Model processing method, device and computer equipment for oblivious transmission rules

Abstract

The present application relates to a model processing method of an oblivious transmission rule, wherein the method comprises: obtaining a mixed model set, the mixed model set comprising a target model and a plurality of randomly generated simulation models; obtaining target position information of the target model in the mixed model set, and transmitting the mixed model set to a model training end; receiving an encrypted model training result set transmitted by the model training end based on the oblivious transmission rule, and obtaining the target model training result based on the target position information and the encrypted model training result set, wherein the model training result set is obtained by training the model in the mixed model set based on the training data set stored in the model training end. Through the present application, the problem that the security of the model owned by the model provider end cannot be guaranteed in the related art is solved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a model processing method, device and computer equipment for inadvertent transmission rules. Background Art

[0002] With the advancement of science and technology and the development of big data technology, the importance of data is increasing, and the development of enterprises is increasingly dependent on big data and advanced big data technology. Various types of enterprises need a large amount of different types of data for them to analyze and process data, so as to provide customers with better services. Some computer-type companies have built more complete models for companies with a large amount of data and advanced machine learning algorithms. However, due to the lack of a large amount of data for machine learning to train models, many companies can only seek cooperation with data providers, hoping to obtain the required model training data from data providers. However, due to various reasons such as legal constraints and privacy protection, data providers cannot directly provide certain data to companies that need these data. For example, pharmaceutical companies need to obtain patients' clinical data from hospitals, but due to legal constraints, hospitals cannot leak patients' clinical data. Therefore, how to maximize the role of data while ensuring that the data provided by the data provider is not leaked will become one of the problems that technicians in this field need to solve urgently.

[0003] In the related art, the model provider directly sends its own model to the model training end, and the model training end trains the received model based on its own data to obtain the model training result, and sends the model training result to the model provider. However, this method will leak the model owned by the model provider, and the security of the model owned by the model provider cannot be guaranteed, which may cause huge economic losses to the model provider.

[0004] Currently, no effective solution has been proposed for the problem in related technologies that the security of the model owned by the model provider cannot be guaranteed. Summary of the invention

[0005] The embodiments of the present application provide a model processing method, apparatus and computer device for inadvertent transmission rules, so as to at least solve the problem in the related art that the security of the model owned by the model provider cannot be guaranteed.

[0006] In a first aspect, an embodiment of the present application provides a first model processing method for inadvertent transmission rules, which is applied to a model provider, including:

[0007] Acquire a hybrid model set, wherein the hybrid model set includes a target model and a plurality of randomly generated simulation models;

[0008] Obtaining target position information of the target model in the hybrid model set, and transmitting the hybrid model set to a model training end;

[0009] Based on the oblivious transmission rule, an encrypted model training result set transmitted by the model training end is received, and a target model training result is obtained based on the target location information and the encrypted model training result set, wherein the model training result set is obtained by training the model in the mixed model set based on the training data set stored in the model training end.

[0010] In some embodiments, obtaining target position information of the target model in the hybrid model set includes:

[0011] The models in the mixed model set are randomly sorted, and the target position information of the target model in the mixed model set after the random sorting is obtained.

[0012] In some embodiments, the receiving, based on the oblivious transmission rule, the encrypted model training result set transmitted by the model training end, and obtaining the target model training result based on the target location information and the encrypted model training result set includes:

[0013] Obtain a preset prime number set, a first random number, and a second random number, and obtain a random parameter according to the first random number and the second random number;

[0014] Sending the random parameter to the model training end, and receiving a first transmission parameter sent by the model training end, wherein the first transmission parameter is obtained by encrypting the model training result set based on the Chinese remainder theorem, the preset prime number set, the random parameter, and the random number set generated by the model training end;

[0015] Obtaining a second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target location information, and sending the second transmission parameter to the model training end;

[0016] receiving a third transmission parameter sent by the model training end, and obtaining the target model training result according to the third transmission parameter, the second random number and the target position information, wherein the third transmission parameter is obtained based on the second transmission parameter and the random number set; or,

[0017] Obtain a preset prime number set, a first random number, and a second random number, and obtain a random parameter according to the first random number and the second random number;

[0018] Sending the random parameter to the model training end, and receiving a first transmission parameter sent by the model training end, where the first transmission parameter is obtained according to the Chinese remainder theorem, the preset prime number set, the random parameter, and the random number set generated by the model training end;

[0019] Obtaining a second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target location information, and sending the second transmission parameter to the model training end;

[0020] Receive the third transmission parameter sent by the model training end, and obtain the target model training result according to the third transmission parameter, the second random number and the target position information, wherein the third transmission parameter is obtained by encrypting the model training result set based on the second transmission parameter and the random number set.

[0021] In a second aspect, an embodiment of the present application provides a second model processing method for inadvertent transmission rules, which is applied to a model training end, including:

[0022] Receiving a hybrid model set transmitted by a model provider, the hybrid model set comprising a target model stored in the model provider and a plurality of randomly generated simulation models;

[0023] Acquire a training data set, and train the models in the hybrid model set based on the training data set to obtain a model training result set corresponding to the hybrid model set;

[0024] The model training result set is encrypted based on the oblivious transmission rule, and the encrypted model training result set is transmitted to the model provider, so that the model provider obtains the target model training result based on the encrypted model training result set and the pre-acquired target position information, and the target position information represents the position information of the target model in the mixed model set.

[0025] In some of the embodiments, the target position information represents position information of the target model in a set of mixed models after random sorting.

[0026] In some embodiments, encrypting the model training result set based on the oblivious transmission rule, and transmitting the encrypted model training result set to the model provider, so that the model provider obtains the target model training result based on the encrypted model training result set and the pre-acquired target location information, includes:

[0027] Get preset prime number sets and random number sets;

[0028] Receive the random parameters sent by the model provider, and encrypt the model training result set based on the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set to obtain a first transmission parameter;

[0029] The first transmission parameter is sent to the model providing end, and a second transmission parameter sent by the model providing end is received; the second transmission parameter is obtained according to the first transmission parameter, the preset prime number set, the target position information stored in the model providing end, and the first random number generated by the model providing end;

[0030] Obtaining a third transmission parameter according to the second transmission parameter and the random number set, and sending the third transmission parameter to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number; or,

[0031] Get preset prime number sets and random number sets;

[0032] Receiving the random parameters sent by the model providing end, and obtaining the first transmission parameters according to the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set generated by the model training end;

[0033] The first transmission parameter is sent to the model providing end, and a second transmission parameter sent by the model providing end is received; the second transmission parameter is obtained according to the first transmission parameter, the preset prime number set, the target position information stored in the model providing end, and the first random number generated by the model providing end;

[0034] The model training result set is encrypted according to the second transmission parameter and the random number set to obtain a third transmission parameter, and the third transmission parameter is sent to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number.

[0035] In a third aspect, an embodiment of the present application provides a first model processing device for inadvertently transmitting rules, which is applied to a model providing end, including:

[0036] A model acquisition module, used to acquire a hybrid model set, wherein the hybrid model set includes a target model and a plurality of randomly generated simulation models;

[0037] A module processing module, used for obtaining target position information of the target model in the mixed model set, and transmitting the mixed model set to a model training end;

[0038] An encrypted transmission module is used to receive the encrypted model training result set transmitted by the model training end based on the oblivious transmission rule, and obtain the target model training result based on the target location information and the encrypted model training result set, wherein the model training result set is obtained by training the model in the hybrid model set based on the training data set stored in the model training end.

[0039] In a fourth aspect, an embodiment of the present application provides a second model processing device for inadvertent transmission rules, which is applied to a model training end, including:

[0040] A model receiving module, used for receiving a mixed model set transmitted by a model providing end, wherein the mixed model set includes a target model stored in the model providing end and a plurality of randomly generated simulation models;

[0041] A model training module, used to obtain a training data set, and train the models in the hybrid model set based on the training data set to obtain a model training result set corresponding to the hybrid model set;

[0042] A data transmission module is used to encrypt the model training result set based on the oblivious transmission rule, and transmit the encrypted model training result set to the model provider, so that the model provider obtains the target model training result based on the encrypted model training result set and the pre-acquired target location information, wherein the target location information represents the location information of the target model in the mixed model set.

[0043] In a fifth aspect, an embodiment of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the model processing method for inadvertent transmission rules as described in the first and second aspects above is implemented.

[0044] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the model processing method for inadvertent transmission rules as described in the first and second aspects above.

[0045] Compared with the related art, the model processing method, device and computer equipment of the inadvertent transmission rule provided in the embodiment of the present application obtain a mixed model set, the mixed model set includes a target model and a plurality of randomly generated simulation models; obtain the target position information of the target model in the mixed model set, and transmit the mixed model set to the model training end; receive the encrypted model training result set transmitted by the model training end based on the inadvertent transmission rule, and obtain the target model training result based on the target position information and the encrypted model training result set, the model training result set is obtained by training the model in the mixed model set based on the training data set stored in the model training end. The present application obtains a mixed model set by mixing the target model owned by the model provider end into a plurality of randomly generated simulation models, and transmits the mixed model set to the model training end, so that the model training end cannot know which model in the mixed model set is the target model owned by the model provider end, thereby ensuring the security of the model owned by the model provider end, and solving the problem that the security of the model owned by the model provider end cannot be guaranteed in the related art.

[0046] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0048] Figure 1 This is a flow chart of a first model processing method for inadvertent transmission rules according to an embodiment of the present application;

[0049] Figure 2 The process of obtaining the target model training results based on the oblivious transmission rule in the embodiment of this application Figure 1 ;

[0050] Figure 3 The process of obtaining the target model training results based on the oblivious transmission rule in the embodiment of this application Figure 2 ;

[0051] Figure 4 This is a flow chart of a second model processing method for inadvertent transmission rules according to an embodiment of the present application;

[0052] Figure 5 The process of encrypting and transmitting the model training result set based on the oblivious transmission rule in the embodiment of the present application Figure 1 ;

[0053] Figure 6 The process of encrypting and transmitting the model training result set based on the oblivious transmission rule in the embodiment of the present application Figure 2 ;

[0054] Figure 7 This is a structural block diagram of a model processing device for a first type of inadvertent transmission rule according to an embodiment of the present application;

[0055] Figure 8 This is a structural block diagram of a model processing device for a second type of inadvertent transmission rule according to an embodiment of the present application;

[0056] Fig. 9 A schematic diagram of the hardware structure of a model processing device for inadvertent transmission rules according to an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0058] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0059] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0060] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a", "the" and the like involved in this application do not indicate a quantitative limitation, and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like involved in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The "multiple" involved in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that there may be three relationships, for example, "A and / or B" can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific ordering of the objects.

[0061] The various technologies described in this application can be applied to the joint training scenarios of various models to solve the problem in related technologies that the security of the model owned by the model provider cannot be guaranteed.

[0062] This application provides a model processing method for inadvertent transmission rules, which is applied to a model provider. Figure 1 Flow chart of the first model processing method of inadvertent transmission rules in the embodiment of the present application, such as Figure 1 As shown, the process includes the following steps:

[0063] Step S110, obtaining a hybrid model set, where the hybrid model set includes a target model and a plurality of randomly generated simulation models.

[0064] The letter H = {h1,...,h n+1} represents a mixed model set, which includes a target model and n randomly generated simulation models.

[0065] It should be noted that the number of simulation models can be set according to the actual application scenario, and the embodiments of the present application do not impose any specific limitations.

[0066] Step S120, obtaining target position information of the target model in the hybrid model set, and transmitting the hybrid model set to the model training end.

[0067] It should be noted that the target location information may be a location number or other location information, which is not limited in this embodiment, as long as the location information of the target model in the mixed model set can be identified. For example, the target location information may be a location number composed of numbers.

[0068] Step S130, receiving the encrypted model training result set transmitted by the model training end based on the oblivious transmission rule, and obtaining the target model training result based on the target location information and the encrypted model training result set, the model training result set is obtained by training the model in the hybrid model set based on the training data set stored in the model training end.

[0069] The letter F = {f1,...,f n+1} represents the model training result set, and the letter q can be used to represent the target location information, so the target model training result can be expressed as f q .

[0070] Obtain the target model training result f based on the target location information q and the encrypted model training result set F q , that is, F,q→f q , where F represents the model training result set, q represents the target location information, and f q Represents the target model training result.

[0071] Among them, the oblivious transfer rules refer to the data transmission rules set based on the oblivious transfer protocol and specific application scenarios. The oblivious transfer protocol is a privacy-protecting two-party communication protocol that enables the communicating parties to transmit messages in a selectively ambiguous manner. The oblivious transfer protocol is a basic protocol in cryptography. It allows the receiver of the service to obtain certain messages input by the sender of the service in an inadvertent manner, so that the privacy of the receiver can be kept unknown by the sender.

[0072] Through the above steps S110 to S130, a hybrid model set is obtained, the hybrid model set includes a target model and a plurality of randomly generated simulation models; the target position information of the target model in the hybrid model set is obtained, and the hybrid model set is transmitted to the model training end; the encrypted model training result set transmitted by the model training end is received based on the oblivious transmission rule, and the target model training result is obtained based on the target position information and the encrypted model training result set, and the model training result set is obtained by training the models in the hybrid model set based on the training data set stored in the model training end. In this implementation, the target model owned by the model provider end is mixed into a plurality of randomly generated simulation models to obtain a hybrid model set, and the hybrid model set is transmitted to the model training end, so that the model training end cannot know which model in the hybrid model set is the target model owned by the model provider end, thereby ensuring the security of the model owned by the model provider end, and solving the problem that the security of the model owned by the model provider end cannot be guaranteed in the related art.

[0073] In addition, the model provider and the model training end interact through an oblivious transmission system. The model provider will not transmit the stored target location information to the model training end, and the model training end will not directly transmit the acquired model training results to the model provider. This can avoid the leakage of model training results and target location information during data transmission, as well as the problem of the model provider exploring the training data set of the model training end through certain statistical laws, thereby improving the security of data transmission.

[0074] In some embodiments, the step S120 further includes: performing random sorting processing on the models in the mixed model set, and obtaining target position information of the target model in the mixed model set after the random sorting processing.

[0075] This embodiment can further improve the confidentiality of the position information of the target model in the mixed model set by randomly sorting the models in the mixed model set and obtaining the target position information of the mixed model set after the random sorting, thereby further improving the security of the data.

[0076] In some of these embodiments, Figure 2 The process of obtaining the target model training results based on the oblivious transmission rule in the embodiment of this application Figure 1 ,like Figure 2 As shown, the process includes the following steps:

[0077] Step S210, obtaining a preset prime number set, a first random number and a second random number, and obtaining a random parameter according to the first random number and the second random number.

[0078] The preset prime number set represents the set determined by the model provider and the model training end through negotiation, which can be represented by the set N = {n1, n2, ..., n k+1 The first random number and the second random number represent two random numbers generated by the model provider. The first random number can be represented by letter t, the second random number can be represented by letter l, and the random parameter can be represented by letter s.

[0079] Step S220, sending the random parameters to the model training end, and receiving the first transmission parameters sent by the model training end, the first transmission parameters are obtained by encrypting the model training result set based on the Chinese remainder theorem, a preset prime number set, random parameters and a random number set generated by the model training end.

[0080] The Chinese Remainder Theorem is also known as the Sun Tzu Theorem. The Sun Tzu Theorem is a method used in ancient China to solve a linear congruence system (see congruence), also known as the Chinese Remainder Theorem.

[0081] The random number set includes multiple random numbers, which can be represented by the set R = {r1, r2, ..., r k+1} to represent the random number set. The letter x can be used to represent the first transmission parameter, and the first transmission parameter x can be expressed as:

[0082] F,s,R,N→x Formula (A-1)

[0083] Among them, x represents the first transmission parameter, s represents the random parameter, R represents the random number set, N represents the preset prime number set, and F represents the model training result set.

[0084] Step S230, obtaining a second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target position information, and sending the second transmission parameter to the model training end.

[0085] The letter q can be used to represent the target position information, and the letter m can be used to represent the second transmission parameter. The second transmission parameter m can be expressed as:

[0086] x,q,t,N→m Formula (A-2)

[0087] Among them, x represents the first transmission parameter, q represents the target position information, t represents the first random number, N represents the preset prime number set, and m represents the second transmission parameter.

[0088] Step S240, receiving the third transmission parameter sent by the model training end, and obtaining the target model training result according to the third transmission parameter, the second random number and the target position information, wherein the third transmission parameter is obtained based on the second transmission parameter and the random number set.

[0089] The third transmission parameter can be represented by M', and the third transmission parameter is obtained according to the second transmission parameter m and the random number set R, so the third transmission parameter M'={m'1,m'2,...,m' k+1} can be expressed as:

[0090] m,R→M' Formula (A-3)

[0091] Wherein, m represents the second transmission parameter, R represents a random number set, and M' represents the third transmission parameter.

[0092] It should be noted that there are multiple third transmission parameters M', and their number is consistent with the number of random numbers in the random number set R. Therefore, M' may also be referred to as a third transmission parameter set. This embodiment does not limit the specific name of M'.

[0093] According to the third transmission parameter M' and the second random number l, the target model training result f corresponding to the target position information q is obtained. q , then the target model training result f q It can be expressed as:

[0094] M',q,l→f q Formula (A-4)

[0095] Wherein, M' represents the third transmission parameter, q represents the target location information, l represents the second random number, and f q Represents the target model training result.

[0096] It should be noted that the first random number and the second random number can be interchanged, that is, the second transmission parameter can be obtained based on the first random number, the first transmission parameter, the preset prime number set and the pre-stored target position information, and correspondingly, the target model training result corresponding to the target position information can be obtained based on the third transmission parameter, the second random number and the target position information; the second transmission parameter can also be obtained based on the second random number, the first transmission parameter, the preset prime number set and the pre-stored target position information, and correspondingly, the target model training result corresponding to the target position information can be obtained based on the third transmission parameter, the first random number and the target position information, and this is not limited to the present embodiment.

[0097] Through the above steps S210 to S240, the model training result set stored in the model training end is encrypted based on the Chinese remainder theorem, and multiple transmission parameters are transmitted back and forth between the model training end and the model providing end based on random numbers, without transmitting the real model training result set and the target location information, thereby improving the security of data transmission. At the same time, by encrypting the model training result set based on the Chinese remainder theorem to obtain the first transmission parameter, the first transmission parameter is used to replace the model training result set with a large amount of data, which greatly reduces the amount of data transmitted, avoids the need to transmit all the data every time during the data transmission process, and makes it impossible to ensure data security and transmission efficiency, and further improves the data transmission efficiency and data transmission security.

[0098] In some embodiments, the first random number and the second random number are added to obtain a random parameter, that is, the random parameter is s=t+l.

[0099] In some of the embodiments, a target prime number corresponding to the target position information is extracted from a preset prime number set; a modular operation is performed on the target prime number and the first transmission parameter with the target prime number as a divisor and the first transmission parameter as a dividend to obtain a modular operation result; and the modular operation result is subtracted from the first random number to obtain a second transmission parameter.

[0100] Specifically, the target prime number corresponding to the target position information q is extracted from the preset prime number set N, that is, the target prime number is n q . With the target prime number n q is the divisor, the first transmission parameter x is the dividend, and the target prime number n is q The modular operation is performed on the first transmission parameter x to obtain the modular operation result, that is, the modular operation result is x(modn q ). The result of the modulo operation x(modn q ) is subtracted from the first random number t to obtain the second transmission parameter m, that is, the second transmission parameter m is:

[0101] m=x(modn q )-t Formula (A-5)

[0102] Wherein, m represents the second transmission parameter, x represents the first transmission parameter, and n q represents the target prime number, and t represents the first random number.

[0103] In some embodiments, there are multiple third transmission parameters, and the target third transmission parameter corresponding to the target position information is obtained from the multiple third transmission parameters, and the target third transmission parameter is subtracted from the second random number to obtain the target model training result.

[0104] Specifically, the target third transmission parameter m' corresponding to the target position information is obtained from the plurality of third transmission parameters. q , and the target third transmission parameter m' q Subtract it from the second random number l to get the target model training result f q , that is, the target model training result f q for:

[0105] f q =m' q -l Formula (A-6)

[0106] Among them, f q Represents the target model training result, m' q represents the target third transmission parameter, and l represents the second random number.

[0107] In some of these embodiments, Figure 3 The process of obtaining the target model training results based on the oblivious transmission rule in the embodiment of this application Figure 2 ,like Figure 3 As shown, the process includes the following steps:

[0108] Step S310, obtaining a preset prime number set, a first random number and a second random number, and obtaining a random parameter according to the first random number and the second random number.

[0109] The preset prime number set represents the set determined by the model provider and the model training end through negotiation, which can be represented by the set N = {n1, n2, ..., n k+1 The first random number and the second random number represent two random numbers generated by the model provider. The first random number can be represented by letter t, the second random number can be represented by letter l, and the random parameter can be represented by letter s.

[0110] Step S320, sending the random parameters to the model training end, and receiving the first transmission parameters sent by the model training end, the first transmission parameters are obtained according to the Chinese remainder theorem, a preset prime number set, random parameters and a random number set generated by the model training end.

[0111] The random number set includes multiple random numbers, which can be represented by the set R = {r1, r2, ..., r k+1} to represent the random number set. The letter x can be used to represent the first transmission parameter, and the first transmission parameter x can be expressed as:

[0112] s,R,N→x Formula (B-1)

[0113] Among them, x represents the first transmission parameter, s represents the random parameter, R represents the random number set, and N represents the preset prime number set.

[0114] Step S330, obtaining a second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target position information, and sending the second transmission parameter to the model training end.

[0115] The letter q can be used to represent the target position information, and the letter m can be used to represent the second transmission parameter. The second transmission parameter m can be expressed as:

[0116] x,q,t,N→m Formula (B-2)

[0117] Among them, x represents the first transmission parameter, q represents the target position information, t represents the first random number, N represents the preset prime number set, and m represents the second transmission parameter.

[0118] Step S340, receiving the third transmission parameter sent by the model training end, and obtaining the target model training result according to the third transmission parameter, the second random number and the target position information, wherein the third transmission parameter is obtained by encrypting the model training result set based on the second transmission parameter and the random number set.

[0119] The third transmission parameter can be represented by M'. According to the third transmission parameter M' and the second random number l, the target model training result f corresponding to the target position information q is obtained. q , then the target model training result f q It can be expressed as:

[0120] M',q,l→f q Formula (B-3)

[0121] Wherein, M' represents the third transmission parameter, q represents the target location information, l represents the second random number, and f q Represents the target model training result.

[0122] It should be noted that the first random number and the second random number can be interchanged, that is, the second transmission parameter can be obtained based on the first random number, the first transmission parameter, the preset prime number set and the pre-stored target position information, and correspondingly, the target model training result corresponding to the target position information can be obtained based on the third transmission parameter, the second random number and the target position information; the second transmission parameter can also be obtained based on the second random number, the first transmission parameter, the preset prime number set and the pre-stored target position information, and correspondingly, the target model training result corresponding to the target position information can be obtained based on the third transmission parameter, the first random number and the target position information.

[0123] Through the above steps S310 to S340, multiple transmission parameters are transmitted back and forth between the model training end and the model providing end based on random numbers, and the model training result set is encrypted and transmitted according to the transmission parameters and random numbers, without transmitting the real model training result set and the target location information, thereby improving the security of data transmission. At the same time, based on the characteristics of the Chinese remainder theorem, this application integrates a large number of random parameters generated by the model training end and the model providing end into the first transmission parameter, and realizes the use of the first transmission parameter to replace a large number of random parameters for transmission between the model training end and the model providing end, greatly reducing the amount of data transmitted, avoiding the need to transmit all data every time during the data transmission process, making it impossible to guarantee data security and transmission efficiency, and further improving data transmission efficiency and data transmission security.

[0124] In some embodiments, the first random number and the second random number are added to obtain a random parameter, that is, the random parameter is s=t+l.

[0125] In some embodiments, a target prime number corresponding to the target position information is extracted from a preset prime number set. The target prime number is used as a divisor and the first transmission parameter is used as a dividend, and a modular operation is performed on the target prime number and the first transmission parameter to obtain a modular operation result. The modular operation result is subtracted from the first random number to obtain a second transmission parameter.

[0126] Specifically, the target prime number corresponding to the target position information q is extracted from the preset prime number set N, that is, the target prime number is n q . With the target prime number n q is the divisor, the first transmission parameter x is the dividend, and the target prime number n is q The modular operation is performed on the first transmission parameter x to obtain the modular operation result, that is, the modular operation result is x(modn q ). The result of the modulo operation x(modn q ) is subtracted from the first random number t to obtain the second transmission parameter m, that is, the second transmission parameter m is:

[0127] m=x(modn q )-t Formula (B-4)

[0128] Wherein, m represents the second transmission parameter, x represents the first transmission parameter, and n q represents the target prime number, and t represents the first random number.

[0129] In some of the embodiments, there are multiple third transmission parameters; the target third transmission parameter corresponding to the target position information is obtained from the multiple third transmission parameters; the target third transmission parameter is subtracted from the second random number to obtain the target model training result.

[0130] Specifically, the target third transmission parameter m' corresponding to the target position information q is obtained from the plurality of third transmission parameters. q . Set the target third transmission parameter m' q Subtract it from the second random number l to get the target model training result f q , that is, the target model training result f q for:

[0131] f q =m' q -l Formula (B-5)

[0132] Among them, f q Represents the target model training result, m' q represents the target third transmission parameter, and l represents the second random number.

[0133] The following describes and illustrates the first data transmission method based on the oblivious transmission rule in the embodiments of the present application through three specific embodiments.

[0134] In specific embodiment 1, a mixed model set is obtained, the mixed model set includes a target model and multiple randomly generated simulation models; the models in the mixed model set are randomly sorted, the target position information of the target model in the mixed model set after the random sorting is obtained, and the mixed model set is transmitted to the model training end; based on the oblivious transmission rule, the encrypted model training result set transmitted by the model training end is received, and the target model training result is obtained based on the target position information and the encrypted model training result set, and the model training result set is obtained by training the models in the mixed model set based on the training data set stored in the model training end.

[0135] In specific embodiment 2, a mixed model set is obtained, the mixed model set includes a target model and multiple randomly generated simulation models; the models in the mixed model set are randomly sorted, the target position information of the target model in the mixed model set after the random sorting is obtained, and the mixed model set is transmitted to the model training end.

[0136] A preset prime number set, a first random number and a second random number are obtained, and random parameters are obtained according to the first random number and the second random number; the random parameters are sent to the model training end, and a first transmission parameter sent by the model training end is received, the first transmission parameter is obtained by encrypting a model training result set based on the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set generated by the model training end; a second transmission parameter is obtained according to the first random number, the first transmission parameter, the preset prime number set and the target position information, and the second transmission parameter is sent to the model training end; a third transmission parameter is received from the model training end, and a target model training result is obtained according to the third transmission parameter, the second random number and the target position information, the third transmission parameter is obtained based on the second transmission parameter and the random number set.

[0137] In specific embodiment 3, a mixed model set is obtained, the mixed model set includes a target model and multiple randomly generated simulation models; the models in the mixed model set are randomly sorted, the target position information of the target model in the mixed model set after the random sorting is obtained, and the mixed model set is transmitted to the model training end.

[0138] Obtain a preset prime number set, a first random number and a second random number, and obtain random parameters based on the first random number and the second random number; send the random parameters to the model training end, and receive the first transmission parameter sent by the model training end, the first transmission parameter is obtained according to the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set generated by the model training end; obtain the second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target position information, and send the second transmission parameter to the model training end; receive the third transmission parameter sent by the model training end, and obtain the target model training result according to the third transmission parameter, the second random number and the target position information, the third transmission parameter is obtained by encrypting the model training result set based on the second transmission parameter and the random number set.

[0139] Figure 4 Flow chart of the second model processing method of inadvertent transmission rule according to the embodiment of the present application, as shown in FIG. Figure 4 As shown, the process includes the following steps:

[0140] Step S410: receiving a hybrid model set transmitted by a model provider, where the hybrid model set includes a target model stored in the model provider and a plurality of randomly generated simulation models.

[0141] Step S420, obtaining a training data set, and training the models in the hybrid model set based on the training data set to obtain a model training result set corresponding to the hybrid model set.

[0142] The letter Y can be used to represent the training data set, and the model training result set corresponding to the mixed model set H can be expressed as: Y, H→F, where Y represents the training data set, H represents the mixed model set, and F represents the model training result set.

[0143] Step S430, encrypt the model training result set based on the oblivious transmission rule, and transmit the encrypted model training result set to the model provider, so that the model provider obtains the target model training result based on the encrypted model training result set and the pre-acquired target location information, and the target location information represents the location information of the target model in the mixed model set.

[0144] This embodiment obtains a mixed model set by mixing the target model owned by the model provider into multiple randomly generated simulation models, and transmits the mixed model set to the model training end, so that the model training end cannot know which model in the mixed model set is the target model owned by the model provider, thereby ensuring the security of the model owned by the model provider and solving the problem in related technologies that the security of the model owned by the model provider cannot be guaranteed.

[0145] In addition, the model provider and the model training end interact through an oblivious transmission system. The model provider will not transmit the stored target location information to the model training end, and the model training end will not directly transmit the acquired model training results to the model provider. This can avoid the leakage of model training results and target location information during data transmission, and can also avoid the problem of the model provider exploring the training data set of the model training end through certain statistical laws, thereby improving the security of data transmission.

[0146] In some of the embodiments, the target position information represents the position information of the target model in the mixed model set after random sorting.

[0147] This embodiment can further improve the confidentiality of the position information of the target model in the mixed model set by randomly sorting the models in the mixed model set and obtaining the target position information of the mixed model set after the random sorting, thereby further improving the security of the data.

[0148] In some of these embodiments, Figure 5 The process of encrypting and transmitting the model training result set based on the oblivious transmission rule in the embodiment of the present application Figure 1 ,like Figure 5 As shown, the process includes the following steps:

[0149] Step S510, obtaining a preset prime number set and a random number set.

[0150] Step S520, receiving the random parameters sent by the model provider, and encrypting the model training result set based on the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set to obtain the first transmission parameter.

[0151] Step S530, sending the first transmission parameter to the model provider, and receiving the second transmission parameter sent by the model provider; the second transmission parameter is obtained according to the first transmission parameter, a preset prime number set, the target position information stored in the model provider, and the first random number generated by the model provider.

[0152] Step S540, obtain a third transmission parameter based on the second transmission parameter and the random number set, and send the third transmission parameter to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number.

[0153] Through the above steps S510 to S540, the model training result set stored in the model training end is encrypted based on the Chinese remainder theorem, and multiple transmission parameters are transmitted back and forth between the model training end and the model providing end based on random numbers, without transmitting the real model training result set and the target location information, thereby improving the security of data transmission. At the same time, by encrypting the model training result set based on the Chinese remainder theorem to obtain the first transmission parameter, the first transmission parameter is used to replace the model training result set with a large amount of data, which greatly reduces the amount of data transmitted, avoids the need to transmit all the data every time during the data transmission process, and makes it impossible to ensure data security and transmission efficiency, and further improves the data transmission efficiency and data transmission security.

[0154] In some of the embodiments, step S410, a sum operation is performed on the model training result set, the random parameter and the random number set to obtain a sum result. Step S420, all prime numbers in the preset prime number set are multiplied to obtain a first formula parameter. Step S430, the first formula parameter is divided by each random number in the random number set to obtain a second formula parameter. Step S440, a third formula parameter is obtained according to the second formula parameter and the preset prime number set, and the product of the sum result, the second formula parameter and the third formula parameter is obtained. Step S450, with the product as the dividend and the first formula parameter as the divisor, a modulo operation is performed on the product and the first formula parameter to obtain a first transmission parameter.

[0155] Specifically, for any model training result f in the model training result set F i , random parameter s and each random number r in the random number set R i Perform a sum operation to obtain the sum result, that is, the sum result is f i +s+ri The first formula parameter can be represented by the letter M. By multiplying all the prime numbers in the preset prime number set, the first formula parameter M can be obtained, that is,

[0156] You can use the letter M i The first formula parameter M is respectively combined with each prime number r in the preset prime number set R. i Divide and get the second formula parameter M i ,Right now Among them, M i Represents the second formula parameter, r i Represents any random number in the random number set R, M i Represents the second formula parameter.

[0157] You can use the letter t i To represent the third formula parameter. According to the second formula parameter and each prime number n in the preset prime number set N i , we get the third formula parameter t i ,Right now

[0158] Get the sum result f i +s+r i 、The second formula parameter M i and the third formula parameter t i The product of i +s+r i )t i M i . Take the product (f i +s+r i )t i M i The first formula parameter M is used as the divisor, and a modulo operation is performed on the product and the first formula parameter to obtain the first transmission parameter. Therefore, the first transmission parameter can be expressed as:

[0159]

[0160] Where x represents the first transmission parameter, f i represents any model training result in the model training result set F, s represents a random parameter, r i represents any random number in the random number set R, t i Represents the third formula parameter, M i represents the second formula parameter, and M represents the first formula parameter.

[0161] In some embodiments, the second transmission parameter m is combined with each random number r in the random number set R. iSubtract, and get the third transmission parameter M'=mr i (i=1,...,k+1), that is, the third transmission parameter is M'={m'1=m-r1,m'2=m-r2,...,m' k =mr k+1}.

[0162] In some of these embodiments, Figure 6 The process of encrypting and transmitting the model training result set based on the oblivious transmission rule in the embodiment of the present application Figure 2 ,like Figure 6 As shown, the process includes the following steps:

[0163] Step S610, obtaining a preset prime number set and a random number set.

[0164] Step S620, receiving the random parameters sent by the model providing end, and obtaining the first transmission parameters according to the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set generated by the model training end.

[0165] Step S630, sending the first transmission parameter to the model provider, and receiving the second transmission parameter sent by the model provider; the second transmission parameter is obtained according to the first transmission parameter, a preset prime number set, the target position information stored in the model provider, and the first random number generated by the model provider.

[0166] Step S640, encrypt the model training result set according to the second transmission parameter and the random number set to obtain a third transmission parameter, and send the third transmission parameter to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number.

[0167] The third transmission parameter is obtained according to the second transmission parameter m and the random number set R, so the third transmission parameter M'={m'1, m'2, ..., m' k} can be expressed as:

[0168] m,R,F→M' Formula (B-6)

[0169] Among them, m represents the second transmission parameter, R represents the random number set, F represents the model training result set, and M' represents the third transmission parameter.

[0170] It should be noted that there are multiple third transmission parameters M', and their number is consistent with the number of random numbers in the random number set R and the number of model training results in the model training result set F. Therefore, M' can also be called the third transmission parameter set. This embodiment does not limit the specific name of M'.

[0171] Through the above steps S610 to S640, multiple transmission parameters are transmitted back and forth between the model training end and the model providing end based on random numbers, and the model training result set is encrypted and transmitted according to the transmission parameters and random numbers, without transmitting the real model training result set and the target location information, thereby improving the security of data transmission. At the same time, based on the characteristics of the Chinese remainder theorem, this application integrates a large number of random parameters generated by the model training end and the model providing end into the first transmission parameter, and realizes the use of the first transmission parameter to replace a large number of random parameters for transmission between the model training end and the model providing end, greatly reducing the amount of data transmitted, avoiding the need to transmit all data every time during the data transmission process, making it impossible to guarantee data security and transmission efficiency, and further improving data transmission efficiency and data transmission security.

[0172] In some embodiments, a summation operation is performed on the random parameter and the random number set to obtain a summation result. All prime numbers in the preset prime number set are multiplied to obtain a first formula parameter. The first formula parameter is divided by each random number in the random number set to obtain a second formula parameter. According to the second formula parameter and the preset prime number set, a third formula parameter is obtained, and the product of the summation result, the second formula parameter and the third formula parameter is obtained. With the product as the dividend and the first formula parameter as the divisor, a modular operation is performed on the product and the first formula parameter to obtain a first transmission parameter.

[0173] Specifically, for the random parameter s and each random number r in the random number set R i Perform a sum operation to obtain the sum result, that is, the sum result is s+r i The first formula parameter can be represented by the letter M. By multiplying all the prime numbers in the preset prime number set N, the first formula parameter M can be obtained, that is,

[0174] You can use the letter M i To represent the second formula parameter. The first formula parameter M is respectively combined with each random number r in the random number set R i Divide and get the second formula parameter M i ,Right now Among them, M i Represents the second formula parameter, r i Represents any random number in the random number set R, M i Represents the second formula parameter.

[0175] You can use the letter t i To represent the third formula parameter. According to the second formula parameter M i And every prime number n in the preset prime number set N i , we get the third formula parameter t i ,Right now Get the sum result s+r i 、The second formula parameter M i and the third formula parameter t i The product of (s+r i )t i M i .

[0176] The product (s+r i )t i M i The first formula parameter M is used as the divisor, and a modulo operation is performed on the product and the first formula parameter to obtain the first transmission parameter x. Therefore, the first transmission parameter x can be expressed as:

[0177]

[0178] Where x represents the first transmission parameter, s represents the random parameter, and r i represents any random number in the random number set R, t i Represents the third formula parameter, M i represents the second formula parameter, and M represents the first formula parameter.

[0179] In some of the embodiments, the second transmission parameter is subtracted from each random number in the random number set to obtain an encryption parameter set; the model training result set is encrypted according to the encryption parameter set to obtain a third transmission parameter.

[0180] Specifically, the second transmission parameter m is combined with each random number r in the random number set R i Subtract them and get the encryption parameter set L = {m-r1,m-r2,...,mr k+1}. According to each encryption parameter mr in the encryption parameter set L i The model training results f in the model training result set F that have the same position information as the encryption parameters are respectively i Encryption is performed to obtain a third transmission parameter, that is, the third transmission parameter M' can be expressed as:

[0181] mr i ,f i →M' Formula (B-8)

[0182] Wherein, m represents the second transmission parameter, r i represents any random number in the random number set R, f i represents any model training result in the model training result set F, and M' represents the third transmission parameter.

[0183] The following describes and illustrates the first data transmission method based on the oblivious transmission rule in the embodiment of the present application through a specific embodiment.

[0184] In specific embodiment 4, a mixed model set transmitted by a model provider is received, the mixed model set including a target model stored in the model provider and a plurality of randomly generated simulation models; a training data set is obtained, and the models in the mixed model set are trained based on the training data set to obtain a model training result set corresponding to the mixed model set; the model training result set is encrypted based on an oblivious transmission rule, and the encrypted model training result set is transmitted to the model provider, so that the model provider obtains a target model training result based on the encrypted model training result set and pre-acquired target position information, wherein the target position information represents the position information of the target model in the mixed model set after random sorting.

[0185] It should be noted that the steps shown in the above process or the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0186] This embodiment also provides a model processing device for inadvertent transmission rules, which is used to implement the above embodiments and preferred implementations, and will not be repeated here. As used below, the terms "module", "unit", "subunit", etc. can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0187] Figure 7 This is a structural block diagram of a model processing device for a first type of inadvertent transmission rule according to an embodiment of the present application, such as Figure 7 As shown, the device comprises:

[0188] The model acquisition module 710 is used to acquire a hybrid model set, where the hybrid model set includes a target model and a plurality of randomly generated simulation models.

[0189] The module processing module 720 is used to obtain the target position information of the target model in the mixed model set and transmit the mixed model set to the model training end.

[0190] The encrypted transmission module 730 is used to receive the encrypted model training result set transmitted by the model training end based on the oblivious transmission rule, and obtain the target model training result based on the target location information and the encrypted model training result set. The model training result set is obtained by training the model in the mixed model set based on the training data set stored in the model training end.

[0191] In some of the embodiments, the module processing module 720 is further used to perform random sorting processing on the models in the mixed model set, and obtain target position information of the target model in the mixed model set after the random sorting processing.

[0192] In some embodiments, the encryption transmission module 730 includes a first data acquisition unit, a first parameter transmission unit, a first parameter generation unit, and a first target acquisition unit, wherein:

[0193] The first data acquisition unit is used to acquire a preset prime number set, a first random number and a second random number, and obtain a random parameter according to the first random number and the second random number.

[0194] The first parameter transmission unit is used to send random parameters to the model training end and receive the first transmission parameters sent by the model training end. The first transmission parameters are obtained by encrypting the model training result set based on the Chinese remainder theorem, a preset prime number set, random parameters and a random number set generated by the model training end.

[0195] The first parameter generating unit is used to obtain the second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target position information, and send the second transmission parameter to the model training end.

[0196] The first target acquisition unit is used to receive the third transmission parameter sent by the model training end, and obtain the target model training result according to the third transmission parameter, the second random number and the target position information, and the third transmission parameter is obtained based on the second transmission parameter and the random number set.

[0197] In some of the embodiments, the encryption transmission module 730 further includes a second data acquisition unit, a second parameter transmission unit, a second parameter generation unit, and a second target acquisition unit, wherein:

[0198] The second data acquisition unit is used to acquire a preset prime number set, a first random number and a second random number, and obtain a random parameter according to the first random number and the second random number.

[0199] The second parameter transmission unit is used to send the random parameters to the model training end and receive the first transmission parameters sent by the model training end, where the first transmission parameters are obtained according to the Chinese remainder theorem, a preset prime number set, random parameters and a random number set generated by the model training end.

[0200] The second parameter generating unit is used to obtain the second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target position information, and send the second transmission parameter to the model training end.

[0201] The second target acquisition unit is used to receive the third transmission parameter sent by the model training end, and obtain the target model training result according to the third transmission parameter, the second random number and the target position information. The third transmission parameter is obtained by encrypting the model training result set based on the second transmission parameter and the random number set.

[0202] Figure 8 This is a structural block diagram of a model processing device for a second inadvertent transmission rule according to an embodiment of the present application. Figure 8 As shown, the device comprises:

[0203] The model receiving module 810 is used to receive a mixed model set transmitted by a model providing end, where the mixed model set includes a target model stored in the model providing end and a plurality of randomly generated simulation models.

[0204] The model training module 820 is used to obtain a training data set, and train the models in the hybrid model set based on the training data set to obtain a model training result set corresponding to the hybrid model set.

[0205] The data transmission module 830 is used to encrypt the model training result set based on the oblivious transmission rule, and transmit the encrypted model training result set to the model provider, so that the model provider obtains the target model training result based on the encrypted model training result set and the pre-acquired target location information, and the target location information represents the location information of the target model in the mixed model set.

[0206] In some of the embodiments, the target position information represents the position information of the target model in the mixed model set after random sorting.

[0207] In some embodiments, the data transmission module 830 includes a first data set acquisition unit, a first data encryption unit, a first data transmission unit, and a second data transmission unit, wherein:

[0208] The first number set acquisition unit is used to acquire a preset prime number set and a random number set.

[0209] The first data encryption unit is used to receive the random parameters sent by the model provider, and encrypt the model training result set based on the Chinese remainder theorem, a preset prime number set, random parameters and a random number set to obtain a first transmission parameter.

[0210] The first data transmission unit is used to send the first transmission parameter to the model provider and receive the second transmission parameter sent by the model provider; the second transmission parameter is obtained according to the first transmission parameter, a preset prime number set, target position information stored in the model provider and a first random number generated by the model provider.

[0211] The second data transmission unit is used to obtain a third transmission parameter based on the second transmission parameter and the random number set, and send the third transmission parameter to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target position information and the pre-generated second random number.

[0212] In some embodiments, the data transmission module 830 further includes a second data set acquisition unit, a transmission parameter generation unit, a transmission parameter transceiver unit, and a second data encryption unit, wherein:

[0213] The second number set acquisition unit is used to acquire a preset prime number set and a random number set.

[0214] The transmission parameter generation unit is used to receive the random parameters sent by the model providing end, and obtain the first transmission parameter according to the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set generated by the model training end.

[0215] A transmission parameter transceiver unit is used to send a first transmission parameter to a model provider and receive a second transmission parameter sent by the model provider; the second transmission parameter is obtained based on the first transmission parameter, a preset prime number set, target position information stored in the model provider and a first random number generated by the model provider.

[0216] The second data encryption unit is used to encrypt the model training result set according to the second transmission parameter and the random number set to obtain a third transmission parameter, and send the third transmission parameter to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number.

[0217] It should be noted that the above modules can be functional modules or program modules, and can be implemented by software or hardware. For modules implemented by hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0218] In addition, combined Figure 1 The model processing method for inadvertent transmission rules described in the embodiment of the present application can be implemented by a model processing device for inadvertent transmission rules. Fig. 9 A schematic diagram of the hardware structure of a model processing device for inadvertent transmission rules according to an embodiment of the present application.

[0219] The model processing device for oblivious transfer rules may include a processor 91 and a memory 92 storing computer program instructions.

[0220] Specifically, the processor 91 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0221] Among them, the memory 92 may include a large-capacity memory for data or instructions. By way of example and not limitation, the memory 92 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 92 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 92 may be inside or outside a data processing device. In a specific embodiment, the memory 92 is a non-volatile memory. In a specific embodiment, the memory 92 includes a read-only memory (ROM) and a random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (Programmable Read-Only Memory, PROM for short), an erasable PROM (Erasable ProgrammableRead-Only Memory, EPROM for short), an electrically erasable PROM (Electrically Erasable ProgrammableRead-Only Memory, EEPROM for short), an electrically alterable ROM (Electrically Alterable Read-Only Memory, EAROM for short) or a flash memory (FLASH) or a combination of two or more of these. Under appropriate circumstances, the RAM can be a static random access memory (SRAM) or a dynamic random access memory (DRAM), wherein the DRAM can be a fast page mode dynamic random access memory (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0222] The memory 92 may be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 91 .

[0223] The processor 91 reads and executes computer program instructions stored in the memory 92 to implement any one of the model processing methods for inadvertent transmission rules in the above embodiments.

[0224] In some of the embodiments, the model processing device for oblivious transmission rules may further include a communication interface 93 and a bus 90. Fig. 9 As shown, the processor 91, the memory 92, and the communication interface 93 are connected via a bus 90 and communicate with each other.

[0225] The communication interface 93 is used to implement communication between the modules, devices, units and / or equipment in the embodiment of the present application. The communication interface 93 can also implement data communication with other components such as: external devices, image / data acquisition equipment, databases, external storage, and image / data processing workstations.

[0226] The bus 90 includes hardware, software or both, and couples the components of the model processing device of the oblivious transmission rule to each other. The bus 90 includes but is not limited to at least one of the following: a data bus, an address bus, a control bus, an expansion bus, and a local bus. By way of example and not limitation, bus 90 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, bus 90 may include one or more buses. Although embodiments of the present application describe and illustrate a particular bus, the present application contemplates any suitable bus or interconnect.

[0227] The model processing device for inadvertent transmission rules can execute the model processing method for inadvertent transmission rules in the embodiment of the present application based on the acquired model processing of the inadvertent transmission rules, thereby realizing the combination of Figure 1 Describes the model processing approach for oblivious transfer rules.

[0228] In addition, in combination with the model processing method of the oblivious transmission rule in the above embodiment, the embodiment of the present application can provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by the processor, any one of the model processing methods of the oblivious transmission rule in the above embodiment is implemented.

[0229] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0230] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A model processing method for inadvertent transmission rules, applied to a model provider, characterized in that: include: Acquire a hybrid model set, wherein the hybrid model set includes a target model and a plurality of randomly generated simulation models; Obtaining target position information of the target model in the hybrid model set, and transmitting the hybrid model set to a model training end; receiving an encrypted model training result set transmitted by the model training end based on an oblivious transmission rule, and obtaining a target model training result based on the target location information and the encrypted model training result set, wherein the model training result set is obtained by training a model in the hybrid model set based on a training data set stored in the model training end, and the oblivious transmission rule is a data transmission rule set based on an oblivious transmission protocol and a specific application scenario; The receiving, based on the oblivious transmission rule, the encrypted model training result set transmitted by the model training end, and obtaining the target model training result based on the target location information and the encrypted model training result set includes: Obtain a preset prime number set, a first random number, and a second random number, and obtain a random parameter according to the first random number and the second random number; Sending the random parameter to the model training end, and receiving a first transmission parameter sent by the model training end, wherein the first transmission parameter is obtained by encrypting the model training result set based on the Chinese remainder theorem, the preset prime number set, the random parameter, and the random number set generated by the model training end; Obtaining a second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target location information, and sending the second transmission parameter to the model training end; receiving a third transmission parameter sent by the model training end, and obtaining the target model training result according to the third transmission parameter, the second random number and the target position information, wherein the third transmission parameter is obtained based on the second transmission parameter and the random number set; or, Obtain a preset prime number set, a first random number, and a second random number, and obtain a random parameter according to the first random number and the second random number; Sending the random parameter to the model training end, and receiving a first transmission parameter sent by the model training end, where the first transmission parameter is obtained according to the Chinese remainder theorem, the preset prime number set, the random parameter, and the random number set generated by the model training end; Obtaining a second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target location information, and sending the second transmission parameter to the model training end; Receive the third transmission parameter sent by the model training end, and obtain the target model training result according to the third transmission parameter, the second random number and the target position information, wherein the third transmission parameter is obtained by encrypting the model training result set based on the second transmission parameter and the random number set.

2. The method according to claim 1, characterized in that The acquiring target position information of the target model in the mixed model set comprises: The models in the mixed model set are randomly sorted, and the target position information of the target model in the mixed model set after the random sorting is obtained.

3. A model processing method for inadvertent transmission rules, applied to a model training end, characterized in that: include: Receiving a hybrid model set transmitted by a model provider, the hybrid model set comprising a target model stored in the model provider and a plurality of randomly generated simulation models; Acquire a training data set, and train the models in the hybrid model set based on the training data set to obtain a model training result set corresponding to the hybrid model set; Encrypting the model training result set based on the oblivious transmission rule, and transmitting the encrypted model training result set to the model provider, so that the model provider obtains the target model training result based on the encrypted model training result set and the pre-acquired target location information, wherein the target location information represents the location information of the target model in the mixed model set, and the oblivious transmission rule is a data transmission rule set based on the oblivious transmission protocol and a specific application scenario; The step of encrypting the model training result set based on the oblivious transmission rule and transmitting the encrypted model training result set to the model provider so that the model provider obtains the target model training result based on the encrypted model training result set and the pre-acquired target location information comprises: Get preset prime number sets and random number sets; Receive the random parameters sent by the model provider, and encrypt the model training result set based on the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set to obtain a first transmission parameter; The first transmission parameter is sent to the model providing end, and a second transmission parameter sent by the model providing end is received; the second transmission parameter is obtained according to the first transmission parameter, the preset prime number set, the target position information stored in the model providing end, and the first random number generated by the model providing end; According to the second transmission parameter and the random number set, a third transmission parameter is obtained, and the third transmission parameter is sent to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number, or, Get preset prime number sets and random number sets; Receiving the random parameters sent by the model providing end, and obtaining the first transmission parameters according to the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set generated by the model training end; The first transmission parameter is sent to the model providing end, and a second transmission parameter sent by the model providing end is received; the second transmission parameter is obtained according to the first transmission parameter, the preset prime number set, the target position information stored in the model providing end, and the first random number generated by the model providing end; The model training result set is encrypted according to the second transmission parameter and the random number set to obtain a third transmission parameter, and the third transmission parameter is sent to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number.

4. The method according to claim 3, characterized in that The target position information represents the position information of the target model in the mixed model set after random sorting.

5. A model processing device for inadvertent transmission rules, applied to a model provider, characterized in that: include: A model acquisition module, used to acquire a hybrid model set, wherein the hybrid model set includes a target model and a plurality of randomly generated simulation models; A module processing module, used for obtaining target position information of the target model in the mixed model set, and transmitting the mixed model set to a model training end; an encrypted transmission module, configured to receive an encrypted model training result set transmitted by the model training end based on an oblivious transmission rule, and obtain a target model training result based on the target location information and the encrypted model training result set, wherein the model training result set is obtained by training the model in the hybrid model set based on the training data set stored in the model training end, and the oblivious transmission rule is a data transmission rule set based on an oblivious transmission protocol and a specific application scenario; The receiving, based on the oblivious transmission rule, the encrypted model training result set transmitted by the model training end, and obtaining the target model training result based on the target location information and the encrypted model training result set includes: Obtain a preset prime number set, a first random number, and a second random number, and obtain a random parameter according to the first random number and the second random number; Sending the random parameter to the model training end, and receiving a first transmission parameter sent by the model training end, wherein the first transmission parameter is obtained by encrypting the model training result set based on the Chinese remainder theorem, the preset prime number set, the random parameter, and the random number set generated by the model training end; Obtaining a second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target location information, and sending the second transmission parameter to the model training end; receiving a third transmission parameter sent by the model training end, and obtaining the target model training result according to the third transmission parameter, the second random number and the target position information, wherein the third transmission parameter is obtained based on the second transmission parameter and the random number set; or, Obtain a preset prime number set, a first random number, and a second random number, and obtain a random parameter according to the first random number and the second random number; Sending the random parameter to the model training end, and receiving a first transmission parameter sent by the model training end, where the first transmission parameter is obtained according to the Chinese remainder theorem, the preset prime number set, the random parameter, and the random number set generated by the model training end; Obtaining a second transmission parameter according to the first random number, the first transmission parameter, the preset prime number set and the target location information, and sending the second transmission parameter to the model training end; Receive the third transmission parameter sent by the model training end, and obtain the target model training result according to the third transmission parameter, the second random number and the target position information, wherein the third transmission parameter is obtained by encrypting the model training result set based on the second transmission parameter and the random number set.

6. A model processing device for inadvertent transmission rules, applied to a model training end, characterized in that: include: A model receiving module, used for receiving a mixed model set transmitted by a model providing end, wherein the mixed model set includes a target model stored in the model providing end and a plurality of randomly generated simulation models; A model training module, used to obtain a training data set, and train the models in the hybrid model set based on the training data set to obtain a model training result set corresponding to the hybrid model set; A data transmission module, used for encrypting the model training result set based on an oblivious transmission rule, and transmitting the encrypted model training result set to the model provider, so that the model provider obtains a target model training result based on the encrypted model training result set and pre-acquired target location information, wherein the target location information represents location information of the target model in the hybrid model set, and the oblivious transmission rule is a data transmission rule set based on an oblivious transmission protocol and a specific application scenario; The step of encrypting the model training result set based on the oblivious transmission rule and transmitting the encrypted model training result set to the model provider so that the model provider obtains the target model training result based on the encrypted model training result set and the pre-acquired target location information comprises: Get preset prime number sets and random number sets; Receive the random parameters sent by the model provider, and encrypt the model training result set based on the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set to obtain a first transmission parameter; The first transmission parameter is sent to the model providing end, and a second transmission parameter sent by the model providing end is received; the second transmission parameter is obtained according to the first transmission parameter, the preset prime number set, the target position information stored in the model providing end, and the first random number generated by the model providing end; According to the second transmission parameter and the random number set, a third transmission parameter is obtained, and the third transmission parameter is sent to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number, or, Get preset prime number sets and random number sets; Receiving the random parameters sent by the model providing end, and obtaining the first transmission parameters according to the Chinese remainder theorem, the preset prime number set, the random parameters and the random number set generated by the model training end; The first transmission parameter is sent to the model providing end, and a second transmission parameter sent by the model providing end is received; the second transmission parameter is obtained according to the first transmission parameter, the preset prime number set, the target position information stored in the model providing end, and the first random number generated by the model providing end; The model training result set is encrypted according to the second transmission parameter and the random number set to obtain a third transmission parameter, and the third transmission parameter is sent to the model provider, so that the model provider obtains the target model training result based on the third transmission parameter, the target location information and the pre-generated second random number.

7. A computer 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 computer program, the model processing method of the oblivious transmission rule according to any one of claims 1 to 4 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the model processing method of the oblivious transmission rule according to any one of claims 1 to 4 is implemented.

Images