Database storage resource security scheduling method based on artificial intelligence

Abstract

The present invention discloses a database storage resource security scheduling method based on artificial intelligence, which belongs to the technical field of data security scheduling. By using network traffic characteristics and security detection labels corresponding to the network traffic characteristics as training data and minimizing the loss function as the goal, a data resource scheduling process detection model is trained. The trained data resource scheduling process detection model is used to identify real-time network traffic characteristics and determine the security detection results corresponding to the database storage resource scheduling process. Therefore, it can be identified whether the database storage resource scheduling process is safe. When it is unsafe, the database storage resource scheduling process can be controlled, thereby effectively ensuring the security of the database and greatly improving the security of the database storage resource scheduling process.

Description

Technical Field

[0001] The present invention belongs to the technical field of data security scheduling, and in particular relates to a database storage resource security scheduling method based on artificial intelligence. Background Art

[0002] With the advent of the big data era, database systems are increasingly being used in various fields. The secure scheduling of database storage resources has become a pressing issue. Secure scheduling of database storage resources involves ensuring the efficient, reliable, and secure use of database storage resources (such as hard drives, solid-state storage, and network storage) through a series of strategies and technical means. Existing technologies primarily rely on data encryption to ensure secure scheduling of database storage resources. While this effectively prevents data leakage during the scheduling process, it cannot effectively guarantee database security. Data attacks can easily lead to data loss, threatening the security of the entire database. Summary of the Invention

[0003] The present invention provides a database storage resource security scheduling method based on artificial intelligence, which is used to solve the problem of insufficient security protection of databases in the prior art.

[0004] A database storage resource security scheduling method based on artificial intelligence, comprising:

[0005] Use artificial intelligence models to build a data resource scheduling process detection model, and construct a loss function corresponding to the data resource scheduling process detection model;

[0006] Obtain network traffic characteristics and security detection tags corresponding to the network traffic characteristics during the database storage resource scheduling process of the specified data source;

[0007] Using network traffic features and the security detection labels corresponding to the network traffic features as training data and minimizing the loss function as the goal, the data resource scheduling process detection model is trained to obtain the trained data resource scheduling process detection model.

[0008] Deploy the trained data resource scheduling process detection model on the database platform or system, and collect real-time network traffic characteristics during the database storage resource scheduling process;

[0009] The trained data resource scheduling process detection model is used to identify real-time network traffic characteristics and determine the security detection results corresponding to the database storage resource scheduling process;

[0010] Based on the security detection results corresponding to the database storage resource scheduling process, the database storage resource scheduling process is controlled to achieve secure scheduling of database storage resources.

[0011] Furthermore, an artificial intelligence model is used to construct a data resource scheduling process detection model, including: using a convolutional neural network model, a bidirectional long short-term memory network model or a BP neural network model to construct a data resource scheduling process detection model.

[0012] Furthermore, the loss function corresponding to the data resource scheduling process detection model is constructed as follows:

[0013]

[0014] in, represents the loss function, Indicates the i The network traffic characteristics corresponding to the j Actual output; Indicates the i The network traffic characteristics corresponding to the j expected outputs, I represents the total number of network traffic features used for training, and J represents the output dimension of the data resource scheduling process detection model.

[0015] Furthermore, the network traffic features and the security detection labels corresponding to the network traffic features are used as training data, and the loss function is minimized as the goal to train the data resource scheduling process detection model. The trained data resource scheduling process detection model is obtained, including:

[0016] Randomly initialize the hyperparameters of the data resource scheduling process detection model and determine multiple hyperparameter individuals;

[0017] For each hyperparameter individual, network traffic characteristics and the corresponding security detection labels are used as training data, and the loss function is used to obtain the loss function value corresponding to the hyperparameter individual;

[0018] Determine the optimal individual, the first best individual, and the second best individual according to the loss function value, and obtain the mean individual according to all hyperparameter individuals;

[0019] The best individual, the first best individual, the second best individual and the average individual form an elite individual pool;

[0020] For any hyperparameter individual, a corresponding collaborative guidance individual is randomly matched from the elite individual pool, and based on the collaborative guidance individual, the elite collaborative guidance mechanism is used to perform optimal guidance search on the hyperparameter individual to obtain the hyperparameter individual after the optimal guidance search;

[0021] The similarity evolution mechanism is used to conduct local collaborative search on the hyperparameter individuals after the optimal guided search to obtain the hyperparameter individuals after the local collaborative search;

[0022] The three-point interpolation information fusion mechanism is used to perform local expansion search on the hyperparameter individuals after local collaborative search to obtain the hyperparameter individuals after local expansion search;

[0023] Adopting the adaptive mutation mechanism to conduct global mutation search on the hyperparameter individuals after local expansion search, and obtain the hyperparameter individuals after global mutation search;

[0024] The elite collaborative guidance mechanism, similarity evolution mechanism, three-point interpolation information fusion mechanism and adaptive mutation mechanism are repeatedly executed until the maximum number of training times is reached. According to the hyperparameter individuals after the last global mutation search, the final hyperparameters of the data resource scheduling process detection model are obtained, and the trained data resource scheduling process detection model is obtained.

[0025] Furthermore, for any hyperparameter individual, a corresponding collaborative guidance individual is randomly matched from the elite individual pool, and based on the collaborative guidance individual, the elite collaborative guidance mechanism is used to perform optimal guidance search on the hyperparameter individual, and the hyperparameter individual after the optimal guidance search is obtained, including:

[0026] For any hyperparameter individual, a corresponding collaborative guiding individual is randomly matched from the elite individual pool;

[0027] Based on the current number of training times, the adaptive guidance coefficient is obtained as: ;in, Indicates the t The adaptive guidance coefficient in the training process, T represents the maximum number of training times, cos represents the cosine function, represents pi;

[0028] According to the adaptive guidance coefficient, the guidance information of the collaborative guidance individual to the hyperparameter individual is obtained as follows: ;in, Indicates collaborative guidance of individuals to m The guidance information of each hyperparameter, Represents a random number between (0,1), Represents a random number between (0,1), Indicates the t During the training m Hyperparameter individuals, Represents the hyperparameter individual The corresponding collaborative guidance individuals, m=1,2,…,M, M represents the total number of hyperparameter individuals;

[0029] According to the guidance information of the collaborative guidance individual on the hyperparameter individual, the optimal guidance search is performed on the hyperparameter individual, and the hyperparameter individual after the optimal guidance search is obtained as follows: ;in, represents the hyperparameter individual after the optimal guided search .

[0030] Furthermore, a similarity evolution mechanism is used to perform local collaborative search on the hyperparameter individuals after the optimal guided search, and the hyperparameter individuals after the local collaborative search are obtained, including:

[0031] For any hyperparameter individual after the optimal guided search, randomly match another hyperparameter individual to obtain the similarity evolution individual corresponding to the hyperparameter individual;

[0032] For any hyperparameter individual after the optimal guided search, the similarity between the hyperparameter individual and the similarity evolution individual is obtained as follows: ;in, Indicates the n The similarity between a hyperparameter individual and its corresponding similarity evolution individual, Indicates the t During the training n The first hyperparameter individual d dimensional hyperparameters, Indicates the n The first hyperparameter individual and its corresponding similarity evolution individual d Dimensional hyperparameter, D represents the total dimension corresponding to the hyperparameter individual;

[0033] For any hyperparameter individual after the optimal guided search, the similarity between the hyperparameter individual and the optimal individual is obtained as: ; Indicates the n The similarity between the hyperparameter individuals and the optimal individual, The optimal individual d dimensional hyperparameters;

[0034] Determining similarity Is it greater than the similarity If so, the optimal individual is used as the target parameter individual, and the target parameter individual is used to perform a local collaborative search on the hyperparameter individual after the optimal guided search. Otherwise, the similarity evolution individual is used as the target parameter individual, and the target parameter individual is used to perform a local collaborative search on the hyperparameter individual after the optimal guided search.

[0035] According to the target parameter individual, a local collaborative search is performed on the hyperparameter individual after the optimal guided search, and the hyperparameter individual after the local collaborative search is obtained as follows: ;in, Indicates the nth parameter individual in the tth training process, n=1,2,…,M, M represents the total number of hyperparameter individuals, represents the target parameter individual, Represents the hyperparameter individual after local collaborative search , represents the preset learning rate, Represents the individual with target parameters And hyperparameter individual Different first random hyperparameter individuals, Represents the individual with target parameters And hyperparameter individual Different second random hyperparameter individuals.

[0036] Furthermore, a three-point interpolation information fusion mechanism is used to perform a local extended search on the hyperparameter individuals after the local collaborative search, and the hyperparameter individuals after the local extended search are obtained, including:

[0037] For any hyperparameter individual after local collaborative search, randomly match the third random hyperparameter individual and the fourth random hyperparameter individual for interpolation;

[0038] According to the third random hyperparameter individual and the fourth random hyperparameter individual, a local expansion search is performed on the hyperparameter individual, and the hyperparameter individual after the local expansion search is obtained as follows: ;in, Indicates the t During the training k The d-th dimension hyperparameter of the hyperparameter individual after local collaborative search, k =1,2,…,M, where M represents the total number of hyperparameter individuals, represents the d-th dimension hyperparameter of the third random hyperparameter individual, represents the fourth random hyperparameter individual d dimensional hyperparameters, Indicates the k The hyperparameter individual after the local expansion search d dimensional hyperparameters, represents a preset constant term, and Less than 0.001; Indicates the k The fitness of the hyperparameter individuals after local collaborative search, represents the fitness of the third random hyperparameter individual, Indicates the fitness of the fourth random hyperparameter individual, fitness = 1 / (loss function value + ).

[0039] Furthermore, an adaptive mutation mechanism is used to perform a global mutation search on the hyperparameter individuals after the local expansion search, and the hyperparameter individuals after the global mutation search are obtained, including:

[0040] The adaptive mutation probability is obtained as: ;in, p represents the adaptive mutation probability, represents the gamma function, e represents a natural constant, v represents the mutation probability control factor;

[0041] For any hyperparameter individual after local expansion search, based on the adaptive mutation probability, the roulette wheel method is used to determine whether the hyperparameter individual has mutated. If so, the coefficient of variation is obtained. Otherwise, the hyperparameter individual is directly used as the hyperparameter individual after global mutation search.

[0042] Get the coefficient of variation as: ;in, represents the coefficient of variation, represents the variation control factor, e represents a natural constant, cos represents the cosine function, Indicates (-2.5 , 2.5 ), ; Indicates the preset maximum value of the variation control factor, represents a constant factor;

[0043] According to the coefficient of variation, a global variation search is performed on the hyperparameter individual after the local expansion search, and the hyperparameter individual after the global variation search is obtained as follows: ;in, Indicates the t During the training h Hyperparameter individuals after local expansion search, Represents the hyperparameter individual after global mutation search , represents the upper limit individual, represents the lower limit individual, h =1,2,…,M, where M represents the total number of hyperparameter individuals;

[0044] Determine whether the loss function value of the hyperparameter individual after the global mutation search is reduced. If so, accept the global mutation search, otherwise reject the global mutation search.

[0045] Furthermore, based on the security detection result corresponding to the database storage resource scheduling process, the database storage resource scheduling process is controlled, including:

[0046] Determine whether the security detection result corresponding to the database storage resource scheduling process is of an abnormal type. If so, perform speed limiting and blocking operations; otherwise, continue monitoring.

[0047] Furthermore, after controlling the database storage resource scheduling process, it also includes: when the security detection result is monitored to be an abnormal type, generating security warning information, and transmitting the security warning information to the device designated by the staff.

[0048] The present invention provides an artificial intelligence-based database storage resource security scheduling method, which uses network traffic characteristics and security detection labels corresponding to the network traffic characteristics as training data, and takes minimizing the loss function as the goal to train a data resource scheduling process detection model. The trained data resource scheduling process detection model identifies real-time network traffic characteristics and determines the security detection results corresponding to the database storage resource scheduling process, so as to identify whether the database storage resource scheduling process is safe. When it is unsafe, the database storage resource scheduling process can be controlled, thereby effectively ensuring the security of the database and greatly improving the security of the database storage resource scheduling process. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0050] Figure 1 A flowchart of a method for securely scheduling database storage resources based on artificial intelligence is provided in an embodiment of the present invention.

[0051] The above drawings illustrate specific embodiments of the present invention, which will be described in more detail below. These drawings and the accompanying description are not intended to limit the scope of the present invention in any way, but rather to illustrate the concept of the present invention to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0052] Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. In the following description, when referring to the drawings, like numbers in different figures represent like or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present invention, as detailed in the appended claims.

[0053] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0054] like Figure 1As shown, an embodiment of the present invention provides a database storage resource security scheduling method based on artificial intelligence, comprising:

[0055] S1. Use an artificial intelligence model to build a data resource scheduling process detection model, and construct a loss function corresponding to the data resource scheduling process detection model;

[0056] In an embodiment of the present invention, an artificial intelligence model is used to construct a data resource scheduling process detection model, including: using a convolutional neural network model, a bidirectional long short-term memory network model, or a BP neural network model to construct the data resource scheduling process detection model. It is worth noting that in addition to the artificial intelligence model mentioned above, other artificial intelligence models can also be used to construct the data resource scheduling process detection model to achieve the same purpose of the invention.

[0057] The cross entropy loss function, mean square loss function or root mean square loss function can be used as the loss function corresponding to the data resource scheduling process detection model, so as to facilitate the evaluation of the training effect.

[0058] S2. Obtain network traffic characteristics and security detection tags corresponding to the network traffic characteristics during the database storage resource scheduling process of the specified data source;

[0059] Optionally, you can use the data in KDD Cup99 as network traffic features and corresponding security detection labels. KDD Cup99 is the preferred dataset for applying deep learning to abnormal intrusion detection. Studies have shown that by screening relevant features of network traffic data, creating optimal parameters for data mining models, and selecting the optimal data processing method, the speed and accuracy of network intrusion detection can be improved.

[0060] The security detection label can be of normal type or abnormal type, and the abnormal type can be set to multiple types, each type representing an attack method.

[0061] S3. Using network traffic features and security detection labels corresponding to the network traffic features as training data and minimizing the loss function as the goal, train a data resource scheduling process detection model to obtain a trained data resource scheduling process detection model.

[0062] After using the artificial intelligence model to build the data resource scheduling process detection model, the data resource scheduling process detection model cannot be used directly. Therefore, the data resource scheduling process detection model needs to be trained. Intelligent optimization algorithms such as particle swarm optimization can be used to train the data resource scheduling process detection model.

[0063] S4. Deploy the trained data resource scheduling process detection model on the platform or system where the database is located, and collect real-time network traffic characteristics during the database storage resource scheduling process;

[0064] It is worth noting that in order to better identify attacks, the data structure and data type of the real-time network traffic features and the network traffic features used for training should be the same, so that attack detection can be achieved.

[0065] S5. Identify real-time network traffic characteristics through the trained data resource scheduling process detection model to determine the security detection results corresponding to the database storage resource scheduling process;

[0066] The security detection result can be of normal type or abnormal type. When it is of abnormal type, security control can be performed.

[0067] S6. Based on the security detection result corresponding to the database storage resource scheduling process, the database storage resource scheduling process is controlled to achieve secure scheduling of database storage resources.

[0068] In the embodiment of the present invention, the loss function corresponding to the data resource scheduling process detection model is constructed as follows:

[0069]

[0070] in, represents the loss function, Indicates the i The network traffic characteristics corresponding to the j Actual output; Indicates the i The network traffic characteristics corresponding to the j expected outputs, I represents the total number of network traffic features used for training, and J represents the output dimension of the data resource scheduling process detection model.

[0071] It is worth noting that the loss function corresponding to the above-mentioned data resource scheduling process detection model is only an example of an embodiment of the present invention, and other loss functions can also be used to evaluate the training effect.

[0072] In an embodiment of the present invention, network traffic features and security detection labels corresponding to the network traffic features are used as training data, and a data resource scheduling process detection model is trained with the goal of minimizing a loss function. The trained data resource scheduling process detection model is obtained, including:

[0073] Randomly initialize the hyperparameters of the data resource scheduling process detection model and determine multiple hyperparameter individuals;

[0074] For example, when a convolutional neural network is preferably used to construct a data resource scheduling process detection model, the hyperparameters to be trained of the data resource scheduling process detection model are weights and thresholds, and each hyperparameter to be trained has a corresponding upper limit and lower limit. Therefore, random initialization can be performed between the upper and lower limits of the hyperparameter to be trained, and the hyperparameter can be encoded into a vector to obtain a hyperparameter individual. After multiple initializations, multiple different hyperparameter individuals can be obtained.

[0075] Optionally, in order to improve the search speed of the algorithm, the chaotic mapping method can be used to randomly initialize the hyperparameters of the data resource scheduling process detection model, thereby improving the distribution of the initial solution in the solution space and making it easier to find the optimal solution.

[0076] For each hyperparameter individual, network traffic characteristics and the corresponding security detection labels are used as training data, and the loss function is used to obtain the loss function value corresponding to the hyperparameter individual;

[0077] Determine the optimal individual, the first best individual, and the second best individual according to the loss function value, and obtain the mean individual according to all hyperparameter individuals;

[0078] The best individual, the first best individual, the second best individual and the average individual form an elite individual pool;

[0079] For any hyperparameter individual, a corresponding collaborative guidance individual is randomly matched from the elite individual pool, and based on the collaborative guidance individual, the elite collaborative guidance mechanism is used to perform optimal guidance search on the hyperparameter individual to obtain the hyperparameter individual after the optimal guidance search;

[0080] The similarity evolution mechanism is used to conduct local collaborative search on the hyperparameter individuals after the optimal guided search to obtain the hyperparameter individuals after the local collaborative search;

[0081] The three-point interpolation information fusion mechanism is used to perform local expansion search on the hyperparameter individuals after local collaborative search to obtain the hyperparameter individuals after local expansion search;

[0082] Adopting the adaptive mutation mechanism to conduct global mutation search on the hyperparameter individuals after local expansion search, and obtain the hyperparameter individuals after global mutation search;

[0083] The elite collaborative guidance mechanism, similarity evolution mechanism, three-point interpolation information fusion mechanism and adaptive mutation mechanism are repeatedly executed until the maximum number of training times is reached. According to the hyperparameter individual after the last global mutation search, the final hyperparameters of the data resource scheduling process detection model are obtained (that is, the hyperparameters in the optimal individual are used as the final hyperparameters), and the trained data resource scheduling process detection model is obtained.

[0084] Existing technologies (such as gradient descent and particle swarm optimization) are prone to falling into local optima when updating hyperparameters of artificial intelligence models. This results in poor recognition of network traffic characteristics by the trained data resource scheduling detection model, reducing the scheduling security of database storage resources. Therefore, embodiments of the present invention provide a new training method to improve training speed and effectiveness, thereby enhancing the security of database storage resources during scheduling.

[0085] In the embodiment of the present invention, for any hyperparameter individual, a corresponding collaborative guidance individual is randomly matched from the elite individual pool, and based on the collaborative guidance individual, the elite collaborative guidance mechanism is used to perform optimal guidance search on the hyperparameter individual to obtain the hyperparameter individual after the optimal guidance search, including:

[0086] For any hyperparameter individual, a corresponding collaborative guiding individual is randomly matched from the elite individual pool;

[0087] Based on the current number of training times, the adaptive guidance coefficient is obtained as: ;in, Indicates the t The adaptive guidance coefficient in the training process, T represents the maximum number of training times, cos represents the cosine function, represents pi;

[0088] According to the adaptive guidance coefficient, the guidance information of the collaborative guidance individual to the hyperparameter individual is obtained as follows: ;in, Indicates collaborative guidance of individuals to m The guidance information of each hyperparameter, Represents a random number between (0,1), Represents a random number between (0,1), Indicates the t During the training m Hyperparameter individuals, Represents the hyperparameter individual The corresponding collaborative guidance individuals, m=1,2,…,M, M represents the total number of hyperparameter individuals;

[0089] According to the guidance information of the collaborative guidance individual on the hyperparameter individual, the optimal guidance search is performed on the hyperparameter individual, and the hyperparameter individual after the optimal guidance search is obtained as follows: ;in, represents the hyperparameter individual after the optimal guided search .

[0090] The elite collaborative guidance mechanism provided by the embodiment of the present invention can effectively improve the search speed of the algorithm, allowing the hyperparameter individuals to quickly move to the optimal area, thereby realizing the search for local optimal values.

[0091] In an embodiment of the present invention, a similarity evolution mechanism is used to perform local collaborative search on the hyperparameter individuals after the optimal guided search, and the hyperparameter individuals after the local collaborative search are obtained, including:

[0092] For any hyperparameter individual after the optimal guided search, randomly match another hyperparameter individual to obtain the similarity evolution individual corresponding to the hyperparameter individual;

[0093] For any hyperparameter individual after the optimal guided search, the similarity between the hyperparameter individual and the similarity evolution individual is obtained as follows: ;in, Indicates the n The similarity between a hyperparameter individual and its corresponding similarity evolution individual, Indicates the t During the training n The first hyperparameter individual d dimensional hyperparameters, Indicates the n The first hyperparameter individual and its corresponding similarity evolution individual d Dimensional hyperparameter, D represents the total dimension corresponding to the hyperparameter individual;

[0094] For any hyperparameter individual after the optimal guided search, the similarity between the hyperparameter individual and the optimal individual is obtained as: ; Indicates the n The similarity between the hyperparameter individuals and the optimal individual, The optimal individual d dimensional hyperparameters;

[0095] Determining similarity Is it greater than the similarity If so, the optimal individual is used as the target parameter individual, and the target parameter individual is used to perform a local collaborative search on the hyperparameter individual after the optimal guided search. Otherwise, the similarity evolution individual is used as the target parameter individual, and the target parameter individual is used to perform a local collaborative search on the hyperparameter individual after the optimal guided search.

[0096] According to the target parameter individual, a local collaborative search is performed on the hyperparameter individual after the optimal guided search, and the hyperparameter individual after the local collaborative search is obtained as follows: ;in, Indicates the nth parameter individual in the tth training process, n=1,2,…,M, M represents the total number of hyperparameter individuals, represents the target parameter individual, Represents the hyperparameter individual after local collaborative search , represents the preset learning rate, Represents the individual with target parameters And hyperparameter individual Different first random hyperparameter individuals, Represents the individual with target parameters And hyperparameter individual Different second random hyperparameter individuals.

[0097] The similarity evolution mechanism provided by the embodiment of the present invention is more inclined to interact with individuals with smaller similarity. At the same time, combined with random hyperparameter individuals, it can effectively enhance information exchange between individuals. While increasing the possibility of exploring the optimal solution, it can preserve individual diversity during the training process and avoid falling into local optimality.

[0098] In an embodiment of the present invention, a three-point interpolation information fusion mechanism is used to perform a local extended search on the hyperparameter individuals after the local collaborative search, and the hyperparameter individuals after the local extended search are obtained, including:

[0099] For any hyperparameter individual after local collaborative search, randomly match the third random hyperparameter individual and the fourth random hyperparameter individual for interpolation;

[0100] According to the third random hyperparameter individual and the fourth random hyperparameter individual, a local expansion search is performed on the hyperparameter individual, and the hyperparameter individual after the local expansion search is obtained as follows: ;in, Indicates the t During the training k The d-th dimension hyperparameter of the hyperparameter individual after local collaborative search, k =1,2,…,M, where M represents the total number of hyperparameter individuals, represents the d-th dimension hyperparameter of the third random hyperparameter individual, represents the fourth random hyperparameter individual d dimensional hyperparameters, Indicates the k The hyperparameter individual after the local expansion search d dimensional hyperparameters, represents a preset constant term, and Less than 0.001; Indicates the k The fitness of the hyperparameter individuals after local collaborative search, represents the fitness of the third random hyperparameter individual, Indicates the fitness of the fourth random hyperparameter individual, fitness = 1 / (loss function value + ).

[0101] The three-point interpolation information fusion mechanism provided by the embodiment of the present invention can search for unfamiliar areas within a local area and greatly improve the convergence speed of the algorithm.

[0102] In an embodiment of the present invention, an adaptive mutation mechanism is used to perform a global mutation search on the hyperparameter individuals after the local expansion search to obtain the hyperparameter individuals after the global mutation search, including:

[0103] The adaptive mutation probability is obtained as: ;in, p represents the adaptive mutation probability, represents the gamma function, e represents a natural constant, v represents the mutation probability control factor;

[0104] For any hyperparameter individual after local expansion search, based on the adaptive mutation probability, the roulette wheel method is used to determine whether the hyperparameter individual has mutated. If so, the coefficient of variation is obtained. Otherwise, the hyperparameter individual is directly used as the hyperparameter individual after global mutation search.

[0105] Get the coefficient of variation as: ;in, represents the coefficient of variation, represents the variation control factor, e represents a natural constant, cos represents the cosine function, Indicates (-2.5 , 2.5 ), ; Indicates the preset maximum value of the variation control factor, represents a constant factor;

[0106] According to the coefficient of variation, a global variation search is performed on the hyperparameter individual after the local expansion search, and the hyperparameter individual after the global variation search is obtained as follows: ;in, Indicates the t During the training h Hyperparameter individuals after local expansion search, Represents the hyperparameter individual after global mutation search , represents the upper limit individual (i.e., the individual composed of the upper limit of each dimension hyperparameter), represents the lower limit individual (i.e., the individual composed of the lower limit of each dimension hyperparameter), h =1,2,…,M, where M represents the total number of hyperparameter individuals;

[0107] Determine whether the loss function value of the hyperparameter individual after the global mutation search is reduced. If so, accept the global mutation search, otherwise reject the global mutation search.

[0108] The adaptive mutation mechanism provided by the embodiment of the present invention can provide a greater mutation probability in the early stage of the algorithm and a smaller mutation probability in the later stage of the algorithm, which can effectively assist the algorithm to escape from the local optimal solution, and combined with the adaptive mutation probability and the greedy algorithm, it can effectively ensure the convergence speed of the algorithm.

[0109] In an embodiment of the present invention, the database storage resource scheduling process is controlled based on the security detection result corresponding to the database storage resource scheduling process, including:

[0110] Determine whether the security detection result corresponding to the database storage resource scheduling process is of an abnormal type. If so, perform speed limiting and blocking operations; otherwise, continue monitoring.

[0111] Speed ​​limiting and blocking operations can be performed to: limit the data parsing speed of illegal users, block the IP of illegal users, or block the port of the database. Various operations can be set according to actual needs.

[0112] In an embodiment of the present invention, after controlling the database storage resource scheduling process, it also includes: when the security detection result is monitored to be an abnormal type, a security warning information (such as a text warning information) is generated, and the security warning information is transmitted to the device designated by the staff, so that the security warning can be effectively realized.

[0113] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and variations can be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A database storage resource security scheduling method based on artificial intelligence, characterized in that: include: Use artificial intelligence models to build a data resource scheduling process detection model, and construct a loss function corresponding to the data resource scheduling process detection model; Obtain network traffic characteristics and security detection tags corresponding to the network traffic characteristics during the database storage resource scheduling process of the specified data source; Using network traffic features and the security detection labels corresponding to the network traffic features as training data and minimizing the loss function as the goal, the data resource scheduling process detection model is trained to obtain the trained data resource scheduling process detection model. Deploy the trained data resource scheduling process detection model on the database platform or system, and collect real-time network traffic characteristics during the database storage resource scheduling process; The trained data resource scheduling process detection model is used to identify real-time network traffic characteristics and determine the security detection results corresponding to the database storage resource scheduling process; Based on the security detection results corresponding to the database storage resource scheduling process, the database storage resource scheduling process is controlled to achieve secure scheduling of database storage resources; Using network traffic features and the corresponding security detection labels as training data, and minimizing the loss function as the goal, the data resource scheduling process detection model is trained. The trained data resource scheduling process detection model is obtained, which includes: Randomly initialize the hyperparameters of the data resource scheduling process detection model and determine multiple hyperparameter individuals; For each hyperparameter individual, network traffic characteristics and the corresponding security detection labels are used as training data, and the loss function is used to obtain the loss function value corresponding to the hyperparameter individual; Determine the optimal individual, the first best individual, and the second best individual according to the loss function value, and obtain the mean individual according to all hyperparameter individuals; The best individual, the first best individual, the second best individual and the average individual form an elite individual pool; For any hyperparameter individual, a corresponding collaborative guidance individual is randomly matched from the elite individual pool, and based on the collaborative guidance individual, the elite collaborative guidance mechanism is used to perform optimal guidance search on the hyperparameter individual to obtain the hyperparameter individual after the optimal guidance search; The method adopts an elite collaborative guidance mechanism to conduct an optimal guidance search for the hyperparameter individual, including: obtaining an adaptive guidance coefficient based on the current number of training times; obtaining guidance information of the collaborative guidance individual on the hyperparameter individual according to the adaptive guidance coefficient; conducting an optimal guidance search on the hyperparameter individual according to the guidance information of the collaborative guidance individual on the hyperparameter individual, and obtaining the hyperparameter individual after the optimal guidance search; The similarity evolution mechanism is used to conduct local collaborative search on the hyperparameter individuals after the optimal guided search to obtain the hyperparameter individuals after the local collaborative search; The three-point interpolation information fusion mechanism is used to perform local expansion search on the hyperparameter individuals after local collaborative search to obtain the hyperparameter individuals after local expansion search; Adopting the adaptive mutation mechanism to conduct global mutation search on the hyperparameter individuals after local expansion search, and obtain the hyperparameter individuals after global mutation search; The elite collaborative guidance mechanism, similarity evolution mechanism, three-point interpolation information fusion mechanism and adaptive mutation mechanism are repeatedly executed until the maximum number of training times is reached. According to the hyperparameter individuals after the last global mutation search, the final hyperparameters of the data resource scheduling process detection model are obtained, and the trained data resource scheduling process detection model is obtained.

2. The method for securely scheduling database storage resources based on artificial intelligence according to claim 1, characterized in that: An artificial intelligence model is used to construct a data resource scheduling process detection model, including: using a convolutional neural network model, a bidirectional long short-term memory network model or a BP neural network model to construct a data resource scheduling process detection model.

3. The method for securely scheduling database storage resources based on artificial intelligence according to claim 1, characterized in that: The loss function corresponding to the data resource scheduling process detection model is: in, represents the loss function, Indicates the i The network traffic characteristics corresponding to the j Actual output; Indicates the i The network traffic characteristics corresponding to the j expected outputs, I represents the total number of network traffic features used for training, and J represents the output dimension of the data resource scheduling process detection model.

4. The method for securely scheduling database storage resources based on artificial intelligence according to claim 1, characterized in that: For any hyperparameter individual, a corresponding collaborative guidance individual is randomly matched from the elite individual pool, and based on the collaborative guidance individual, the elite collaborative guidance mechanism is used to perform optimal guidance search on the hyperparameter individual, and the hyperparameter individual after the optimal guidance search is obtained, including: For any hyperparameter individual, a corresponding collaborative guiding individual is randomly matched from the elite individual pool; Based on the current number of training times, the adaptive guidance coefficient is obtained as: ;in, Indicates the t The adaptive guidance coefficient in the training process, T represents the maximum number of training times, cos represents the cosine function, represents pi; According to the adaptive guidance coefficient, the guidance information of the collaborative guidance individual to the hyperparameter individual is obtained as follows: ;in, Indicates collaborative guidance of individuals to m The guidance information of each hyperparameter, Represents a random number between (0,1), Represents a random number between (0,1), Indicates the t During the training m Hyperparameter individuals, Represents the hyperparameter individual The corresponding collaborative guidance individuals, m=1,2,…,M, M represents the total number of hyperparameter individuals; According to the guidance information of the collaborative guidance individual on the hyperparameter individual, the optimal guidance search is performed on the hyperparameter individual, and the hyperparameter individual after the optimal guidance search is obtained as follows: ;in, represents the hyperparameter individual after the optimal guided search .

5. The method for securely scheduling database storage resources based on artificial intelligence according to claim 4, characterized in that: The similarity evolution mechanism is used to conduct local collaborative search on the hyperparameter individuals after the optimal guided search, and the hyperparameter individuals after the local collaborative search are obtained, including: For any hyperparameter individual after the optimal guided search, randomly match another hyperparameter individual to obtain the similarity evolution individual corresponding to the hyperparameter individual; For any hyperparameter individual after the optimal guided search, the similarity between the hyperparameter individual and the similarity evolution individual is obtained as follows: ;in, Indicates the n The similarity between a hyperparameter individual and its corresponding similarity evolution individual, Indicates the t During the training n The first hyperparameter individual d dimensional hyperparameters, Indicates the n The first hyperparameter individual and its corresponding similarity evolution individual d Dimensional hyperparameter, D represents the total dimension corresponding to the hyperparameter individual; For any hyperparameter individual after the optimal guided search, the similarity between the hyperparameter individual and the optimal individual is obtained as: ; Indicates the n The similarity between the hyperparameter individuals and the optimal individual, The optimal individual d dimensional hyperparameters; Determining similarity Is it greater than the similarity If so, the optimal individual is used as the target parameter individual, and the target parameter individual is used to perform a local collaborative search on the hyperparameter individual after the optimal guided search. Otherwise, the similarity evolution individual is used as the target parameter individual, and the target parameter individual is used to perform a local collaborative search on the hyperparameter individual after the optimal guided search. According to the target parameter individual, a local collaborative search is performed on the hyperparameter individual after the optimal guided search, and the hyperparameter individual after the local collaborative search is obtained as follows: ;in, Indicates the nth parameter individual in the tth training process, n=1,2,…,M, M represents the total number of hyperparameter individuals, represents the target parameter individual, Represents the hyperparameter individual after local collaborative search , Represents the preset learning rate, Represents the individual with target parameters And the hyperparameter individual Different first random hyperparameter individuals, Represents the individual with target parameters And the hyperparameter individual Different second random hyperparameter individuals.

6. The method for securely scheduling database storage resources based on artificial intelligence according to claim 5, characterized in that: The three-point interpolation information fusion mechanism is used to perform local expansion search on the hyperparameter individuals after local collaborative search, and the hyperparameter individuals after local expansion search are obtained, including: For any hyperparameter individual after local collaborative search, randomly match the third random hyperparameter individual and the fourth random hyperparameter individual for interpolation; According to the third random hyperparameter individual and the fourth random hyperparameter individual, a local expansion search is performed on the hyperparameter individual, and the hyperparameter individual after the local expansion search is obtained as follows: ;in, Indicates the t During the training k The d-th dimension hyperparameter of the hyperparameter individual after local collaborative search, k =1,2,…,M, where M represents the total number of hyperparameter individuals, represents the d-th dimension hyperparameter of the third random hyperparameter individual, represents the fourth random hyperparameter individual d dimensional hyperparameters, Indicates the k The hyperparameter individual after the local expansion search d dimensional hyperparameters, represents a preset constant term, and Less than 0.001; Indicates the k The fitness of the hyperparameter individuals after local collaborative search, represents the fitness of the third random hyperparameter individual, Indicates the fitness of the fourth random hyperparameter individual, fitness = 1 / (loss function value + ).

7. The method for securely scheduling database storage resources based on artificial intelligence according to claim 6, characterized in that: Adopting the adaptive mutation mechanism to perform global mutation search on the hyperparameter individuals after the local expansion search, the hyperparameter individuals after the global mutation search are obtained, including: The adaptive mutation probability is obtained as: ;in, p represents the adaptive mutation probability, represents the gamma function, e represents a natural constant, v represents the mutation probability control factor; For any hyperparameter individual after local expansion search, based on the adaptive mutation probability, the roulette wheel method is used to determine whether the hyperparameter individual has mutated. If so, the coefficient of variation is obtained. Otherwise, the hyperparameter individual is directly used as the hyperparameter individual after global mutation search. Get the coefficient of variation as: ;in, represents the coefficient of variation, represents the variation control factor, e represents a natural constant, cos represents the cosine function, Indicates (-2.5 , 2.5 ), ; Indicates the preset maximum value of the variation control factor, represents a constant factor; According to the coefficient of variation, a global variation search is performed on the hyperparameter individual after the local expansion search, and the hyperparameter individual after the global variation search is obtained as follows: ;in, Indicates the t During the training h Hyperparameter individuals after local expansion search, Represents the hyperparameter individual after global mutation search , represents the upper limit individual, represents the lower limit individual, h =1,2,…,M, where M represents the total number of hyperparameter individuals; Determine whether the loss function value of the hyperparameter individual after the global mutation search is reduced. If so, accept the global mutation search, otherwise reject the global mutation search.

8. The method for securely scheduling database storage resources based on artificial intelligence according to claim 1, characterized in that: Based on the security detection results corresponding to the database storage resource scheduling process, the database storage resource scheduling process is controlled, including: Determine whether the security detection result corresponding to the database storage resource scheduling process is of an abnormal type. If so, perform speed limiting and blocking operations; otherwise, continue monitoring.

9. The method for securely scheduling database storage resources based on artificial intelligence according to claim 8, characterized in that: After controlling the database storage resource scheduling process, the method further includes: when it is monitored that the security detection result is an abnormal type, generating security warning information, and transmitting the security warning information to the device designated by the staff.

Images