect the position update of the gray wolf, so that the gray wolf with higher fitness has greater influence in the group. This mechanism enables the watermark vector with higher fitness to be used more effectively in the optimization process of watermark embedding, thereby improving the quality of the final embedded watermark data set. In the improved IGWO algorithm, the weight of the α wolf is higher than that of the β wolf and the δ wolf, and different weights are set according to their fitness to achieve position update.

[0089] X(t+1)=c 1 r 2 (ω α X 1 +ω β X 2 +ω δ X 3 )+c 2 r 3 (X α -X(t)) (7)

[0090]

[0091] Among them, c 1 =c 2 =0.6 represents the social behavior learning factor of gray wolves, ω α ,ω β ,ω δ Respectively represent the weight coefficients of α, β, δ gray wolf individuals to adjust the update of individual positions, X αrepresents the current optimal position of the gray wolf individual α, r 2 ,r 3 ∈[0,1] represents a random factor, ω j represents the updated weight of the j-th gray wolf, f(X j (t)) represents the fitness value of the jth gray wolf, f(X α (t)),f(X β (t)),f(X δ (t)) represent the fitness values ​​of α, β, and δ gray wolves respectively.

[0092] The present invention is further described below by means of specific examples:

[0093] The experimental data adopts the simulated data set (Electrical Grid Stability Simulated Data) of the power system stability research in the UCI machine learning public data set. The experimental data set contains 12 attributes and real classification labels. Each row of the data set represents a transient stability event, and each column corresponds to different characteristic parameters of the event. Based on these characteristic parameters, it is predicted whether the power system is in a stable state. The present invention implements mutual information-based screening and watermark solution based on the MATLAB platform and compares the classification effect of the watermark data generated by the algorithm of the present invention with the original data and EMR.

[0094] The specific process is as follows:

[0095] ① Calculate the mutual information of each attribute, and filter out the attribute set J to be watermarked by designing the importance evaluation index of the attribute data. A , to remove unimportant attributes;

[0096] ② Take the watermark vector wm as the optimization object, and only in the attribute set J in the original data D A Embed the watermark on the attribute column to obtain the watermark data D w .

[0097] ③ Based on the watermark embedding processing results, an optimization model is established to perform the optimization operation of the IGWO algorithm. w The classification effect of EMR watermark data is compared to verify the feasibility and effectiveness of the algorithm in improving the availability of irreversible watermark data.

[0098] The mutual information values ​​of 12 attributes are calculated as follows: Figure 1 As shown, the data set used in the present invention selects 8 attributes for watermark embedding based on mutual information screening.

[0099] like Figure 2As shown in the figure, the IGWO algorithm for obtaining 10 watermark vectors reaches convergence at 100 iterations. At this time, the watermark vector solution achieves a high accuracy rate for the corresponding watermark data while satisfying the constraints.

[0100] like Figure 3 As shown in the figure, 10 groups of watermark data obtained by the WMIIGWO algorithm proposed in the present invention are classified, and the average of the recall rate and F score value are obtained respectively, and compared with the original data and EMR watermark data. From the comparison results, the algorithm proposed in the present invention has similar classification effects as the original data in terms of recall rate and F score value. Compared with the EMR watermark data, the recall rate is still high while maintaining the F score value. The present invention pays attention to the selection and trade-off of attributes in the watermark generation process. By using mutual information as a correlation measure between attributes and classification labels, the present invention can more accurately screen out attributes that have a greater impact on the classification task. This helps to maintain the characteristics of the data and reduce the decline in classification performance after the introduction of the watermark. At the same time, the algorithm searches in the watermark vector space through the gray wolf optimization algorithm and finds a watermark embedding scheme that can obtain a higher recall rate while maintaining a certain F score value. This further demonstrates the superiority of the algorithm of the present invention in balancing data availability and classification performance.

[0101] It should be emphasized that the embodiments of the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific implementation manner. Any other implementation manners derived by those skilled in the art based on the technical solution of the present invention also fall within the scope of protection of the present invention.

Claims

1. A watermarking method based on mutual information screening and IGWO optimization, characterized in that: The following steps are involved: Step 1: Use the mutual information method to screen attributes, obtain the attribute set that is highly correlated with classification accuracy, and determine the unimportant attribute set to form a complete attribute set after mutual information screening; Step 2: Based on the attribute set selected by the mutual information attribute in step 1, a watermark embedding operation is performed on the unimportant attribute set to obtain a watermark data set containing the watermark; Step 3: Based on the watermark data set generated in step 2, the watermark data is classified through the decision tree model, and the IGWO algorithm is applied to optimize the watermark embedding strength and position.

2. A watermarking method based on mutual information screening and IGWO optimization according to claim 1, characterized in that: The specific steps of step 1 include: (1) The concept of mutual information is used to measure the relationship between attributes and classification labels. First, the input continuous value attribute data is read and then divided into several intervals. The equal frequency or equal width method can be selected according to the number of samples. (2) Number each interval and convert the continuous attribute into the corresponding discrete interval value; (3) Calculate the frequency distribution of each interval to ensure that the interval division is reasonable, and normalize the discretized attribute data so that all attribute values ​​are in the same range. The processed attribute data will be stored to generate a new discrete attribute data set, thereby obtaining an attribute set that is highly correlated with classification accuracy; at the same time, filter out the attribute set with lower importance.

3. A watermarking method based on mutual information screening and IGWO optimization according to claim 1, characterized in that: The specific steps of step 2 include: (1) Sort the mutual information attribute set selected based on step 1. The sorted data set will be divided into several intervals. The size and number of intervals are determined according to the sorting result; (2) Determine the mapping interval corresponding to the data points in each attribute set through the sorting results, maintain the distribution consistency of the data before and after the watermark is embedded, and ensure that adding the watermark does not significantly change the statistical characteristics of the data; (3) In the process of embedding watermarks into data points, the relative order of the data points is kept unchanged through interval mapping, so that the data distribution after watermark embedding is consistent with the original data distribution, and finally a watermarked data set containing the watermark is obtained.

4. The watermark method based on mutual information screening and IGWO optimization according to claim 1, characterized in that: The specific steps of step 3 include: (1) Based on the watermark data set generated in step 2, the watermark data is classified using a decision tree model; (2) For the classified watermark data set, the IGWO algorithm is introduced to optimize the embedding strength and position of the watermark: the IGWO algorithm takes the digital watermark vector as the optimization object, calculates the fitness of each watermark vector, and helps determine the optimal watermark embedding scheme.

5. A watermarking method based on mutual information screening and IGWO optimization according to claim 4, characterized in that: The calculation formula of step (2) of step 3 is: The digital watermark vector wm is used as the optimization object of the IGWO algorithm. According to different watermark vectors, the original data is converted into a watermark data set D according to the watermark embedding method proposed in step 2. w , and use D w The classification accuracy of is used as the fitness function; Among them, f(D w ) is the classification result of the watermark data based on the decision tree model, and WM is the watermark set in the existing established watermark index table.

Images