Sample data access method and device of large model and related equipment

Abstract

The invention provides a sample data access method and device for a large model and related equipment, and belongs to the technical field of network security, and the method comprises the steps: employing a progressive key generation mode to determine public keys and private keys corresponding to a plurality of distributed data source nodes; based on the first test data corresponding to the plurality of distributed data source nodes and the public key and the private key corresponding to the plurality of distributed data source nodes, performing a tamper-free attack aggregation detection operation to obtain a detection result; generating an encrypted ciphertext and a sequence bit aggregation key according to the detection result and the total sample data corresponding to the plurality of distributed data source nodes; and sending the encrypted ciphertext to a large model according to the sequence bit aggregation key, and obtaining full sample data corresponding to the plurality of distributed data source nodes by the large model according to the encrypted ciphertext and a preset decryption algorithm. According to the method, the security of large model sample data access can be improved.

Description

Technical Field

[0001] This application relates to the field of network security technology, and in particular to a method, apparatus and related equipment for obtaining sample data of a large model. Background Technology

[0002] With the advent of the era of general artificial intelligence, the artificial intelligence industry is developing rapidly. Currently, large models (such as Large Language Models (LLM), Inception convolutional neural network structures in computer vision, and Deep Q-Network models) can improve performance through distributed training; however, in the actual process of optimizing large models using different distributed nodes, the security of each distributed data node is uncontrollable, and some malicious data nodes may launch data poisoning attacks, resulting in low security for the use of sample data for large models. Summary of the Invention

[0003] This application provides a method, apparatus, and related equipment for obtaining sample data from large models, which can solve the technical problem of low security in obtaining sample data from large models.

[0004] In a first aspect, embodiments of this application provide a method for personalized service content recommendation, the method comprising:

[0005] A progressive key generation method is used to determine the public and private keys for each of the multiple distributed data source nodes.

[0006] Obtain the first test data corresponding to each of the multiple distributed data source nodes;

[0007] Based on the first test data corresponding to each of the multiple distributed data source nodes, and the public and private keys corresponding to each of the multiple distributed data source nodes, a tamper-free attack aggregation detection operation T-PAADO is performed to obtain the detection results;

[0008] Based on the detection results and the full sample data corresponding to each of the multiple distributed data source nodes, an encrypted ciphertext and a sequence bit aggregation key are generated.

[0009] The encrypted ciphertext is sent to the large model according to the sequence bit aggregation key, and the large model obtains the full sample data corresponding to each of the multiple distributed data source nodes according to the encrypted ciphertext and the preset decryption algorithm.

[0010] Optionally, the step of determining the public and private keys corresponding to each of the multiple distributed data source nodes using a progressive key generation method includes:

[0011] Based on the node identification information and hash function of the first data source node, determine the intermediate parameters of the first data source node;

[0012] Based on the intermediate parameters and the pre-acquired system master key, the private key of the first data source node is determined, and based on the first preset random parameters of the first data source node and the generator of the pre-acquired cyclic addition group, the public key of the first data source node is determined.

[0013] Based on the node identifier information of the second data source node, the public key of the first data source node and the hash function, the public key of the second data source node is determined, and based on the first preset random parameter of the second data source node, the system master key and the public key of the second data source node, the private key of the second data source node is determined.

[0014] The first data node and the second data source node are any one of the plurality of distributed data source nodes.

[0015] Optionally, the method further includes:

[0016] The private key of the third data source node is determined based on the node identifier information of the third data source node and the system master key;

[0017] The public key of the third data source node is determined based on the system master key and the generator.

[0018] The third data source node is any one of the multiple distributed data source nodes.

[0019] Optionally, the step of performing T-PAADO based on the first test data corresponding to each of the plurality of distributed data source nodes, and the public and private keys corresponding to each of the plurality of distributed data source nodes, to obtain the detection result includes:

[0020] Based on the first test data corresponding to each of the multiple distributed data source nodes, the public key corresponding to each of the multiple distributed data source nodes, the private key corresponding to each of the multiple distributed data source nodes, and preset system parameters, multiple ciphertexts corresponding to each of the multiple distributed data source nodes are determined respectively. The preset system parameters include a preset prime number, a cyclic additive group of the order of the preset prime number, the generator of the cyclic additive group, a preset system public key, and a hash function. The preset system public key is determined by the preset system master key and the generator.

[0021] The multiple ciphertexts corresponding to each of the multiple distributed data source nodes are aggregated to obtain the aggregable ciphertexts corresponding to each of the multiple distributed data source nodes.

[0022] The potential attack content of obfuscated multi-ciphertext was tested on the aggregateable signatures corresponding to the multiple distributed data source nodes respectively, and the detection results were obtained.

[0023] Optionally, generating encrypted ciphertext based on the detection results and the full sample data from the multiple distributed data source nodes includes the following:

[0024] If the test result is that the test fails, the encrypted ciphertext generation operation will be terminated.

[0025] If the test result is passed, the full sample data of the multiple distributed data source nodes and the aggregateable signatures of the multiple distributed data source nodes are concatenated to obtain the concatenation results corresponding to each of the multiple distributed data source nodes. The concatenation results corresponding to each of the multiple distributed data source nodes are then aggregated to obtain the encrypted ciphertext.

[0026] Optionally, generating the sequential bit aggregation key based on the detection result and the full sample data of the multiple distributed data source nodes includes the following:

[0027] If the test result is "test fails", stop the sequence bit aggregation key generation operation;

[0028] If the test result is passed, a first segment key is generated. If the concatenation results of multiple distributed data source nodes are aggregated to obtain encrypted ciphertext, a second segment key is generated. The first segment key and the second segment key are concatenated to obtain a sequence bit aggregation key.

[0029] Secondly, embodiments of this application provide a sample data acquisition device for a large model, the device comprising:

[0030] The first processing module is used to determine the public and private keys of each of the multiple distributed data source nodes using a progressive key generation method.

[0031] The acquisition module is used to acquire the first test data corresponding to each of the multiple distributed data source nodes;

[0032] The second processing module is used to perform T-PAADO based on the first test data corresponding to each of the multiple distributed data source nodes, as well as the public and private keys corresponding to each of the multiple distributed data source nodes, to obtain the detection results;

[0033] The third processing module is used to generate encrypted ciphertext and sequence bit aggregation key based on the detection results and the full sample data corresponding to each of the multiple distributed data source nodes;

[0034] The fourth processing module is used to send the encrypted ciphertext to the large model according to the sequence bit aggregation key, and the large model obtains the full sample data corresponding to each of the multiple distributed data source nodes according to the encrypted ciphertext and the preset decryption algorithm.

[0035] Thirdly, embodiments of this application provide an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, it implements the steps of the large model sample data retrieval method as described in the first aspect.

[0036] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the large model sample data retrieval method as described in the first aspect.

[0037] Fifthly, embodiments of this application provide a computer program product, including computer instructions that, when executed by a processor, implement the steps of the sample data retrieval method for a large model as described in the first aspect.

[0038] In this embodiment, by using progressive key generation and tamper-free attack aggregation detection operations, the data security of each distributed data source node is detected before generating encrypted ciphertext, preventing the large model from using sample data that has been tampered with or is at risk of attack. In addition, encrypted ciphertext and sequence bit aggregation keys are generated based on the detection results and the full sample data of each node, so that the large model does not need to expose the original plaintext data when using data, and can ensure that the data has not been tampered with during the data access process, thereby further improving the security of large model sample data access. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating a method for obtaining sample data from a large model, as provided in an embodiment of this application.

[0041] Figure 2 This is a schematic diagram of a sample data acquisition system for a large model provided in an embodiment of this application;

[0042] Figure 3 This is a schematic diagram of the structure of a personalized service recommendation device provided in an embodiment of this application;

[0043] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0044] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0045] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, not limited in number; for example, the first object can be one or more. Furthermore, "and / or" in this application indicates at least one of the connected objects. For example, the scope of protection of "A and / or B" covers at least three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B. Additionally, the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0046] See Figure 1 , Figure 1 This is a flowchart illustrating a method for obtaining sample data from a large model, as provided in an embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0047] Step 101: Use a progressive key generation method to determine the public and private keys corresponding to each of the multiple distributed data source nodes;

[0048] The progressive key generation method can be that the key of the next data source node is determined by the key of the previous data source node, and both the previous data source node and the next data source node belong to the multiple distributed data source nodes; the key can be a public key and / or a private key.

[0049] The multiple distributed data source nodes can be different nodes that provide different training samples and can be used to store data in different scenarios; for example, distributed data source nodes can be financial databases, economic databases, integrated databases that provide financial data, traffic databases that provide pedestrian flow data, and / or language dialogue databases that provide dialogue text, etc.

[0050] The financial database may include: databases that provide financial market data such as stocks, bonds, foreign exchange, futures, and funds; enterprise databases and / or macro-enterprise databases (EDBs);

[0051] The economic database can be a global macroeconomic and industry economic big data platform, used to provide Chinese macroeconomic data, global macroeconomic data, and industry economic data, etc.

[0052] The traffic database that provides pedestrian flow data can be used to provide daily floating car location data, which can be used for traffic flow analysis; the traffic database that provides pedestrian flow data can provide population data (passenger flow, population profile, population movement, etc.) and / or traffic data (congestion distance, average speed, etc.).

[0053] The language dialogue database can be a comprehensive collection of open-source Chinese datasets, encompassing most tasks related to natural language understanding and generation, and providing a variety of open-source datasets; the language dialogue database can contain language dialogue-related data;

[0054] The databases mentioned above can be relational databases (e.g., Structured Query Language (SQL) databases), unstructured databases, semi-structured databases, and non-relational (Not Only NoSQL, NoSQL) databases.

[0055] In this step, a progressive key generation method is used to determine the public and private keys for each of the multiple distributed data source nodes, which increases the security and unpredictability of the public and private keys.

[0056] Step 102: Obtain the first test data corresponding to each of the multiple distributed data source nodes;

[0057] The first test data corresponding to each of the multiple distributed data source nodes can be a part of the full sample data corresponding to each of the multiple distributed data source nodes.

[0058] The step of obtaining the first test data corresponding to each of the multiple distributed data source nodes can be: randomly downloading a portion of data from each of the multiple distributed data source nodes to obtain the first test data corresponding to each of the multiple distributed data source nodes.

[0059] In this step, the first test data corresponding to each of the multiple distributed data source nodes is obtained, thereby providing an accurate data foundation for the large model to use data.

[0060] Step 103: Based on the first test data corresponding to each of the multiple distributed data source nodes, and the public and private keys corresponding to each of the multiple distributed data source nodes, perform a tamper-proof attack aggregation detection operation (T-PAADO) to obtain the detection results;

[0061] The T-PAADO can be obtained by extracting non-full data (e.g., the first test data) from data samples (e.g., the full sample data mentioned above) and performing single-point aggregation on the non-full data, thereby detecting whether there are abnormal points in the aggregated data, and further obtaining the security of the data on the multiple distributed data source nodes.

[0062] It is worth noting that T-PAADO itself can also be used as an independent security detection algorithm;

[0063] In this step, the T-PAADO process introduces the public and private keys corresponding to the multiple distributed data source nodes, enabling encrypted verification and integrity checks of test data from different distributed data source nodes. This ensures that the data source is trustworthy and the content has not been tampered with, achieving tamper-free attack aggregation detection. This step can detect potential data tampering, malicious injection, or abnormal distribution without exposing all original data (such as the full sample data mentioned above), thereby assessing the security of data at each node with lightweight overhead and providing a secure data foundation for large models to use sample data.

[0064] Step 104: Based on the detection results and the full sample data corresponding to each of the multiple distributed data source nodes, generate encrypted ciphertext and sequence bit aggregation key;

[0065] The encrypted ciphertext can be obtained by aggregating and concatenating the detection results and the full sample data corresponding to each of the multiple distributed data source nodes; the encrypted ciphertext can include all the information of the full sample data;

[0066] In this step, based on the detection results and the full sample data corresponding to each of the multiple distributed data source nodes, encrypted ciphertext and sequential bit aggregation key are generated. This can determine the confidentiality and integrity of the sample data during transmission and storage, thereby ensuring the security and reliability of the large model's use of sample data.

[0067] Step 105: Send the encrypted ciphertext to the large model according to the sequence bit aggregation key, and the large model obtains the full sample data corresponding to each of the multiple distributed data source nodes according to the encrypted ciphertext and the preset decryption algorithm.

[0068] The large model refers to a deep learning model supported by a large number of parameters and / or a large amount of training data, which can learn high-dimensional feature representations on general or specific tasks and has strong generalization ability and complex task processing ability.

[0069] Typical large models can be pre-trained LLMs, which can be used for natural language understanding and generation; convolutional neural network models based on the Inception architecture, which can be used for image feature extraction and recognition; or deep Q-network (DQN) models, which can be used for policy learning and decision-making in reinforcement learning scenarios.

[0070] It should be noted that the types of large models mentioned above are not limited examples. The large models can also be other deep learning models that meet the preset parameter scale, computational complexity and performance requirements, such as other forms of Transformer models, convolutional neural network models, graph neural network models, sequence decision models, etc.

[0071] The preset decryption algorithm may include the inverse operation of T-PAADO, or may be set by those skilled in the art as needed.

[0072] In this embodiment, by using progressive key generation and tamper-free attack aggregation detection operations, the data security of each distributed data source node is detected before generating encrypted ciphertext, preventing the large model from using sample data that has been tampered with or is at risk of attack. In addition, encrypted ciphertext and sequence bit aggregation keys are generated based on the detection results and the full sample data of each node, so that the large model does not need to expose the original plaintext data when using data, and can ensure that the data has not been tampered with during the data access process, thereby further improving the security of large model sample data access.

[0073] In some implementations, the step of using a progressive key generation method to determine the public and private keys corresponding to each of the multiple distributed data source nodes includes:

[0074] Based on the node identification information and hash function of the first data source node, determine the intermediate parameters of the first data source node;

[0075] Based on the intermediate parameters and the pre-acquired system master key, the private key of the first data source node is determined, and based on the first preset random parameters of the first data source node and the generator of the pre-acquired cyclic addition group, the public key of the first data source node is determined.

[0076] Based on the node identifier information of the second data source node, the public key of the first data source node and the hash function, the public key of the second data source node is determined, and based on the first preset random parameter of the second data source node, the system master key and the public key of the second data source node, the private key of the second data source node is determined.

[0077] The first data node and the second data source node are any one of the plurality of distributed data source nodes.

[0078] The pre-acquired system master key and the generator of the pre-acquired cyclic addition group can be part of the preset system parameters and can be set as needed by those skilled in the art;

[0079] The preset system parameters may include a preset prime number, a cyclic additive group of the preset prime number, a generator of the cyclic additive group, a preset system public key, and a hash function. The preset system public key is determined by the preset system master key and the generator. The preset system parameters can be set as needed by those skilled in the art.

[0080] In this embodiment, intermediate parameters are determined based on node identification information and hash function, and the public and private keys of the first data source node are generated under the combined action of the system master key and the generator of the cyclic addition group. Furthermore, the public and private keys of the second data source node are generated progressively using the public key of the first data source node, node identification information and system master key.

[0081] On the one hand, it can ensure that the keys of different data source nodes are strongly bound to their unique node identifiers, preventing key forgery and identity impersonation, and enhancing the verifiability of node identities and the consistency of key management. On the other hand, by introducing the public key of the previous node to participate in the key generation of subsequent nodes, a progressive associated key structure is realized, which enables the key system to have traceability and overall controllability while ensuring the independence of each node's key. This is conducive to achieving higher security and anti-attack capabilities in the generation and transmission of detection results and encrypted ciphertext, thereby improving the security of large model sample data acquisition.

[0082] It should be noted that the first data source node and the second data source node can be specified by those skilled in the art according to specific business needs or actual application scenarios, and can also be nodes with an associated relationship. In addition, the selection method of the first data source node and the second data source node can be extended to the correspondence between multiple sets of nodes (that is, the multiple distributed data source nodes may include multiple first data source nodes and second data source nodes), or to continue to generate more data source node key pairs (the key pair includes a public key and a private key) by continuing to advance along the node sequence.

[0083] In some embodiments, the method further includes:

[0084] The private key of the third data source node is determined based on the node identifier information of the third data source node and the system master key;

[0085] The public key of the third data source node is determined based on the system master key and the generator.

[0086] The third data source node is any one of the multiple distributed data source nodes.

[0087] The plurality of distributed data source nodes may include one or more third data source nodes;

[0088] In this embodiment, by using the same system master key and generator key generation mechanism, a consistent and scalable key system is formed among multiple distributed data source nodes. This facilitates the inclusion of third data source nodes into the overall security framework during subsequent distributed encrypted transmission, aggregation, and secure computation, thereby further improving the security of sample data access in the large model.

[0089] For details, please see Figure 2 , Figure 2 This is a schematic diagram of a big data sample data retrieval system provided in an embodiment of this application, which can be used to implement the big data sample data retrieval method, such as... Figure 2 As shown, the system may include: a key security center, multiple distributed data source nodes, test nodes, aggregation nodes, distribution nodes, and a large model;

[0090] The key security center is used to generate public and private keys corresponding to each of the multiple distributed data source nodes, and to send the public and private keys of the multiple distributed data source nodes to the multiple distributed data source nodes;

[0091] The plurality of distributed data source nodes are used to obtain the public and private keys corresponding to each of the plurality of distributed data source nodes, and to provide their respective first test data;

[0092] The test node and the aggregation node perform a tamper-free attack aggregation detection operation T-PAADO based on the first test data corresponding to each of the multiple distributed data source nodes, as well as the public and private keys corresponding to each of the multiple distributed data source nodes, to obtain the detection results. Based on the detection results and the full sample data corresponding to each of the multiple distributed data source nodes, they generate encrypted ciphertext and sequence bit aggregation key.

[0093] The distribution node is used to send the encrypted ciphertext and the sequence bit aggregation key to the large model;

[0094] The large model is used to obtain the full sample data corresponding to each of the multiple distributed data source nodes based on the encrypted ciphertext and the preset decryption algorithm;

[0095] In this application, the key security center may be simply referred to as the security center, and the distributed data source node may be simply referred to as the data source node;

[0096] Optional, please see Figure 2 The big data sample data acquisition system may also include a blockchain; the blockchain is used to receive authentication confirmation information sent by the multiple distributed data source nodes and then distribute the encrypted ciphertext (described later) to the distribution nodes.

[0097] For details, please see Figure 2 As the main control node, the security center needs to initialize the system parameter Params before generating keys (including public and private keys);

[0098] It should be noted that, in the above Figure 2 In this context, distributed data source node 1 is also referred to as the first data source node or the first distributed data source node; distributed data source node 2 is also referred to as the second data source node or the second distributed data source node; and distributed data source node 3 is also referred to as the third data source node or the third distributed data source node.

[0099] In the application, the distributed data source node may also be referred to simply as a data source node;

[0100] The security center can acquire preset security parameters. , large prime numbers , ( For the order of (cyclic addition group) yes Generator, system master key s ( (a set of positive integers) and system public key ( ), hash function, to determine system parameter Params= .

[0101] In this application, the large prime numbers are represented in lowercase. The generator is indicated by uppercase letters. express;

[0102] The hash function includes a first hash function. Second hash function Third hash function Fourth hash function Fifth hash function and the sixth hash function :

[0103] Specifically, the hash function can be found in the following formula:

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110] in, Used to represent bit strings of arbitrary length; Used to indicate a length of The bit string can also be used to represent the length of the ciphertext; Used to indicate a length of The bit string; Represents the set of positive integers; "Used to represent the mapping relationship of the corresponding hash function."

[0111] Specifically, taking the multiple distributed data source nodes including a first data source node, a second data source node, and a third data source node as an example; the step of determining the public and private keys corresponding to each of the multiple distributed data source nodes using a progressive key generation method may include the following:

[0112] The node representation of each distributed data source node contains node identification information. ( Used to indicate the first The node identifier information of each data source node is uploaded to the security center. The security center calculates the public and private keys corresponding to the first, second, and third data source nodes. The calculation process specifically includes:

[0113] Determine intermediate parameters , Used to indicate the first Intermediate parameters of each data source node;

[0114] Determine the private key of the first data source node , The entire key is used to represent the private key of the first data source node. middle Used to represent the system master key. Intermediate parameters used to represent the first data source node ;

[0115] The first preset random parameters of the multiple distributed data source nodes are obtained in advance. For example, the first preset random parameter of the first data source node can be used This indicates that the first preset random parameter of the second data source node can be used This indicates that, and so on;

[0116] Determine the public key of the first data source node , The public key used to represent the first data source node. middle Used to represent the generator. The first preset random parameter used to represent the first data source node ;

[0117] Determine the public key of the second data source node ; The public key used to represent the second data source node; "Used to indicate a concatenation character, Indicates will and splicing The entire structure represents the public key of the first data source node. Used to represent , Used to represent the first hash function mentioned above;

[0118] Determine the private key of the second data source node ; The entire string represents the private key of the second data source node; middle Used to represent , middle Used to represent the system master key. The public key used to represent the second data source node;

[0119] Determine the private key of the third data source node ; The entire key is used to represent the private key of the third-party data source node. Used to represent , Used to represent the system master key, "Used to indicate a concatenation character, Used to indicate that and splicing;

[0120] Determine the public key of the third data source node ; The entire structure represents the public key of the third-party data source node. The entire key is used to represent the private key of the third-party data source node. Used to represent the generator element;

[0121] The security center obtains the public keys corresponding to the first data source node, the second data source node, and the third data source node. And the private keys corresponding to the first data source node, the second data source node, and the third data source node. ;

[0122] More specifically, the security center can use the private key Based on each distributed data source node Send to each of the data source nodes;

[0123] It should be noted that the above explanation uses only three data source nodes as an example; in actual applications, the number of data source nodes is unlimited. When there are multiple data source nodes, the above process can be executed cyclically to generate and distribute a public and private key for each distributed data source node. This ensures that only authorized nodes can access and use the corresponding keys, minimizing the risk of key leakage.

[0124] In some implementations, the step of performing T-PAADO based on the first test data corresponding to each of the plurality of distributed data source nodes, and the public and private keys corresponding to each of the plurality of distributed data source nodes, to obtain the detection result includes:

[0125] Based on the first test data corresponding to each of the multiple distributed data source nodes, the public key corresponding to each of the multiple distributed data source nodes, the private key corresponding to each of the multiple distributed data source nodes, and preset system parameters, multiple ciphertexts corresponding to each of the multiple distributed data source nodes are determined respectively. The preset system parameters include a preset prime number, a cyclic additive group of the order of the preset prime number, the generator of the cyclic additive group, a preset system public key, and a hash function. The preset system public key is determined by the preset system master key and the generator.

[0126] The multiple ciphertexts corresponding to each of the multiple distributed data source nodes are aggregated to obtain the aggregable ciphertexts corresponding to each of the multiple distributed data source nodes.

[0127] The potential attack content of obfuscated multi-ciphertext was tested on the aggregateable signatures corresponding to the multiple distributed data source nodes respectively, and the detection results were obtained.

[0128] In this embodiment, based on the first test data corresponding to each of the multiple distributed data source nodes and their respective public and private keys, multiple ciphertexts are generated under the constraints of preset system parameters. The multiple ciphertexts corresponding to each node are aggregated to obtain their respective aggregable ciphertexts. Based on the aggregable ciphertexts, an equivalent test of the potential attack content of the obfuscated multi-ciphertexts is performed, thereby realizing the detection of potential attack content without exposing the original test data of each node.

[0129] On the one hand, it ensures the privacy and key security of the multiple distributed data source nodes during data transmission and processing, reducing the risk of key and sensitive data leakage; on the other hand, it improves the computation and storage efficiency of the detection process through ciphertext aggregation and equivalence testing mechanisms.

[0130] For details, please see Figure 2 Taking the first data source node of the big data sample data acquisition system as an example to illustrate how aggregateable signatures are obtained, the process by which the other multiple distributed data source nodes obtain detection results is the same as that of the first data source node; wherein, the process by which the first data source node obtains the detection results includes the following:

[0131] Pre-request the second random parameter corresponding to the first data source node Third random parameter and the fourth random parameter ; It is a set of positive integers;

[0132] Determine the first ciphertext of the first data source node ;

[0133] Determine the first ciphertext of the first data source node ;

[0134] Determine intermediate data ;

[0135] in, and The ciphertext for the verification encrypted segment generated based on the system parameters. For intermediate data, This is the above cyclic addition group generator, The system public key; The intermediate parameters are: ;

[0136] Determine message integrity value : ;

[0137] Determine the message hash value : ;

[0138] and Taking the first data source node as an example, based on the scenario in this solution, when the distributed data source nodes are the second and third data source nodes; The number of the distributed data source node. Node identification information used to represent the first data source node; Used to represent constant values ​​that can be set as needed by those skilled in the art; The key corresponds to the public key of different distributed data source nodes, for example... This is the public key of the first data source node; This is the first test data for the first data source node; The second hash function in the system parameter Params. It is the third hash function in the system parameter Params; For the above intermediate data, satisfying ; This represents the system's public key.

[0139] Determine the third ciphertext of the first data source node. ,in ;

[0140] Determine the fourth ciphertext of the first data source node “ "Used to represent the XOR operation;

[0141] Constructing a message dependency chain for a single distributed data source node: , , , ;in, The first test data is the first data source node. This refers to the number of large models that use the full sample data; The fifth hash function in the system parameters; the number of dependencies in the dependency chain depends on how many sets of ciphertext segments need to be provided, i.e. how many sets of full sample data are provided for the large model.

[0142] Based on the message dependency chain, the fifth ciphertext of the first data source node is determined. ,in ;

[0143] Based on the message dependency chain, the sixth ciphertext of the first data source node is determined. ;

[0144] Based on the first encrypted text, second encrypted text, third encrypted text, fourth encrypted text, fifth encrypted text, and sixth encrypted text of the first data source node, and the number of large models, determine the aggregable cryptographic signatures of the first data source node. ;in , For large numbers of models that require the use of distributed data source nodes;

[0145] The aggregateable signatures corresponding to each of the distributed data source nodes (e.g., the signatures corresponding to the first data source node, the second data source node, and the third data source node mentioned above) are used to... Send to the test node; perform equivalent tests on the potential attack content of obfuscated multi-ciphertext on the aggregateable signatures corresponding to the multiple distributed data source nodes respectively, and obtain the detection results.

[0146] It should be noted that during the random selection of sample data (i.e., obtaining the first test data and the second test data described later) between the distributed data source nodes and the test nodes, the test nodes can either predetermine the random selection strategy and / or specific samples and synchronize the selection results to each distributed data source node for execution; or each distributed data source node can independently determine the random selection strategy and / or specific samples and synchronize the relevant information to the test nodes. The specific agreement method described above can be set by the system deployment or maintenance personnel according to actual application needs, and this application does not impose any limitations on it.

[0147] In some implementations, generating encrypted ciphertext based on the detection results and the full sample data from the multiple distributed data source nodes includes the following:

[0148] If the test result is that the test fails, the encrypted ciphertext generation operation will be terminated.

[0149] If the test result is passed, the full sample data of the multiple distributed data source nodes and the aggregateable signatures of the multiple distributed data source nodes are concatenated to obtain the concatenation results corresponding to each of the multiple distributed data source nodes. The concatenation results corresponding to each of the multiple distributed data source nodes are then aggregated to obtain the encrypted ciphertext.

[0150] In this embodiment, when the test fails, the encrypted ciphertext generation operation is terminated directly. This avoids further encryption, splicing, and aggregation of data containing potentially attacking or abnormal content, blocking the channel for insecure sample data to enter the subsequent large model training or inference process, and reducing the risk of malicious or contaminated samples being misused from the source.

[0151] Only when the test result is passed, the full sample data of multiple distributed data source nodes are concatenated and aggregated with the corresponding aggregable signatures. The aggregation results of each node are further concatenated into a unified encrypted ciphertext, so that the subsequent large model uses these sample data only in the form of a whole encrypted ciphertext. This significantly improves the security of sample data access for the large model while ensuring data availability.

[0152] In some implementations, generating the sequential bit aggregation key based on the detection results and the full sample data from the multiple distributed data source nodes includes the following:

[0153] If the test result is "test fails", stop the sequence bit aggregation key generation operation;

[0154] If the test result is passed, a first segment key is generated. If the concatenation results of multiple distributed data source nodes are aggregated to obtain encrypted ciphertext, a second segment key is generated. The first segment key and the second segment key are concatenated to obtain a sequence bit aggregation key.

[0155] In this embodiment, the sequential bit aggregation key generation process is strongly bound to the detection result and the acquisition of encrypted ciphertext. Only when the data is secure and aggregation is completed can a usable sequential bit aggregation key be formed, which significantly reduces the risk of key abuse or being bypassed for direct use, thereby improving the security of large models when using sample data.

[0156] Specifically, taking the sample data acquisition system of the large model as an example, the step of generating encrypted ciphertext based on the detection results and the full sample data of the multiple distributed data source nodes includes:

[0157] Equivalence test trap gates are sent to the multiple distributed data source nodes respectively. ;in, It can be a pre-generated fixed private key known to the tester. This is used to identify different distributed data source nodes; it can also be understood as the tester randomly generating different fixed private keys for different nodes.

[0158] The test nodes respectively obtain the second test data D corresponding to each of the multiple distributed data source nodes, and use the system public key. Encryption is performed to obtain encrypted ciphertext. ; Perform an XOR operation between the private key skj and the encrypted data E. If If it is the expected session key, then the acceptance of this formula is high. If This is not the expected session key. You may need to check the public key, private key, or whether there are any errors in the encryption process.

[0159] Among them, the test node evaluates the acceptance FD ;

[0160] If the acceptance factor (FD) is high at the test node, then each of the n ciphertexts will be evaluated separately. Perform an equivalence test on the content of the obfuscated multi-ciphertext attack.

[0161] The execution of the obfuscated multi-ciphertext potential attack content equivalence test includes:

[0162] The test node check If the condition is met, continue with the following operations; otherwise, terminate the operation and output "". "Prompt message;"

[0163] for Test node calculation Because signature encryption algorithms have Therefore, the test node combines n equations to obtain a system of equations;

[0164]

[0165] And make it possible to hide. ,in The tester obtained a unique solution to the system of equations by inverting the Vandermonde matrix corresponding to the system. ;

[0166] Test node check equation Whether it is valid, among which satisfy:

[0167]

[0168] If the expression is true, the test node will output the test result "test passed" to the key security center and the aggregation node; otherwise, it will output the test result "test failed" to the key security center.

[0169] If the aggregation node receives a "test passed" result from the test node, it means that the sample data owned by the multiple nodes has no potential security risks and can be used to train and optimize large models.

[0170] Further, obtain the full sample data corresponding to each of the multiple distributed data source nodes. The aggregateable cryptographic signatures of the multiple data source nodes, including the first data source node, the second data source node, and the third data source node, etc., are... Each of the multiple distributed data source nodes corresponds to a full set of sample data. The concatenation results corresponding to each of the multiple distributed data source nodes are obtained, and the concatenation results corresponding to each of the multiple distributed data source nodes are aggregated to obtain the encrypted ciphertext; the encrypted ciphertext It can be expressed by the following formula:

[0171] ;

[0172] In this formula, Used to represent encrypted ciphertext; Used to indicate the first Aggregatable secret signatures for distributed data source nodes; Used to indicate the first Full sample data from each distributed data source node; " is used to indicate a splicing character.

[0173] More specifically, the encrypted ciphertext can also be stored in the... Figure 2 On the blockchain.

[0174] Specifically, taking the sample data acquisition system of the large model as an example, the generation of the sequential bit aggregation key based on the detection results and the full sample data of the multiple distributed data source nodes can include the following:

[0175] If the test result is "test fails", stop the sequence bit aggregation key generation operation;

[0176] If the test result is "test passed", the test node sends the test result "test passed" to the key security center. Then, the key security center generates a first segment key SPKA (the first segment key can be generated using the existing Advanced Encryption Standard (AES) key generation algorithm, but is not limited to it; other symmetric key generation algorithms known to those skilled in the art can also be used) and sends it to the test node. If the concatenation results from multiple distributed data source nodes are aggregated to obtain encrypted ciphertext, and the aggregation node sends aggregation execution information, the key security center generates a second segment key SPKB (the second segment key can be generated using the AES algorithm) and sends it to the test node.

[0177] The test node and the aggregation node send the first segment key SPKA and the second segment key SPKB to the distribution node, and synchronize the corresponding encryption public key to the distributed data source node. This key pair can be used for decryption later. The first segment key SPKA and the second segment key SPKB are concatenated to obtain the sequential aggregation key SPKS. SPKS is used as the key for the large model to request training samples from the distributed data source node. SPKS can be controlled by the distribution node.

[0178] More specifically, when a large model needs to use sample data to train and optimize model performance, the large model can make a data request to the distribution node; after the distribution node verifies the universally unique identifier (UUID) carried by the large model with the request, if the UUID is in the preset whitelist, the SPKS is sent to the distributed data source node.

[0179] The distributed data source node receives the SPKS and matches it with the public key. If they match, it sends the encrypted ciphertext to the blockchain, authorizing the distribution node to access it. The authentication confirmation information; after collecting the authentication confirmation information sent by all distributed data source nodes on the blockchain, the confirmation information is specifically confirmed and encrypted messages are sent. To the distribution node; the distribution node can Forward the data to any large model that initiated the data request, so that the large model can use the sample data and improve the security and reliability of the large model's use of the sample data.

[0180] More specifically, the large model can decrypt ciphertext based on a preset decryption algorithm. Perform a decryption operation to obtain the full sample data corresponding to each of the multiple distributed data source nodes, which is used to complete the retrieval of sample data;

[0181] More specifically, you can refer to the following steps:

[0182] calculate ;

[0183] According to the above The value, calculate = , ;

[0184] calculate:

[0185] ,

[0186]

[0187]

[0188] ;

[0189] Check whether the following equations are true simultaneously:

[0190] ,

[0191] ,

[0192] f

[0193] If all the above equations are true, then the decryption is successful, and the large model can use the full sample data normally.

[0194] It should be noted that the data sample acquisition method of the large model described above can be executed by an electronic device, that is, all steps included in the above method are executed by the electronic device, which can be an electronic device such as a server, computer, cloud service node or dedicated hardware device.

[0195] See Figure 3 , Figure 3 This is a schematic diagram of the structure of a sample data acquisition device for a large model provided in an embodiment of this application, as shown below. Figure 3 As shown, the large model sample data acquisition device 300 includes:

[0196] The first processing module 301 is used to determine the public and private keys of each of the multiple distributed data source nodes using a progressive key generation method.

[0197] The acquisition module 302 is used to acquire the first test data corresponding to each of the plurality of distributed data source nodes respectively;

[0198] The second processing module 303 is used to perform T-PAADO based on the first test data corresponding to each of the multiple distributed data source nodes, as well as the public and private keys corresponding to each of the multiple distributed data source nodes, to obtain the detection result;

[0199] The third processing module 304 is used to generate encrypted ciphertext and sequence bit aggregation key based on the detection result and the full sample data corresponding to each of the multiple distributed data source nodes.

[0200] The fourth processing module 305 is used to send the encrypted ciphertext to the large model according to the sequence bit aggregation key, and the large model obtains the full sample data corresponding to each of the multiple distributed data source nodes according to the encrypted ciphertext and the preset decryption algorithm.

[0201] Optionally, the step of determining the public and private keys corresponding to each of the multiple distributed data source nodes using a progressive key generation method includes:

[0202] Based on the node identification information and hash function of the first data source node, determine the intermediate parameters of the first data source node;

[0203] Based on the intermediate parameters and the pre-acquired system master key, the private key of the first data source node is determined, and based on the first preset random parameters of the first data source node and the generator of the pre-acquired cyclic addition group, the public key of the first data source node is determined.

[0204] Based on the node identifier information of the second data source node, the public key of the first data source node and the hash function, the public key of the second data source node is determined, and based on the first preset random parameter of the second data source node, the system master key and the public key of the second data source node, the private key of the second data source node is determined.

[0205] The first data node and the second data source node are any one of the plurality of distributed data source nodes.

[0206] Optionally, the large model sample data acquisition device further includes a fifth processing module 306;

[0207] The fifth processing module 306 is used to determine the private key of the third data source node based on the node identifier information of the third data source node and the system master key;

[0208] The public key of the third data source node is determined based on the system master key and the generator.

[0209] The third data source node is any one of the multiple distributed data source nodes.

[0210] Optionally, the step of performing T-PAADO based on the first test data corresponding to each of the plurality of distributed data source nodes, and the public and private keys corresponding to each of the plurality of distributed data source nodes, to obtain the detection result includes:

[0211] Based on the first test data corresponding to each of the multiple distributed data source nodes, the public key corresponding to each of the multiple distributed data source nodes, the private key corresponding to each of the multiple distributed data source nodes, and preset system parameters, multiple ciphertexts corresponding to each of the multiple distributed data source nodes are determined respectively. The preset system parameters include a preset prime number, a cyclic additive group of the order of the preset prime number, the generator of the cyclic additive group, a preset system public key, and a hash function. The preset system public key is determined by the preset system master key and the generator.

[0212] The multiple ciphertexts corresponding to each of the multiple distributed data source nodes are aggregated to obtain the aggregable ciphertexts corresponding to each of the multiple distributed data source nodes.

[0213] The potential attack content of obfuscated multi-ciphertext was tested on the aggregateable signatures corresponding to the multiple distributed data source nodes respectively, and the detection results were obtained.

[0214] Optionally, generating encrypted ciphertext based on the detection results and the full sample data from the multiple distributed data source nodes includes the following:

[0215] If the test result is that the test fails, the encrypted ciphertext generation operation will be terminated.

[0216] If the test result is passed, the full sample data of the multiple distributed data source nodes and the aggregateable signatures of the multiple distributed data source nodes are concatenated to obtain the concatenation results corresponding to each of the multiple distributed data source nodes. The concatenation results corresponding to each of the multiple distributed data source nodes are then aggregated to obtain the encrypted ciphertext.

[0217] Optionally, generating the sequential bit aggregation key based on the detection result and the full sample data of the multiple distributed data source nodes includes the following:

[0218] If the test result is "test fails", stop the sequence bit aggregation key generation operation;

[0219] If the test result is passed, a first segment key is generated. If the concatenation results of multiple distributed data source nodes are aggregated to obtain encrypted ciphertext, a second segment key is generated. The first segment key and the second segment key are concatenated to obtain a sequence bit aggregation key.

[0220] The large model sample data acquisition device 300 is a series of processes that can realize the above-mentioned large model sample data acquisition method. The technical features are one-to-one and can achieve the same technical effect. To avoid repetition, they will not be described in detail here.

[0221] This application also provides an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the various processes of the above-described large model sample data retrieval method embodiment and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0222] For details, see Figure 4 This application also provides an electronic device, including a bus 401, a transceiver 402, an antenna 403, a bus interface 404, a processor 405, and a memory 406.

[0223] The processor 405 is used to determine the public and private keys of each of the multiple distributed data source nodes using a progressive key generation method.

[0224] The transceiver 402 is used to acquire the first test data corresponding to each of the plurality of distributed data source nodes;

[0225] The processor 405 is further configured to perform a tamper-free attack aggregation detection operation T-PAADO based on the first test data corresponding to each of the plurality of distributed data source nodes, and the public and private keys corresponding to each of the plurality of distributed data source nodes, to obtain the detection result;

[0226] Based on the detection results and the full sample data corresponding to each of the multiple distributed data source nodes, an encrypted ciphertext and a sequence bit aggregation key are generated.

[0227] The encrypted ciphertext is sent to the large model according to the sequence bit aggregation key, and the large model obtains the full sample data corresponding to each of the multiple distributed data source nodes according to the encrypted ciphertext and the preset decryption algorithm.

[0228] Optionally, the step of determining the public and private keys corresponding to each of the multiple distributed data source nodes using a progressive key generation method includes:

[0229] Based on the node identification information and hash function of the first data source node, determine the intermediate parameters of the first data source node;

[0230] Based on the intermediate parameters and the pre-acquired system master key, the private key of the first data source node is determined, and based on the first preset random parameters of the first data source node and the generator of the pre-acquired cyclic addition group, the public key of the first data source node is determined.

[0231] Based on the node identifier information of the second data source node, the public key of the first data source node and the hash function, the public key of the second data source node is determined, and based on the first preset random parameter of the second data source node, the system master key and the public key of the second data source node, the private key of the second data source node is determined.

[0232] The first data node and the second data source node are any one of the plurality of distributed data source nodes.

[0233] Optionally, the processor 405 is further configured to determine the private key of the third data source node based on the node identifier information of the third data source node and the system master key;

[0234] The public key of the third data source node is determined based on the system master key and the generator.

[0235] The third data source node is any one of the multiple distributed data source nodes.

[0236] Optionally, the step of performing T-PAADO based on the first test data corresponding to each of the plurality of distributed data source nodes, and the public and private keys corresponding to each of the plurality of distributed data source nodes, to obtain the detection result includes:

[0237] Based on the first test data corresponding to each of the multiple distributed data source nodes, the public key corresponding to each of the multiple distributed data source nodes, the private key corresponding to each of the multiple distributed data source nodes, and preset system parameters, multiple ciphertexts corresponding to each of the multiple distributed data source nodes are determined respectively. The preset system parameters include a preset prime number, a cyclic additive group of the order of the preset prime number, the generator of the cyclic additive group, a preset system public key, and a hash function. The preset system public key is determined by the preset system master key and the generator.

[0238] The multiple ciphertexts corresponding to each of the multiple distributed data source nodes are aggregated to obtain the aggregable ciphertexts corresponding to each of the multiple distributed data source nodes.

[0239] The potential attack content of obfuscated multi-ciphertext was tested on the aggregateable signatures corresponding to the multiple distributed data source nodes respectively, and the detection results were obtained.

[0240] Optionally, generating encrypted ciphertext based on the detection results and the full sample data from the multiple distributed data source nodes includes the following:

[0241] If the test result is that the test fails, the encrypted ciphertext generation operation will be terminated.

[0242] If the test result is passed, the full sample data of the multiple distributed data source nodes and the aggregateable signatures of the multiple distributed data source nodes are concatenated to obtain the concatenation results corresponding to each of the multiple distributed data source nodes. The concatenation results corresponding to each of the multiple distributed data source nodes are then aggregated to obtain the encrypted ciphertext.

[0243] Optionally, generating the sequential bit aggregation key based on the detection result and the full sample data of the multiple distributed data source nodes includes the following:

[0244] If the test result is "test fails", stop the sequence bit aggregation key generation operation;

[0245] If the test result is passed, a first segment key is generated. If the concatenation results of multiple distributed data source nodes are aggregated to obtain encrypted ciphertext, a second segment key is generated. The first segment key and the second segment key are concatenated to obtain a sequence bit aggregation key.

[0246] exist Figure 4 In this context, a bus architecture (represented by bus 401) is used. Bus 401 can include any number of interconnected buses and bridges, linking various circuits including one or more processors represented by processor 405 and memory represented by memory 406. Bus 401 can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 404 provides an interface between bus 401 and transceiver 402. Transceiver 402 can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 405 is transmitted over a wireless medium via antenna 403, which further receives data and transmits data to processor 405.

[0247] Processor 405 is responsible for managing bus 401 and general processing, and can also provide various functions, including timing, peripheral interface, voltage regulation, power management, and other control functions. Memory 406 can be used to store data used by processor 405 during operation.

[0248] Optionally, the processor 405 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD).

[0249] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the various processes of the above-described large model sample data retrieval method embodiment and achieves the same technical effect. To avoid repetition, it will not be described again here. The computer-readable storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0250] This application also provides a computer program product, including computer instructions. When executed by a processor, these computer instructions implement the various processes of the above-described large model sample data retrieval method embodiment and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0251] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0252] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0253] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for accessing sample data of a large model, characterized in that, The method includes: A progressive key generation method is used to determine the public and private keys for each of the multiple distributed data source nodes. Obtain the first test data corresponding to each of the multiple distributed data source nodes; Based on the first test data corresponding to each of the multiple distributed data source nodes, and the public and private keys corresponding to each of the multiple distributed data source nodes, a tamper-free attack aggregation detection operation T-PAADO is performed to obtain the detection results; Based on the detection results and the full sample data corresponding to each of the multiple distributed data source nodes, an encrypted ciphertext and a sequence bit aggregation key are generated. The encrypted ciphertext is sent to the large model according to the sequence bit aggregation key, and the large model obtains the full sample data corresponding to each of the multiple distributed data source nodes according to the encrypted ciphertext and the preset decryption algorithm.

2. The method of claim 1, wherein, The method of determining the public and private keys corresponding to each of the multiple distributed data source nodes using a progressive key generation approach includes: Based on the node identification information and hash function of the first data source node, determine the intermediate parameters of the first data source node; Based on the intermediate parameters and the pre-acquired system master key, the private key of the first data source node is determined, and based on the first preset random parameters of the first data source node and the generator of the pre-acquired cyclic addition group, the public key of the first data source node is determined. Based on the node identifier information of the second data source node, the public key of the first data source node and the hash function, the public key of the second data source node is determined, and based on the first preset random parameter of the second data source node, the system master key and the public key of the second data source node, the private key of the second data source node is determined. The first data node and the second data source node are any one of the plurality of distributed data source nodes.

3. The method of claim 2, wherein, The method further includes: The private key of the third data source node is determined based on the node identifier information of the third data source node and the system master key; The public key of the third data source node is determined based on the system master key and the generator. The third data source node is any one of the multiple distributed data source nodes.

4. The method according to any of claims 2 or 3, characterized in that, The step involves performing T-PAADO based on the first test data corresponding to each of the multiple distributed data source nodes, as well as the public and private keys corresponding to each of the multiple distributed data source nodes, to obtain the detection results, including: Based on the first test data corresponding to each of the multiple distributed data source nodes, the public key corresponding to each of the multiple distributed data source nodes, the private key corresponding to each of the multiple distributed data source nodes, and preset system parameters, multiple ciphertexts corresponding to each of the multiple distributed data source nodes are determined respectively. The preset system parameters include a preset prime number, a cyclic additive group of the order of the preset prime number, the generator of the cyclic additive group, a preset system public key, and a hash function. The preset system public key is determined by the preset system master key and the generator. The multiple ciphertexts corresponding to each of the multiple distributed data source nodes are aggregated to obtain the aggregable ciphertexts corresponding to each of the multiple distributed data source nodes. The potential attack content of obfuscated multi-ciphertext was tested on the aggregateable signatures corresponding to the multiple distributed data source nodes respectively, and the detection results were obtained.

5. The method of claim 4, wherein, The step of generating encrypted ciphertext based on the detection results and the full sample data from the multiple distributed data source nodes includes the following: If the test result is that the test fails, the encrypted ciphertext generation operation will be terminated. If the test result is passed, the full sample data of the multiple distributed data source nodes and the aggregateable signatures of the multiple distributed data source nodes are concatenated to obtain the concatenation results corresponding to each of the multiple distributed data source nodes. The concatenation results corresponding to each of the multiple distributed data source nodes are then aggregated to obtain the encrypted ciphertext.

6. The method of claim 5, wherein, The step of generating a sequential bit aggregation key based on the detection results and the full sample data from the multiple distributed data source nodes includes the following: If the test result is "test fails", stop the sequence bit aggregation key generation operation; If the test result is passed, a first segment key is generated. If the concatenation results of multiple distributed data source nodes are aggregated to obtain encrypted ciphertext, a second segment key is generated. The first segment key and the second segment key are concatenated to obtain a sequence bit aggregation key. 7.A sample data access device of a large model, characterized by comprising: The device includes: The first processing module is used to determine the public and private keys of each of the multiple distributed data source nodes using a progressive key generation method. The acquisition module is used to acquire the first test data corresponding to each of the multiple distributed data source nodes; The second processing module is used to perform T-PAADO based on the first test data corresponding to each of the multiple distributed data source nodes, as well as the public and private keys corresponding to each of the multiple distributed data source nodes, to obtain the detection results; The third processing module is used to generate encrypted ciphertext and sequence bit aggregation key based on the detection results and the full sample data corresponding to each of the multiple distributed data source nodes; The fourth processing module is used to send the encrypted ciphertext to the large model according to the sequence bit aggregation key, and the large model obtains the full sample data corresponding to each of the multiple distributed data source nodes according to the encrypted ciphertext and the preset decryption algorithm.

8. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes computer instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 6.

Images