Large model detection method and related equipment
By binary encoding, splitting labels and unruly sorting of sensitive data in large models, the problems of non-compliance and sensitive data leakage during large model training are solved, and higher sensitive data recognition capabilities and security are achieved Compliance.
Patent Information
- Application Number
- CN202411438209.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-15
- Publication Date
- 2025-02-25
AI Technical Summary
Large models may experience non-compliance and sensitive data leakage during massive data training, resulting in security issues and data management vulnerabilities.
A large model detection method is proposed to detect the large model by binary encoding, splitting labels, sorting irregularly and generating new labels.
By double splitting and encoding combination of sensitive data, more unlabeled sensitive tags and content are generated, which enhances the identification ability of sensitive data and avoids data breaches and security compliance issues.
Smart Images
Figure CN119513592A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of large model detection, and in particular to a large model detection method and related equipment. Background Art
[0002] Big models refer to machine learning models with large-scale parameters and complex computational structures. These models are usually built with deep neural networks and have billions or even hundreds of billions of parameters. Big models are designed to improve the expressiveness and predictive performance of the model and to handle more complex tasks and data. Big models are widely used in various fields, including natural language processing, computer vision, speech recognition, and recommendation systems. Big models learn complex patterns and features by training massive amounts of data, have stronger generalization capabilities, and can make accurate predictions on unseen data.
[0003] However, when large models perform complex learning through massive data, they may become non-compliant. The root cause lies in loopholes in the management of training data. However, for the training of massive data, if one blindly pursues the control of training data, it will greatly inhibit the growth of large models. However, if left unchecked, it will cause security issues such as loss of control of large models and leakage of sensitive data. Summary of the invention
[0004] In view of this, the purpose of this application is to propose a large model detection method and related equipment.
[0005] Based on the above objectives, the present application provides a large model detection method, including:
[0006] Binary encode the sensitive data in the acquired large model to obtain an encoding result;
[0007] Based on the encoding result, split the obtained first tag into at least one sub-tag; the first tag is used to mark the sensitive data;
[0008] Randomly sorting the sub-tags, and generating a second tag according to the sorting result;
[0009] The large model is detected based on the second label.
[0010] In a possible implementation, the step of performing binary encoding on the acquired sensitive data in the large model to obtain an encoding result includes:
[0011] Based on the word length and word content of the sensitive data, the sensitive data is binary-encoded to obtain the encoding result; the binary code corresponding to the word length is the starting code of the encoding result.
[0012] In a possible implementation, splitting the acquired first tag into at least one sub-tag based on the encoding result includes:
[0013] Based on the word length, split the encoding result corresponding to the word content to obtain a split result;
[0014] The first tag is split based on a splitting rule corresponding to the splitting result to obtain the sub-tags.
[0015] In a possible implementation, the encoding result corresponding to the word content is split based on the word length to obtain the split result, including:
[0016] Based on the word length, the encoding result corresponding to the word content is split to obtain a first character and a second character; wherein the second character is a special character; and the first character is different from the second character.
[0017] In a possible implementation, the randomly sorting the sub-tags and generating the second tag according to the sorting result includes:
[0018] Randomly sorting the sub-tags, and generating new sensitive data based on the sorted sub-tags;
[0019] The second tag is generated based on the new sensitive data.
[0020] In a possible implementation, the randomly sorting the sub-tags and generating new sensitive data based on the sorted sub-tags includes:
[0021] Randomly sorting the sub-tags to obtain sorted sub-tags;
[0022] The binary code of the word length corresponding to the sub-tag in the sorted sub-tag is used as the first part of the binary code of the new sensitive data; the first part is the starting code in the binary code;
[0023] Using the binary encoding of the word content corresponding to the sub-tag in the sorted sub-tag as the second part of the binary encoding of the new sensitive data;
[0024] The binary encoding of the new sensitive data is decoded to obtain the new sensitive data.
[0025] Based on the same inventive concept, the embodiment of the present application also provides a large model detection device, including:
[0026] An encoding module is configured to perform binary encoding on the acquired sensitive data in the large model to obtain an encoding result;
[0027] a splitting module, configured to split the obtained first tag into at least one sub-tag based on the encoding result; the first tag is used to mark the sensitive data;
[0028] A sorting module, configured to sort the sub-tags randomly and generate a second tag according to the sorting result;
[0029] A detection module is configured to detect the large model based on the second label.
[0030] Based on the same inventive concept, an embodiment of the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a large model detection method as described in any one of the above items is implemented.
[0031] Based on the same inventive concept, an embodiment of the present application further provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute any of the above-mentioned large model detection methods.
[0032] Based on the same inventive concept, an embodiment of the present application further provides a computer program product, which includes computer program instructions, and the computer instructions are used to enable the computer program product to execute any of the large model detection methods described above.
[0033] From the above, it can be seen that the large model detection method and related equipment provided by the present application obtain a coding result by binary encoding the sensitive data in the acquired large model; based on the coding result, the acquired first label is split into at least one sub-label; the first label is used to mark the sensitive data; the sub-labels are randomly sorted, and a second label is generated according to the sorting result; the large model is detected based on the second label. The embodiment of the present application redefines the binary encoding of the sensitive data corresponding to the sensitive label, and splits the content corresponding to the sensitive data, thereby enriching the content of the sensitive label, avoiding the situation where symbols, characters or annotations cannot be effectively identified, and at the same time increasing the recognition ability of sensitive data, so that sensitive information can be mined. The present invention double-splits the sensitive content corresponding to the sensitive label, and at the same time generates more unlabeled sensitive labels and corresponding sensitive content based on the encoding combination of "word length" + "word content" (it can also rely on the above method to detect the security and compliance prevention capabilities of the large model. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] In order to more clearly illustrate the technical solutions in the present application or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0035] Figure 1 This is a schematic diagram of the large model detection method flow of an embodiment of the present application;
[0036] Figure 2 This is a schematic diagram of the structure of a large model detection device according to an embodiment of the present application;
[0037] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION
[0038] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.
[0039] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be the usual meanings understood by people with ordinary skills in the field to which the present application belongs. The "first", "second" and similar words used in the embodiments of the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing in front of the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0040] It is understandable that before using the technical solutions of each embodiment of the present disclosure, the type, scope of use, usage scenarios, etc. of the personal information involved will be informed to the user in an appropriate manner, and the user's authorization will be obtained.
[0041] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly remind the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can independently choose whether to provide personal information to software or hardware such as an electronic device, application, server, or storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.
[0042] As an optional but non-limiting implementation, in response to receiving the user's active request, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.
[0043] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that meet relevant laws and regulations may also be applied to the implementation of the present disclosure.
[0044] As mentioned in the background technology section, a big model refers to a machine learning model with large-scale parameters and complex computational structures. These models are usually built with deep neural networks and have billions or even hundreds of billions of parameters. Big models are designed to improve the expressiveness and predictive performance of the model and to handle more complex tasks and data. Big models are widely used in various fields, including natural language processing, computer vision, speech recognition, and recommendation systems. Big models learn complex patterns and features by training massive amounts of data, have stronger generalization capabilities, and can make accurate predictions on unseen data.
[0045] However, when large models perform complex learning through massive data, they may become non-compliant. The root cause lies in loopholes in the management of training data. However, for the training of massive data, if one blindly pursues the control of training data, it will greatly inhibit the growth of large models. However, if left unchecked, it will cause security issues such as loss of control of large models and leakage of sensitive data.
[0046] Based on the above considerations, the embodiment of the present application proposes a large model detection method, which obtains the encoding result by binary encoding the sensitive data in the acquired large model; splits the acquired first label into at least one sub-label based on the encoding result; the first label is used to mark the sensitive data; the sub-labels are randomly sorted, and a second label is generated according to the sorting result; the large model is detected based on the second label. The embodiment of the present application redefines the binary encoding of the sensitive data corresponding to the sensitive label, and splits the content corresponding to the sensitive data, thereby enriching the content of the sensitive label, avoiding the situation where symbols, characters or annotations cannot be effectively identified, and at the same time increasing the recognition ability of sensitive data, so that sensitive information can be mined. The present invention double splits the sensitive content corresponding to the sensitive label, and generates more unlabeled sensitive labels and corresponding sensitive content based on the encoding combination of "word length" + "word content", and can also rely on the above method to detect the security and compliance prevention capabilities of the large model.
[0047] The technical solutions of the embodiments of the present application are described in detail below through specific examples.
[0048] refer to Figure 1 The large model detection method of the embodiment of the present application comprises the following steps:
[0049] Step S101, binary encoding is performed on the sensitive data in the acquired large model to obtain an encoding result;
[0050] Step S102: split the obtained first tag into at least one sub-tag based on the encoding result; the first tag is used to mark the sensitive data;
[0051] Step S103, sorting the sub-tags randomly, and generating a second tag according to the sorting result;
[0052] Step S104: Detect the large model based on the second label.
[0053] With respect to step S101, the sensitive data in the acquired large model is binary-encoded to obtain an encoding result.
[0054] In some embodiments, the sensitive data in the acquired large model is binary-encoded to obtain an encoding result, including: based on the word length and word content of the sensitive data, the sensitive data is binary-encoded to obtain the encoding result; the binary code corresponding to the word length is the starting code of the encoding result.
[0055] In some embodiments, it is crucial to properly convert sensitive data in the initial stage of data processing to facilitate subsequent analysis and processing. This process can not only improve the availability of data, but also effectively protect privacy. In this regard, converting raw data into binary encoding is an effective strategy.
[0056] Binary encoding allows information to be stored and transmitted in a more compact and efficient way. By converting complex data types (such as text, images, or numbers) into binary form, we can reduce the space occupied by the data and improve computer processing efficiency. This is especially important for processing large data sets, because fast access and processing of data is the key to success in data analysis.
[0057] It also provides greater flexibility for subsequent data analysis. Many modern machine learning algorithms and data analysis tools perform better when processing binary data. This conversion can not only speed up the calculation process, but also improve the performance of the model, making the analysis results more reliable and effective.
[0058] In addition, from a security perspective, binary encoding can enhance data privacy protection. By representing sensitive information in binary form, we can better control access rights to the data and reduce the risk of data leakage. Even if the data is intercepted during transmission, the binary encoding form will make it difficult for unauthorized users to understand and use the data.
[0059] In summary, converting sensitive data into binary code not only lays a solid foundation for subsequent data analysis, but also improves data storage efficiency, analysis flexibility and security. This process plays an indispensable role in the initial stage of data processing and provides support for subsequent in-depth analysis and decision-making.
[0060] In this embodiment, sensitive data is first obtained:
[0061] Suppose the sensitive data that needs to be processed is an email address, such as johndoe@example.com.
[0062] After obtaining the sensitive data, calculate the word length:
[0063] After calculation, it is known that the above email address consists of 19 characters.
[0064] Furthermore, the word length 19 is converted to binary form: 10011.
[0065] In addition, for each character, use ASCII conversion and then convert to binary:
[0066] "j": ASCII 106, converted to binary 01101010;
[0067] "o": ASCII 111, converted to binary 01101111;
[0068] "h": ASCII 104, converted to binary 01101000;
[0069] "n": ASCII 110, converted to binary 01101110;
[0070] "d": ASCII 100, converted to binary 01100100;
[0071] "o": ASCII 111, converted to binary 01101111;
[0072] "e": ASCII 101, converted to binary 01100101;
[0073] "@": ASCII 64, converted to binary 01000000;
[0074] "e": ASCII 101, converted to binary 01100101;
[0075] "x": ASCII 120, converted to binary 01111000;
[0076] "a": ASCII 97, converted to binary 01100001;
[0077] "m": ASCII 109, converted to binary 01101101;
[0078] "p": ASCII 112, converted to binary 01110000;
[0079] "l": ASCII 108, converted to binary 01101100;
[0080] "e": ASCII 101, converted to binary 01100101;
[0081] ".": ASCII 46, converted to binary 00101110;
[0082] "c": ASCII 99, converted to binary 01100011;
[0083] "o": ASCII 111, converted to binary 01101111;
[0084] "m": ASCII 109, converted to binary 01101101;
[0085] Furthermore, the length code is combined with the binary code of each character to form a complete sequence: 10011 + the binary sequence corresponding to the above character.
[0086] Further, with respect to step S102, based on the encoding result, the obtained first tag is split into at least one sub-tag; the first tag is used to mark the sensitive data.
[0087] In some embodiments, by splitting the coding results, the data structure becomes clearer. This clear structure helps analysts quickly understand the relationship between each feature and its importance in the entire data set, so as to make more effective decisions. In this way, the data will not be confused by complexity during the analysis process, but will be more focused.
[0088] In some embodiments, based on the encoding result, the obtained first tag is split into at least one sub-tag, including: based on the word length, splitting the encoding result corresponding to the word content to obtain a split result; based on the splitting rule corresponding to the splitting result, splitting the first tag to obtain the sub-tag.
[0089] In some embodiments, the encoding result corresponding to the word content is split based on the word length to obtain a split result, including: based on the word length, the encoding result corresponding to the word content is split to obtain a first character and a second character; wherein the second character is a special character; the first character is different from the second character.
[0090] The result of the split includes the first characters johndoe and example.com, as well as the special character @.
[0091] It should be noted that the above-mentioned splitting method is one of the embodiments of the present application. For the convenience of subsequent explanation, in the actual splitting process, it can be split into single letters and other special characters. In this way, in the subsequent reorganization process, more types of new sensitive data can be formed, which is more beneficial for the detection of large models.
[0092] Furthermore, each part obtained by splitting will be labeled to ensure that the data can be accurately identified and processed. Naming each part can not only enhance the readability of the data, but also effectively support subsequent data analysis and operations.
[0093] The first character 1johndoe is renamed NameSegment1, and the corresponding classification gets the sub-label: NameSegment1-EmailSensitive.
[0094] The first character 2example.com is renamed DomainSegment2, and the corresponding classification obtains the sub-label: Connector-EmailSensitive.
[0095] The second character @ is renamed as Connector, and the corresponding classification gets the sub-label: DomainSegment2-EmailSensitive.
[0096] With respect to step S103, the sub-tags are randomly sorted, and a second tag is generated according to the sorting result.
[0097] Next, the generated sub-tags are randomly sorted to increase the complexity of sensitive data.
[0098] In some embodiments, the step of randomly sorting the sub-tags and generating a second tag based on the sorting result includes: randomly sorting the sub-tags and generating new sensitive data based on the sorted sub-tags; and generating the second tag based on the new sensitive data.
[0099] In some embodiments, the sub-tags are randomly sorted, and new sensitive data is generated based on the sorted sub-tags, including: randomly sorting the sub-tags to obtain sorted sub-tags; using the binary code of the word length corresponding to the sub-tag in the sorted sub-tags as the first part of the binary code of the new sensitive data; the first part is the starting code in the binary code; using the binary code of the word content corresponding to the sub-tag in the sorted sub-tags as the second part of the binary code of the new sensitive data; decoding the binary code of the new sensitive data to obtain the new sensitive data.
[0100] In this embodiment, the sorting is performed by randomization or a specific algorithm, for example, Connector-EmailSensitive is placed in front and NameSegment1-EmailSensitive is placed in the back.
[0101] The sorting result is: Connector-EmailSensitive-DomainSegment2-EmailSensitive-NameSegment1-EmailSensitive.
[0102] It should be noted that the step of randomly sorting the generated sub-tags here can, in actual applications, be randomly sorted as described in the above embodiment, or all sorting results can be generated first, and then sensitive data can be selected from the sorting results, because the purpose of this step is mainly to generate new sensitive data by randomly arranging the sub-tags, and the new sensitive data can be obtained by randomly permuting and combining the original sensitive data.
[0103] Furthermore, by sorting and reorganizing the tags, new sensitive data codes can be generated.
[0104] The first part is the binary code of the length of the sorted sub-tags. The second part is the binary code of the word content corresponding to the sub-tags.
[0105] In this embodiment, the obtained sorting order is as follows: Connector-EmailSensitive, DomainSegment2-EmailSensitive, NameSegment1-EmailSensitive.
[0106] The length of the sensitive data corresponding to each sub-tag needs to be converted into binary code.
[0107] Length calculation:
[0108] Connector corresponds to the character @ and has a length of 1.
[0109] DomainSegment2 corresponds to example.com and has a length of 11.
[0110] NameSegment1 corresponds to johndoe and has a length of 7.
[0111] Convert the length to binary:
[0112] @: 1 - binary 1;
[0113] example.com: 11 - binary 1011;
[0114] johndoe: 7-binary 111.
[0115] Convert the actual content of each subtag character by character into binary encoding.
[0116] The code is as follows:
[0117] @: ASCII64 - binary 01000000
[0118] example.com:
[0119] "e": ASCII 101 - binary 01100101
[0120] "x": ASCII120-binary 01111000
[0121] And so on, please refer to the contents in the above steps for details.
[0122] johndoe:
[0123] "j": ASCII106 - binary 01101010
[0124] And so on. For details, please refer to the contents in the above steps.
[0125] The length encoding and content encoding of each subtag are combined to form a complete binary sequence.
[0126] Example of a complete sequence:
[0127] Connector:1(length)+01000000(content).
[0128] DomainSegment2:1011(length)+
[0129] 0110010101111000011000010110110101011100000110110001100101(content).
[0130] NameSegment1:111(length)+
[0131] 011010100110111101101110011001000110111101100101(content).
[0132] Furthermore, the binary codes of all sub-tags are merged to construct a new sensitive data code.
[0133] Final combination:
[0134] Combine the binary encodings of all labels in sorted order:
[0135] 1010000001011011001010111100001100001011011010111000001101100101011110110101001101010100100011011110110010101.
[0136] Then, based on the new sensitive data, a second label is generated.
[0137] In this embodiment, based on the newly generated sensitive data, a unique second tag is created for identification and protection in the large model.
[0138] Specifically, a unique second label is generated based on the characteristics of the new sensitive data.
[0139] For example, EmailDetectionTag is generated based on Connector-EmailSensitive-DomainSegment2-EmailSensitive-NameSegment1-EmailSensitive.
[0140] It should be noted that the long description mentioned above can be used directly here, or it can be renamed as a short description. Both expressions are acceptable and there is no corresponding restriction here.
[0141] It can be seen from the above embodiments that the large model detection method described in the embodiment of the present application obtains the encoding result by binary encoding the sensitive data in the acquired large model; based on the encoding result, the acquired first label is split into at least one sub-label; the first label is used to mark the sensitive data; the sub-labels are randomly sorted, and a second label is generated according to the sorting result; the large model is detected based on the second label. The embodiment of the present application redefines the binary encoding of the sensitive data corresponding to the sensitive label, and splits the content corresponding to the sensitive data, thereby enriching the content of the sensitive label, avoiding the situation where symbols, characters or annotations cannot be effectively identified, and at the same time increasing the recognition ability of sensitive data, so that sensitive information can be mined. The present invention double-splits the sensitive content corresponding to the sensitive label, and generates more unlabeled sensitive labels and corresponding sensitive content based on the encoding combination of "word length" + "word content", and can also rely on the above method to detect the security and compliance prevention capabilities of the large model.
[0142] It should be noted that the method of the embodiment of the present application can be performed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present application, and the multiple devices will interact with each other to complete the described method.
[0143] It should be noted that the above describes some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0144] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a large model detection device.
[0145] refer to Figure 2 , the large model detection device comprises:
[0146] The encoding module 21 is configured to perform binary encoding on the sensitive data in the acquired large model to obtain an encoding result;
[0147] A splitting module 22 is configured to split the obtained first tag into at least one sub-tag based on the encoding result; the first tag is used to mark the sensitive data;
[0148] A sorting module 23 is configured to sort the sub-tags randomly and generate a second tag according to the sorting result;
[0149] The detection module 24 is configured to detect the large model based on the second label.
[0150] For the convenience of description, the above device is described in terms of functions divided into various modules. Of course, when implementing the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.
[0151] The device of the above embodiment is used to implement the corresponding large model detection method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.
[0152] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the large model detection method described in any of the above embodiments is implemented.
[0153] Figure 3A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other through the bus 1050 in the device.
[0154] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.
[0155] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.
[0156] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.
[0157] The communication interface 1040 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).
[0158] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).
[0159] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.
[0160] The electronic device of the above embodiment is used to implement the corresponding large model detection method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.
[0161] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the large model detection method described in any of the above embodiments.
[0162] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.
[0163] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the large model detection method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0164] Based on the same inventive concept, corresponding to the large model detection method described in any of the above embodiments, the present disclosure also provides a computer program product, which includes computer program instructions. In some embodiments, the computer program instructions can be executed by one or more processors of a computer so that the computer and / or the processor executes the large model detection method. Corresponding to the execution subject corresponding to each step in each embodiment of the large model detection method, the processor that executes the corresponding step may belong to the corresponding execution subject.
[0165] The computer program product of the above embodiment is used to enable the computer and / or the processor to execute the large model detection method described in any of the above embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0166] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application (including the claims) is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.
[0167] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present application difficult to understand, the known power supply / ground connection with the integrated circuit (IC) chip and other components may or may not be shown in the provided drawings. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present application difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform to be implemented in the embodiments of the present application (that is, these details should be fully within the scope of understanding of those skilled in the art). In the case of elaborating specific details (e.g., circuits) to describe exemplary embodiments of the present application, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.
[0168] Although the present application has been described in conjunction with specific embodiments of the present application, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.
[0169] The embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the scope of protection of the present application.
Claims
1. A large model detection method, characterized in that: include: Binary encode the sensitive data in the acquired large model to obtain an encoding result; Based on the encoding result, split the obtained first tag into at least one sub-tag; The first label is used to mark the sensitive data; Randomly sorting the sub-tags, and generating a second tag according to the sorting result; The large model is detected based on the second label.
2. The method according to claim 1, characterized in that: The sensitive data in the acquired large model is binary-encoded to obtain the encoding result, including: Based on the word length and word content of the sensitive data, the sensitive data is binary-encoded to obtain the encoding result; the binary code corresponding to the word length is the starting code of the encoding result.
3. The method according to claim 2, characterized in that The step of splitting the acquired first tag into at least one sub-tag based on the encoding result includes: Based on the word length, split the encoding result corresponding to the word content to obtain a split result; The first tag is split based on a splitting rule corresponding to the splitting result to obtain the sub-tags.
4. The method according to claim 3, characterized in that The step of splitting the encoding result corresponding to the word content based on the word length to obtain the split result includes: Based on the word length, the encoding result corresponding to the word content is split to obtain a first character and a second character; wherein the second character is a special character; and the first character is different from the second character.
5. The method according to claim 2, characterized in that: The step of randomly sorting the sub-tags and generating a second tag according to the sorting result includes: Randomly sorting the sub-tags, and generating new sensitive data based on the sorted sub-tags; The second tag is generated based on the new sensitive data.
6. The method according to claim 5, characterized in that The step of randomly sorting the sub-tags and generating new sensitive data based on the sorted sub-tags includes: Randomly sorting the sub-tags to obtain sorted sub-tags; The binary code of the word length corresponding to the sub-tag in the sorted sub-tag is used as the first part of the binary code of the new sensitive data; the first part is the starting code in the binary code; Using the binary encoding of the word content corresponding to the sub-tag in the sorted sub-tag as the second part of the binary encoding of the new sensitive data; The binary encoding of the new sensitive data is decoded to obtain the new sensitive data.
7. A large model detection device, characterized in that: include: An encoding module is configured to perform binary encoding on the acquired sensitive data in the large model to obtain an encoding result; a splitting module, configured to split the acquired first tag into at least one sub-tag based on the encoding result; The first label is used to mark the sensitive data; A sorting module, configured to sort the sub-tags randomly and generate a second tag according to the sorting result; A detection module is configured to detect the large model based on the second label.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.
9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 6.
10. A computer program product, comprising computer program instructions, which, when executed on a computer, cause the computer to execute the method according to any one of claims 1 to 6.