Real-time data classification and hierarchical encryption system and method based on deep learning

Abstract

The application provides a real-time data classification and hierarchical encryption system and method based on deep learning, which comprises the following steps: dividing real-time data into a plurality of independent data frames, identifying the data attributes corresponding to each independent data frame, constructing a positive data representation of the corresponding real-time data according to the data attributes, dividing the real-time data into corresponding data sets according to the positive data representation, constructing a random ciphertext sequence based on the data attributes to encrypt the corresponding independent data frame, constructing primary encrypted real-time data according to the encryption effect of each independent data frame, selecting a plurality of random encrypted modes contained in the data set to locally encrypt the primary encrypted real-time data for several times, determining the secret level of the real-time data and saving it, which can classify different data first and then encrypt them accordingly, thereby guaranteeing the confidentiality of the data to be encrypted and reducing the processing of non-encrypted data, and fully guaranteeing the security of the data to be encrypted.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a real-time data classification and hierarchical encryption system and method based on deep learning. BACKGROUND

[0002] With the continuous development of science and technology, human beings have entered the digital era, and information technology and digital technology have developed rapidly and profoundly changed every aspect of society. Under the drive of the data era, people can obtain the information or data they need in a variety of convenient and efficient ways. While enjoying the convenience and intelligence brought by the data era, people also suffer from the confusion and risks it brings. For example: data leakage, data tampering, data leakage and other problems. For individuals: privacy in life cannot be protected, for enterprises and factories: interests cannot be maintained. However, with the rapid development of science and technology, society and individuals are increasingly dependent on digital tools, leading some people or enterprises to over-rely on technology and be unable to resist risks, but also unable to enjoy information technology and digital technology with peace of mind. At present, how to protect data that should not be disclosed and improve the trust and acceptance of the public on information technology and digital technology has become a problem to be solved.

[0003] Therefore, the present application provides a real-time data classification and hierarchical encryption system and method based on deep learning. SUMMARY

[0004] The real-time data classification and hierarchical encryption system and method based on deep learning can classify different data first, and then encrypt accordingly, which not only guarantees the confidentiality of encrypted data, but also reduces the processing of non-encrypted data, and fully guarantees the security of encrypted data.

[0005] The present application provides a real-time data classification and hierarchical encryption system based on deep learning, comprising:

[0006] A frame processing module is used to divide the real-time data into a plurality of independent data frames, and identify the data attributes corresponding to each independent data frame;

[0007] A data classification module is used to construct a positive data representation corresponding to the real-time data according to the data attributes, and divide the real-time data into corresponding data sets according to the positive data representation;

[0008] A preliminary encryption module is used to construct a random ciphertext sequence based on the data attributes, and encrypt the corresponding independent data frame, and construct a preliminary encrypted real-time data according to the encryption effect of each independent data frame;

[0009] The deep encryption module is used to select several random encrypted methods contained in the dataset to perform several local encryptions on the initially encrypted real-time data, determine the security level of the real-time data, and save it.

[0010] In one feasible approach

[0011] The frame segmentation processing module includes:

[0012] The compression processing unit is used to compress and encode real-time data, filter out several whitespace characters contained in the encoding result, and divide the real-time data accordingly based on the distribution characteristics of the whitespace characters in the encoding result to obtain several independent data frames.

[0013] The preliminary screening unit is used to randomly select a corresponding number of datasets based on the number of frames of the independent data frames, perform similarity analysis on each independent data frame and different datasets to construct a similarity matrix of the real-time data, and construct an initial matching dataset for each independent data frame based on the elements corresponding to the same matrix position in different similar matrices.

[0014] The attribute analysis unit is used to construct several data matching methods for the real-time data based on the frame sequence number of each independent data frame in the real-time data and the corresponding several initial matching datasets, obtain the initial matching data attributes corresponding to each dataset, and verify the attribute exclusion features corresponding to each data matching method respectively.

[0015] The attribute determination unit is used to select a target data matching method with an attribute exclusion feature value of 0, and to determine the data attribute corresponding to each independent data frame.

[0016] In one feasible approach

[0017] Also includes:

[0018] An iterative filtering unit is used to randomly select a corresponding number of updated datasets based on the number of frames of the independent data frames when all the attribute exclusion features are not 0, and transmit the updated datasets to the preliminary filtering unit for corresponding analysis.

[0019] In one feasible approach

[0020] The data classification module includes:

[0021] An attribute processing unit is used to perform enhancement processing on the data attributes respectively, determine several enhancement items of the real-time data based on the difference between the results before and after enhancement, find the relevant sub-data corresponding to each enhancement item in the real-time data, assign a value to the enhancement item based on the data value corresponding to the relevant sub-data, and generate a positive data representation of the real-time data.

[0022] The condition generating unit is configured to acquire the intra-set feature and the intra-set data disclosure information of each data set respectively, construct the representation collection condition of the data set according to the intra-set feature, and construct the interval collection condition of the data set according to the intra-set data disclosure information;

[0023] The classification executing unit is configured to match the representation collection condition with the positive data representation, match the interval collection condition with the data specification of the real-time data, obtain the data set matched with the real-time data, and classify the real-time data according to the set category of the data set.

[0024] In an implementable manner,

[0025] The condition generating unit comprises:

[0026] The first generating subunit is configured to acquire the data basic information corresponding to different intra-set data in each data set respectively, construct the strong sub-feature of the data set according to the same information between the data basic information, acquire the data type corresponding to each intra-set data, construct the weak sub-feature of the data set in combination with the different information between the data basic information, and generate the intra-set feature of the data set according to the strong sub-feature and the weak sub-feature.

[0027] The second generating subunit is configured to acquire the data disclosure information corresponding to different intra-set data in each data set respectively, determine the data maximum value and the data minimum value of the intra-set data according to the data disclosure information, acquire a plurality of first test numbers greater than the data maximum value and a plurality of second test numbers less than the data minimum value, and analyze the includable data range of the data set by using the first test numbers and the second test numbers.

[0028] The third generating subunit is configured to construct the representation collection condition of the data set according to a plurality of feature key points contained in the intra-set feature, and construct the interval collection condition of the data set according to the includable data range.

[0029] In an implementable manner,

[0030] The preliminary encryption module comprises:

[0031] The ciphertext construction unit is configured to determine the invalid encryption mode corresponding to the independent data frame according to the data attribute, delete the corresponding invalid ciphertext, determine a plurality of groups of valid ciphertexts of the real-time data, perform chaos processing on the valid ciphertexts, and generate a chaos ciphertext sequence of the real-time data.

[0032] The encryption training unit is configured to select a sequence segment from the chaotic ciphertext sequence, obtain a random ciphertext corresponding to each independent data frame, encrypt the corresponding independent data frame by using the random ciphertext, cover the corresponding independent data frame by using the obtained encrypted data frame, and score the encryption effect of the corresponding random ciphertext according to the coverage rate of each encrypted data frame.

[0033] The encryption execution unit is configured to select a target random ciphertext with the highest score for each independent data frame, sort the target random ciphertext based on the sorting order of the independent data frame, generate an execution ciphertext of the real-time data, select an effective encryption mode of the independent data according to the positive data representation, encrypt the real-time data by using the effective encryption mode and the execution ciphertext, and generate initial encrypted real-time data.

[0034] In an implementable manner,

[0035] Further comprising:

[0036] According to the encryption effect score, a plurality of encryption defect positions corresponding to the encrypted data frame are determined.

[0037] The encryption defect positions are located in the encrypted data frame, and the encryption defect positions are hidden.

[0038] In an implementable manner,

[0039] The deep encryption module comprises:

[0040] The mode screening unit is configured to determine an invalid encryption mode of the real-time data according to the positive data representation, select a plurality of random encryption modes except the invalid encryption mode in the data set, and determine a historical use frequency corresponding to each random encryption mode.

[0041] The encryption preparation unit is configured to sort the random encryption modes in the order from less to more according to the historical use frequency, set a corresponding order level for the random encryption modes, and determine a plurality of encryptable data frames corresponding to each random encryption mode according to an encryption attribute corresponding to each random encryption mode and a data attribute of the independent data frame.

[0042] The local encryption unit is configured to select a corresponding implementation random encryption mode for the independent data frame according to the order level of the random encryption mode corresponding to each independent data frame, and locally encrypt the corresponding independent data frame by using the implementation random encryption mode.

[0043] A secret level determination unit is configured to determine a secret level of the real-time data according to a data source of the real-time data, add a corresponding secret level label to the encrypted real-time data, and save the encrypted real-time data in a secret level storage area in a corresponding data set.

[0044] The application provides a deep learning-based real-time data classification and hierarchical encryption method, which comprises the following steps:

[0045] Step 1: dividing the real-time data into a plurality of independent data frames, and identifying data attributes corresponding to each independent data frame;

[0046] Step 2: constructing a positive data representation corresponding to the real-time data according to the data attributes, and dividing the real-time data into corresponding data sets according to the positive data representation;

[0047] Step 3: constructing a random ciphertext sequence based on the data attributes to perform encryption training on the corresponding independent data frames, and constructing initial encrypted real-time data according to the encryption effect of each independent data frame;

[0048] Step 4: selecting a plurality of random encrypted modes contained in the data sets to perform a plurality of times of local encryption on the initial encrypted real-time data, determining a secret level of the real-time data, and saving the real-time data.

[0049] In an implementable manner,

[0050] The step 2 comprises:

[0051] Step 21: respectively performing enhancement processing on the data attributes, determining a plurality of enhancement items of the real-time data according to a result difference before and after the enhancement processing, finding relevant sub-data corresponding to each enhancement item in the real-time data, assigning a value to the enhancement item according to a data value of the relevant sub-data, and generating a positive data representation of the real-time data;

[0052] Step 22: respectively acquiring an intra-set feature and intra-set data public information of each data set, constructing a representation collection condition corresponding to the data set according to the intra-set feature, and constructing an interval collection condition corresponding to the data set according to the intra-set data public information;

[0053] Step 23: matching the representation collection condition with the positive data representation, matching the interval collection condition with a data specification of the real-time data, obtaining a data set matched with the real-time data, and classifying the real-time data according to a set category of the data set.

[0054] The implementable beneficial effects of the technical scheme are as follows: in order to better classify and encrypt data, the collected real-time data is divided into a plurality of independent data frames, the data attributes of each independent data frame are analyzed to construct a positive data representation of the real-time data, so that the data set to which the real-time data belongs can be determined, the data classification work is completed by dividing the real-time data into the corresponding data set, and then the independent data frame is encrypted and trained by using the data attributes to construct a random ciphertext sequence, thereby generating corresponding initial encrypted real-time data; in order to guarantee the encryption quality and further guarantee the quality of the real-time data encryption, the initial encrypted real-time data is locally encrypted by using a random encryption mode used by existing in-set data in the data set, so that the encryption of weak parts can be compensated, and the encryption mode in the data set of the same classification can be used for deep encryption, thereby improving the quality and efficiency of the encryption result, improving the overall encryption effect of the real-time data, and achieving the purpose of synchronously generating data and completely encrypting in a short time.

[0055] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be achieved and obtained by means of the structure particularly pointed out in the written description and the accompanying drawings.

[0056] The technical scheme of the present application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0057] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation on the present application. In the drawings:

[0058] Figure 1 The composition schematic diagram of the real-time data classification and hierarchical encryption system based on deep learning in the embodiment of the present application is shown in the figure.

[0059] Figure 2 The working flowchart of the real-time data classification and hierarchical encryption method based on deep learning in the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0060] The preferred embodiments of the present application are described below in combination with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not serve as a limitation on the present application.

[0061] Embodiment 1

[0062] The embodiment provides a real-time data classification and hierarchical encryption system based on deep learning, as shown in the figure, which comprises: Figure 1

[0063] ​a frame processing module, configured to divide the real-time data into a plurality of independent data frames, and identify a data attribute corresponding to each of the independent data frames;

[0064] a data classification module, configured to construct a positive data representation corresponding to the real-time data according to the data attribute, and divide the real-time data into a corresponding data set according to the positive data representation;

[0065] a preliminary encryption module, configured to construct a random ciphertext sequence based on the data attribute, and perform encryption training on the corresponding independent data frame, and construct preliminary encrypted real-time data according to an encryption effect corresponding to each of the independent data frames;

[0066] a deep encryption module, configured to select a plurality of random encrypted modes included in the data set to perform a plurality of times of local encryption on the preliminary encrypted real-time data, determine a secret level of the real-time data, and save the secret level.

[0067] In this example, the independent data frame represents a result of dividing the real-time data into sub-data;

[0068] In this example, the data attribute represents an attribute presented by the independent data frame;

[0069] In this example, the positive data representation represents a combined expression of a feature and an attribute of the real-time data;

[0070] In this example, the data set represents a data set used to store the real-time data;

[0071] In this example, the preliminary encrypted real-time data represents a result of encrypting the independent data frame in the real-time data by using the random ciphertext sequence;

[0072] In this example, the local encryption represents a process of encrypting a data part in the preliminary encrypted real-time data by using the random encrypted mode;

[0073] In this example, the real-time data with a corresponding secret level in the same data set is stored in the same area.

[0074] The working principle and beneficial effects of the technical solution are as follows: in order to better classify and encrypt data, the collected real-time data is divided into a plurality of independent data frames, the data attributes of each independent data frame are analyzed to construct a positive data representation of the real-time data, so that the data set to which the real-time data belongs can be determined, the real-time data is classified into the corresponding data set, and then the data attributes are used to construct a random ciphertext sequence to encrypt and train the independent data frame, thereby generating corresponding initial encrypted real-time data. In order to ensure the encryption quality and further ensure the quality of the real-time data encryption, the initial encrypted real-time data is locally encrypted by using the random encryption mode used by the existing data in the data set. Firstly, the encryption of the weak part can be compensated, and secondly, the encryption mode in the data set of the same classification can be used for deep encryption, thereby improving the quality and efficiency of the encryption result, improving the overall encryption effect of the real-time data, and achieving the purpose of synchronously generating data and completely encrypting in a short time.

[0075] Embodiment 2

[0076] Based on the real-time data classification and hierarchical encryption system based on deep learning in embodiment 1, the frame processing module comprises:

[0077] A compression processing unit is configured to compress and encode the real-time data, filter a plurality of blank characters contained in the encoding result, divide the real-time data according to the distribution characteristics of the blank characters in the encoding result, and obtain a plurality of independent data frames.

[0078] A preliminary filtering unit is configured to randomly select a corresponding number of data sets according to the number of independent data frames, respectively analyze the similarity of each independent data frame and different data sets to construct a similarity matrix of the real-time data, and construct an initial matching data set of each independent data frame according to the elements corresponding to the same matrix position in different similarity matrices.

[0079] An attribute analysis unit is configured to construct a plurality of data matching modes of the real-time data based on the frame sequence number of each independent data frame in the real-time data and a plurality of initial matching data sets corresponding thereto, obtain the initial matching data attributes corresponding to each data set, and respectively verify the attribute exclusion characteristics corresponding to each data matching mode.

[0080] An attribute determination unit is configured to select a target data matching mode with a feature value of 0 of the attribute exclusion characteristics, and determine the data attribute corresponding to each independent data frame.

[0081] In this example, compression encoding represents the process of compressing and re-encoding real-time data.

[0082] In this example, the number of data sets is consistent with the number of frames.

[0083] In this example, the similarity matrix represents a matrix generated by counting the similarity between an independent data frame and different data sets;

[0084] In this example, the initial matching data set represents that the independent data frame is matched with the data set;

[0085] In this example, the data matching mode represents a mode of matching the independent data frame in real-time data after combining different initial matching data sets;

[0086] In this example, the attribute exclusion feature represents a feature exhibited when exclusion occurs between multiple initial matching data attributes included in a data matching mode.

[0087] The working principle and beneficial effects of the above technical solution are as follows: In order to ensure the accuracy of data classification and encryption, the data attributes of each independent data frame in real-time data are first determined. The division position of the independent data frame is determined by compression and encoding, thereby dividing the real-time data into several independent data frames. Further, data sets consistent with the number of independent data frames are selected. The similarity matrix is constructed by analyzing the similarity between the independent data frame and different data sets. The independent data frame is matched with the corresponding initial matching data set by using the principle that the matrix element is similar and the data is effective. Further, the initial matching data set is sorted to construct several data matching modes of real-time data. The data attributes corresponding to each independent data frame are determined by analyzing the attribute exclusion feature in different data matching modes. The first step of real-time data analysis is completed by determining the data attributes of the independent data frame, which determines the general direction of subsequent data classification and ensures the effectiveness of real-time data classification, effectively reducing errors.

[0088] Embodiment 3

[0089] Based on embodiment 2, the real-time data classification and hierarchical encryption system based on deep learning further comprises:

[0090] The iterative screening unit is configured to, when all the attribute exclusion features are not 0, randomly select a corresponding number of updated data sets according to the number of frames of the independent data frame, and transmit the updated data sets to the preliminary screening unit for corresponding analysis.

[0091] The working principle and beneficial effects of the above technical solution are as follows: When the first selected data set cannot meet the needs, the data set is reselected for iterative analysis, which ensures that each independent data frame can be matched with the corresponding data attribute.

[0092] Embodiment 4

[0093] On the basis of embodiment 1, the real-time data classification and hierarchical encryption system based on deep learning, the data classification module comprises:

[0094] The attribute processing unit is configured to perform enhancement processing on the data attributes respectively, determine a plurality of enhancement items of the real-time data according to the difference between the results before and after enhancement, find the relevant sub-data corresponding to each enhancement item in the real-time data, assign a value to the enhancement item according to the data value corresponding to the relevant sub-data, and generate a positive data representation of the real-time data.

[0095] The condition generation unit is configured to acquire the intra-set features and intra-set data public information of each data set respectively, construct a representation collection condition corresponding to the data set according to the intra-set features, and construct an interval collection condition corresponding to the data set according to the intra-set data public information.

[0096] The classification execution unit is configured to match the representation collection condition with the positive data representation, match the interval collection condition with the data specification of the real-time data, obtain a data set matched with the real-time data, and classify the real-time data according to the set category of the data set.

[0097] In this example, the enhancement processing represents the process of enhancing the data attributes and strengthening the prominent key attributes therein;

[0098] In this example, the enhancement item represents a data sub that has been enhanced;

[0099] In this example, the intra-set features represent the features presented by the data set, and the intra-set data public information represents the information that all intra-set data in the data set allow to be publicly viewed;

[0100] In this example, the representation collection condition represents the basic condition that needs to be met when collecting data into the data set;

[0101] In this example, the interval collection condition represents the data specification condition that needs to be met when collecting data into the data set;

[0102] In this example, the set category represents the public category to which the data set belongs, for example: the set category of data set A is: financial category.

[0103] The working principle and beneficial effects of the technical solution are as follows: in order to realize effective classification, the enhanced item of real-time data is determined by enhancing the data attributes, the enhanced item is valued by using the related sub-data of the enhanced item, the positive data representation of real-time data is determined, then the information of each data set is processed, the representation collection condition and the interval collection condition of each data set are constructed, and it is judged that the real-time data conforms to which data set, the data classification of the real-time data is divided according to the set category of the data set, and the purpose of classifying the real-time data is achieved. Through this way, the data classification can be completed, and the collection conditions of each data set can be updated in real time, so that multiple real-time data can be classified in a short time, and the intelligence of the system is improved.

[0104] Embodiment 5

[0105] Based on the real-time data classification and hierarchical encryption system based on deep learning in embodiment 4, the condition generation unit comprises:

[0106] The first generation sub-unit is configured to acquire the data basic information corresponding to different in-set data in each data set respectively, construct the strong sub-feature corresponding to the data set according to the same information between the data basic information, acquire the data type corresponding to each in-set data, construct the weak sub-feature corresponding to the data set in combination with the different information between the data basic information, and generate the in-set feature corresponding to the data set according to the strong sub-feature and the weak sub-feature.

[0107] The second generation sub-unit is configured to acquire the data public information corresponding to different in-set data in each data set respectively, determine the data maximum value and the data minimum value corresponding to the in-set data according to the data public information, acquire a plurality of first test numbers greater than the data maximum value and a plurality of second test numbers less than the data minimum value, and analyze the includable data range of the data set by using the first test numbers and the second test numbers.

[0108] The third generation sub-unit is configured to construct the representation collection condition of the data set according to a plurality of feature key points contained in the in-set feature, and construct the interval collection condition of the data set according to the includable data range.

[0109] In this example, the strong sub-feature represents the feature that must be realized if a data conforms to the data set, and the weak sub-feature represents the feature that is not necessarily realized if a data conforms to the data set.

[0110] In this example, the first test number represents a test value used to test and explore the maximum data allowed to be included in the data set, and the second test number represents a test value used to test and explore the minimum data allowed to be included in the data set.

[0111] In this example, the includable data range indicates a data value range in which the data set allows data to be included.

[0112] The working principle and beneficial effects of the above technical solution are as follows: In order to further guarantee the effectiveness of the collection condition of the data set and avoid real-time data classification failure due to errors, the in-set characteristics of the data set and the includable data range are determined by analyzing the basic information and data disclosure information of the data set, thereby constructing the corresponding representation collection condition and interval collection condition, and laying a foundation for subsequent data classification.

[0113] Embodiment 6

[0114] Based on the embodiment 1, the real-time data classification and hierarchical encryption system based on deep learning, the preliminary encryption module comprises:

[0115] A ciphertext construction unit is configured to determine an invalid encryption mode corresponding to the independent data frame according to the data attribute, delete the corresponding invalid ciphertext, determine a plurality of groups of valid ciphertexts of the real-time data, perform chaotic processing on the valid ciphertexts, and generate a chaotic ciphertext sequence of the real-time data.

[0116] An encryption training unit is configured to select a sequence segment in the chaotic ciphertext sequence, obtain a random ciphertext corresponding to each independent data frame, encrypt the corresponding independent data frame by using the random ciphertext, cover the corresponding independent data frame by using the obtained encrypted data frame, and score the encryption effect of the corresponding random ciphertext according to the coverage rate of each encrypted data frame.

[0117] An encryption execution unit is configured to select a target random ciphertext with the highest score for each independent data frame, sort the target random ciphertexts based on the sorting order of the independent data frames, generate an execution ciphertext of the real-time data, select an effective encryption mode of the independent data according to the positive data representation, encrypt the real-time data by using the effective encryption mode and the execution ciphertext, and generate preliminary encrypted real-time data.

[0118] In this example, the invalid encryption mode indicates that the encryption mode cannot encrypt the independent data frame of the attribute.

[0119] In this example, chaotic processing indicates a process of performing behavior analysis and linear analysis on the valid ciphertext and then generating a digital sequence.

[0120] In this example, the random ciphertext is derived from the chaotic ciphertext sequence and is used to encrypt the independent data frame.

[0121] The working principle and beneficial effects of the technical solution are as follows: in order to guarantee the encryption effect, the system encrypts the real-time data twice, and needs to exclude invalid encryption modes first when performing the first encryption, and then performs hash processing on the remaining valid ciphertext to generate a corresponding chaotic ciphertext sequence, thereby guaranteeing the logicality between different weight ciphertexts and avoiding errors during encryption. Then, a corresponding segment is selected from the chaotic ciphertext sequence to construct random ciphertext to encrypt independent data frames, and the original independent data frames are covered. The target random ciphertext is selected by encrypting the random ciphertext and scoring the encryption effect, and then the real-time data is encrypted to obtain initial encrypted real-time data. In this way, each independent data frame in the real-time data can be encrypted, and the encryption modes used by different independent data frames are not necessarily the same, thereby greatly improving the encryption quality.

[0122] Embodiment 7

[0123] Based on the embodiment 6, the real-time data classification and hierarchical encryption system based on deep learning further comprises:

[0124] According to the encryption effect score, a plurality of encryption defect positions corresponding to the encrypted data frames are determined;

[0125] The encryption defect positions are located in the encrypted data frames, and the encryption defect positions are hidden.

[0126] The working principle and beneficial effects of the technical solution are as follows: in order to guarantee the security of real-time data and avoid leakage when encryption is not completed, the encryption defect positions in the encrypted data frames are hidden to avoid serious consequences caused by data loss.

[0127] Embodiment 8

[0128] Based on the embodiment 1, the real-time data classification and hierarchical encryption system based on deep learning, the deep encryption module comprises:

[0129] A mode screening unit is configured to determine invalid encryption modes of the real-time data according to the positive data representation, select a plurality of random encryption modes in the data set except the invalid encryption modes, and determine a historical use frequency corresponding to each random encryption mode.

[0130] An encryption preparation unit is configured to sort the random encryption modes in the order from less to more according to the historical use frequency to set a corresponding order level for the random encryption modes, and determine a plurality of encryptable data frames corresponding to each random encryption mode according to the encryption attribute corresponding to each random encryption mode and the data attribute of the independent data frame.

[0131] a local encryption unit configured to select an implementation random encryption mode for each of the independent data frames according to the order level of the random encryption mode corresponding to the independent data frame, and perform local encryption on the corresponding independent data frame by using the implementation random encryption mode;

[0132] a secret level determination unit configured to determine the secret level of the real-time data according to the data source of the real-time data, add a corresponding secret level label to the encrypted real-time data, and save the encrypted real-time data in a secret level storage area in the corresponding data set.

[0133] The working principle and beneficial effects of the above technical solution are as follows: when performing the second encryption, a valid random encryption mode of a data set is first randomly selected, and the historical use times of each random encryption mode are determined, then the independent data frames are locally encrypted according to the encryptable data frames of each random encryption mode and the order level thereof, and finally the secret level of the real-time data is determined and stored accordingly. Through such a way, the effectiveness of the encryption work can be further improved, and the security of the real-time data is guaranteed.

[0134] Embodiment 9

[0135] The embodiment provides a real-time data classification and hierarchical encryption method based on deep learning, as shown in Figure 2 , which comprises the following steps:

[0136] Step 1: dividing the real-time data into a plurality of independent data frames, and identifying the data attribute corresponding to each of the independent data frames;

[0137] Step 2: constructing a positive data representation corresponding to the real-time data according to the data attribute, and dividing the real-time data into a corresponding data set according to the positive data representation;

[0138] Step 3: constructing a random ciphertext sequence based on the data attribute, performing encryption training on the corresponding independent data frames, and constructing initial encrypted real-time data according to the encryption effect corresponding to each of the independent data frames;

[0139] Step 4: selecting a plurality of random encrypted modes contained in the data set to perform local encryption on the initial encrypted real-time data for a plurality of times, determining the secret level of the real-time data, and saving the real-time data.

[0140] In this example, the independent data frame represents the result of dividing the real-time data into sub-data;

[0141] In this example, the data attribute represents the attribute presented by the independent data frame;

[0142] In this example, the positive data representation represents the combined expression of the features and attributes of the real-time data;

[0143] In this example, the data set represents a data set used to store real-time data;

[0144] In this example, the initial encrypted real-time data represents the result of encrypting an independent data frame in the real-time data using a random ciphertext sequence;

[0145] In this example, local encryption refers to the process of encrypting a data portion in the initial encrypted real-time data using a random encryption method;

[0146] In this example, real-time data with corresponding encryption levels in the same data set are stored in the same area.

[0147] The working principle and beneficial effects of the above technical solution are as follows: In order to better classify and encrypt data, the collected real-time data is divided into several independent data frames, the data attributes of each independent data frame are analyzed to construct a positive data representation of real-time data, so that the data set to which the real-time data belongs can be determined, and the data classification work is completed by dividing it into the corresponding data set. Then, the data attributes are used to construct a random ciphertext sequence to encrypt and train the independent data frame, generating corresponding initial encrypted real-time data. In order to ensure encryption quality and further ensure the quality of real-time data encryption, the initial encrypted real-time data is locally encrypted using the random encryption method used by the existing in-cluster data in the data set. Firstly, it can compensate for encryption of weak parts, and secondly, it can use the encryption method in the same classification data set for deep encryption, improving the quality and efficiency of the encryption result, improving the overall encryption effect of real-time data, and achieving the purpose of synchronously generating data and completely encrypting in a short time.

[0148] Embodiment 10

[0149] Based on embodiment 9, the real-time data classification and hierarchical encryption method based on deep learning, step 2, comprises:

[0150] Step 21: respectively enhance the data attributes, determine a plurality of enhancement items of the real-time data according to the difference between the results before and after enhancement, find the relevant sub-data corresponding to each enhancement item in the real-time data, and assign values to the enhancement items according to the data values corresponding to the relevant sub-data, to generate a positive data representation of the real-time data;

[0151] Step 22: respectively acquire the in-cluster features and in-cluster data public information of each data set, construct a representation collection condition corresponding to the data set according to the in-cluster features, and construct an interval collection condition corresponding to the data set according to the in-cluster data public information;

[0152] Step 23: match the characterization collection condition with the positive data characterization, match the interval collection condition with the data specification of the real-time data, obtain a data set matched with the real-time data, and classify the real-time data according to the set category of the data set.

[0153] In this example, the enhancement processing represents a process of enhancing data attributes, and strengthening prominent key attributes therein;

[0154] In this example, the enhancement item represents a data sub that is enhanced;

[0155] In this example, the set-in feature represents a feature presented by the data set, and the set-in data public information represents information allowed to be publicly viewed by all set-in data in the data set;

[0156] In this example, the characterization collection condition represents a basic condition required to be met when data is collected into the data set;

[0157] In this example, the interval collection condition represents a data specification condition required to be met when data is collected into the data set;

[0158] In this example, the set category represents a public category to which the data set belongs, for example, the set category of the data set A is the financial category.

[0159] Working principle and beneficial effects of the above technical solution: in order to realize effective classification, the corresponding category of the real-time data is divided by first determining the enhancement item of the real-time data through the enhancement processing of the data attributes, further assigning values to the enhancement item by using the related sub data of the enhancement item, determining the positive data characterization of the real-time data, then processing each information of the data set, constructing the characterization collection condition and the interval collection condition of each data set, and then judging which data set the real-time data meets, and dividing the corresponding data category of the real-time data according to the set category of the data set, thus completing the purpose of classifying the real-time data. Through such a way, the data classification can be completed, and the collection conditions of each data set can be updated in real time, which facilitates the classification of multiple real-time data in a short time, and improves the intelligence of the system.

[0160] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the claims of the present application and their equivalents, the present application also intends to include these modifications and variations.

Claims

1. A real-time data classification and hierarchical encryption system based on deep learning, characterized in that, include: The frame processing module is used to divide real-time data into several independent data frames and identify the data attributes corresponding to each independent data frame. The data classification module is used to construct a positive data representation corresponding to the real-time data based on the data attributes, and to divide the real-time data into corresponding datasets based on the positive data representation. The preliminary encryption module is used to construct a random ciphertext sequence based on the data attributes to encrypt the corresponding independent data frames, and to construct the initial encrypted real-time data according to the encryption effect of each independent data frame. The deep encryption module is used to select several random encryption methods contained in the dataset to locally encrypt each independent data frame in the initially encrypted real-time data, determine the security level of the real-time data, and save it. The preliminary encryption module includes: The ciphertext construction unit is used to determine the invalid encryption method corresponding to the independent data frame based on the data attributes, delete the corresponding invalid ciphertext, determine several sets of valid ciphertexts of the real-time data, perform chaotic processing on the valid ciphertexts, and generate a chaotic ciphertext sequence of the real-time data. An encryption training unit is used to select sequence segments from the chaotic ciphertext sequence, obtain random ciphertext corresponding to each independent data frame, encrypt the corresponding independent data frame using the random ciphertext, cover the corresponding independent data frame with the obtained encrypted data frame, and score the encryption effect based on the coverage rate of each encrypted data frame and the corresponding random ciphertext. An encryption execution unit is configured to select the target random ciphertext with the highest score for each of the independent data frames, sort the target random ciphertext based on the sorting order of the independent data frames, generate the execution ciphertext of the real-time data, select an effective encryption method for the independent data according to the positive data representation, and encrypt the real-time data using the effective encryption method and the execution ciphertext to generate initially encrypted real-time data.

2. The deep learning based real-time data classification and grading encryption system of claim 1, wherein, The frame processing module includes: The compression processing unit is used to compress and encode real-time data, filter out several whitespace characters contained in the encoding result, and divide the real-time data accordingly based on the distribution characteristics of the whitespace characters in the encoding result to obtain several independent data frames. The preliminary screening unit is used to randomly select a corresponding number of datasets based on the number of frames of the independent data frames, perform similarity analysis on each independent data frame and different datasets to construct a similarity matrix of the real-time data, and construct an initial matching dataset for each independent data frame based on the elements corresponding to the same matrix position in different similar matrices. The attribute analysis unit is used to construct several data matching methods for the real-time data based on the frame sequence number of each independent data frame in the real-time data and the corresponding several initial matching datasets, obtain the initial matching data attributes corresponding to each dataset, and verify the attribute exclusion features corresponding to each data matching method respectively. The attribute determination unit is used to select a target data matching method with an attribute exclusion feature value of 0, and to determine the data attribute corresponding to each independent data frame.

3. The deep learning based real-time data classification and grading encryption system of claim 2, wherein, Also include: Iterative screening unit, for when all the attribute exclusion features are not 0, a corresponding number of update data sets are randomly selected according to the number of frame of the independent data frame, the update data set is transmitted to the preliminary screening unit for corresponding analysis.

4. The deep learning based real-time data classification and grading encryption system of claim 1, wherein, The data classification module comprises: Attribute processing unit, for respectively enhancing the data attributes, determining a plurality of enhancement items of the real-time data according to the difference between the results before and after enhancement, finding the corresponding relevant sub-data of each enhancement item in the real-time data, assigning values to the enhancement items according to the data values of the relevant sub-data, and generating positive data representation of the real-time data; Condition generation unit, for respectively acquiring the intra-set features and intra-set data public information of each data set, constructing the representation collection condition corresponding to the data set according to the intra-set features, and constructing the interval collection condition corresponding to the data set according to the intra-set data public information; Classification execution unit, for matching the representation collection condition with the positive data representation, matching the interval collection condition with the data specification of the real-time data, obtaining the data set matched with the real-time data, and classifying the real-time data according to the set category of the data set.

5. The deep learning based real-time data classification and grading encryption system of claim 4, wherein, The condition generation unit comprises: First generation subunit, for respectively acquiring the data basic information corresponding to different intra-set data in each data set, constructing strong sub-features corresponding to the data set according to the same information between the data basic information, acquiring the data type corresponding to each intra-set data, constructing weak sub-features corresponding to the data set in combination with the different information between the data basic information, and generating the intra-set features corresponding to the data set according to the strong sub-features and weak sub-features; Second generation subunit, for respectively acquiring the data public information corresponding to different intra-set data in each data set, determining the data maximum value and data minimum value corresponding to the intra-set data according to the data public information, acquiring a plurality of first test numbers greater than the data maximum value and a plurality of second test numbers less than the data minimum value, and analyzing the includable data range of the data set by using the first test number and the second test number; Third generation subunit, for constructing the representation collection condition of the data set according to a plurality of feature key points contained in the intra-set features, and constructing the interval collection condition of the data set according to the includable data range.

6. The deep learning based real-time data classification and rating encryption system of claim 1, wherein, Also include: According to the encryption effect score, a plurality of encryption defect positions corresponding to the encrypted data frame are determined; The encryption defect positions are located in the encrypted data frame, and the encryption defect positions are hidden.

7. The deep learning based real-time data classification and rating encryption system of claim 1, wherein, The deep encryption module comprises: Mode screening unit, for determining the invalid encryption mode of the real-time data according to the positive data representation, selecting a plurality of random encryption modes in the data set except the invalid encryption mode, and determining the historical use times corresponding to each random encryption mode; The encryption preparation unit is used for arranging the random encryption modes in the order from less to more according to the historical usage times, setting corresponding order levels for the random encryption modes, and determining a plurality of encryptable data frames corresponding to each random encryption mode according to the encryption attribute corresponding to each random encryption mode and the data attribute of the independent data frame; The local encryption unit is used for selecting a corresponding real-time random encryption mode for the independent data frame according to the order level of the random encryption mode corresponding to each independent data frame, and locally encrypting the corresponding independent data frame by using the real-time random encryption mode; The secret level determination unit is used for determining the secret level of the real-time data according to the data source of the real-time data, adding a corresponding secret level label to the encrypted real-time data, and inputting the encrypted real-time data to the secret level storage area in the corresponding data set for storage.

8. A real-time data classification and grading encryption method based on deep learning, characterized in that, Comprise: Step 1: divide the real-time data into a plurality of independent data frames, and identify the data attribute corresponding to each independent data frame; Step 2: construct a positive data representation corresponding to the real-time data according to the data attribute, and divide the real-time data into a corresponding data set according to the positive data representation; Step 3: encrypt the corresponding independent data frame based on the data attribute to construct a random ciphertext sequence, and construct the initial encrypted real-time data according to the encryption effect corresponding to each independent data frame; Step 4: select a plurality of random encryption modes contained in the data set to locally encrypt each independent data frame in the initial encrypted real-time data, determine the secret level of the real-time data, and store the secret level; Wherein, step 3, comprising: Step 31: determine the invalid encryption mode corresponding to the independent data frame according to the data attribute, delete the corresponding invalid ciphertext, determine a plurality of groups of valid ciphertexts of the real-time data, perform chaotic processing on the valid ciphertexts to generate a chaotic ciphertext sequence of the real-time data; Step 32: select a sequence segment in the chaotic ciphertext sequence to obtain a random ciphertext corresponding to each independent data frame, encrypt the corresponding independent data frame by using the random ciphertext, cover the corresponding independent data frame by using the obtained encrypted data frame, and score the encryption effect of the corresponding random ciphertext according to the coverage rate of each encrypted data frame; Step 33: select the target random ciphertext with the highest score for each independent data frame, sort the target random ciphertext based on the sorting order of the independent data frame to generate the execution ciphertext of the real-time data, select the effective encryption mode of the independent data according to the positive data representation, and encrypt the real-time data by using the effective encryption mode and the execution ciphertext to generate the initial encrypted real-time data.

9. The deep learning based real-time data classification and hierarchical encryption method of claim 8, wherein, The step 2, comprising: Step 21: respectively enhance the data attribute, determine a plurality of enhancement items of the real-time data according to the difference between the results before and after enhancement, find the relevant sub-data corresponding to each enhancement item in the real-time data, assign a value to the enhancement item according to the data value corresponding to the relevant sub-data, and generate the positive data representation of the real-time data; Step 22: obtaining the in-set features and in-set data public information of each data set respectively, constructing the representation collection condition corresponding to the data set according to the in-set features, and constructing the interval collection condition corresponding to the data set according to the in-set data public information; Step 23: matching the representation collection condition with the positive data representation, matching the interval collection condition with the data specification of the real-time data, obtaining the data set matched with the real-time data, and classifying the real-time data according to the set category of the data set.

Images