Data transmission method and system for artificial intelligence

Abstract

The present invention provides a data transmission method and system for artificial intelligence, belonging to the field of digital information transmission. The method includes: a server updating a first model parameter of a first neural network model; sending the first model parameter encrypted using a first combined key to a terminal device; obtaining a corresponding second model parameter in the first neural network model based on the time the first model parameter was sent, and updating the second key in the server's first combined key based on the second model parameter; the terminal device receiving the first model parameter and decrypting the first model parameter using the second combined key; updating the model parameter of the second neural network model based on the first model parameter; obtaining a corresponding third model parameter in the second neural network model based on the time the first model parameter was received, and updating the fourth key in the terminal device's second combined key based on the third model parameter. Implementing the technical solution of the present invention can improve the security of data transmission.

Description

Technical Field

[0001] The present invention relates to the field of digital information transmission, and in particular to a data transmission method and system for artificial intelligence. Background Art

[0002] In the field of artificial intelligence, servers are often used to train and update various network models. These trained and updated network models are then sent to terminal devices, which then use them to recognize relevant images. For example, in the field of intelligent driving, servers use images captured by individual vehicles and their corresponding recognition information to update and train the image recognition models in autonomous driving systems. The updated image recognition models are then sent to the corresponding vehicles, enabling them to autonomously utilize the latest image recognition models. The confidentiality of data transmitted from servers to vehicles and from vehicles to servers is crucial. A compromise of these keys could easily lead to immeasurable losses. Therefore, improving the security of AI data transmission is an urgent technical challenge.

[0003] Patent publication number CN111565177A discloses a method and device for encrypting and transmitting vehicle-mounted data. The method includes: a data transmitter and a data receiver maintaining the same encryption algorithm and key generation rules, where the key is a string containing a first password and timestamp information; the transmitter generating a key according to a preset rule, and encrypting the vehicle-mounted data and key according to a preset encryption algorithm to generate a first verification string; the vehicle-mounted data and the verification string are packaged and sent; the receiver receives the data packet and parses it to obtain the vehicle-mounted data; generates a key according to the preset rule, and encrypts the parsed vehicle-mounted data according to the preset encryption algorithm to obtain a second verification string; and compares the first and second verification strings to see if they are consistent. If not, the received data packet is discarded.

[0004] Patent publication number CN110912941A discloses a method and apparatus for multicast data transmission and processing. The method comprises: obtaining multicast data generated by encrypting and packaging playback data, wherein the multicast data includes: encrypted playback data, a data check field, a key ID field, and an encryption scheme field; verifying the multicast data based on the data check field; after verification, determining a decryption scheme based on the encryption scheme field and a decryption key based on the key ID field; and decrypting the encrypted playback data based on the decryption scheme and the decryption key to obtain the playback data. This invention improves the security of playback data transmission.

[0005] Patent publication number CN110139273A discloses a security reinforcement method and system for IoT wireless transmission. The method includes: authorization filing; issuing an ID certificate; issuing a new key; issuing an ID certificate; and confirming successful key update. This method utilizes backend authorization, gateway authentication, and random changes in communication data encryption keys to reinforce IoT wireless transmission. This method enhances the security of existing IoT wireless transmission data, while offering high security, low complexity, and low cost, making it suitable for widespread application.

[0006] The invention patent with publication number CN111279732A provides a method, network device, and terminal device for transmitting information. The method includes: a first network device obtains indication information, which is used to indicate that the integrity protection IP check of the data on the data radio bearer DRB has failed; the first network device sends the indication information to a second network device. In an embodiment of the present invention, through the indication information, when the IP check of the data on the DRB fails, the second network device can update the key of the terminal device, or the second network device can release the RRC connection where the DRB is located. In this way, security risks are eliminated, communication security is guaranteed, and the success rate of data transmission is improved.

[0007] If the key of the above-mentioned disclosed technical solution is stolen, it is easy to cause data transmission insecurity. In particular, although the invention with publication number CN110139273A discloses using a key update method to improve key security and thus improve data transmission security, it requires the issuance of a new key. Therefore, the new key is easily stolen during the issuance process, which in turn causes data transmission insecurity. Summary of the Invention

[0008] In order to solve at least one technical problem in the prior art, the present invention provides a data transmission method and system for artificial intelligence to improve the security of data transmission.

[0009] According to a first aspect of the present invention, a data transmission method for artificial intelligence is provided, wherein a server stores a trained first neural network model, and a terminal device stores a trained second neural network model, wherein the first neural network model is identical to the second neural network model, and the method comprises:

[0010] The server updates the first model parameter of the first neural network model;

[0011] The server sends the first model parameter encrypted according to the first combined key to the terminal device, wherein the first combined key includes the combined first key and the second key;

[0012] The server obtains corresponding second model parameters in the first neural network model according to the sending time of the first model parameters, and updates the second key of the server according to the second model parameters;

[0013] The terminal device receives the first model parameter and decrypts the first model parameter according to a second combined key, where the second combined key includes a third key and a fourth key, where the third key is the same as the first key, and the fourth key is the same as the second key;

[0014] The terminal device updates the model parameters of the second neural network model according to the first model parameters;

[0015] The terminal device obtains the corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, and updates the fourth key of the terminal device according to the third model parameters.

[0016] Optionally, the server updates the first model parameter of the first neural network model, including:

[0017] Get training samples;

[0018] The first neural network model is trained according to the training sample to update first model parameters of the first neural network model.

[0019] Optionally, the server obtains corresponding second model parameters in the first neural network model according to the sending time of the first model parameters, including:

[0020] The server obtains corresponding model parameters from the first neural network model as the second model parameters according to the time value of the sending time;

[0021] The terminal device obtains corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, including:

[0022] The terminal device obtains corresponding model parameters from the second neural network model as the third model parameters based on the time value of the receiving time.

[0023] Optionally, when the difference between the receiving time and the whole hour is less than or equal to 1 minute, the terminal device requests the sending time of the first model parameter from the server, and obtains the corresponding model parameter from the second neural network model as the third model parameter based on the time value of the sending time of the first model parameter.

[0024] Optionally, updating the second key of the server according to the second model parameter includes:

[0025] Updating the second key according to the last N values ​​of the second model parameter, where N is the length of the second key;

[0026] Updating the fourth key of the terminal device according to the third model parameter includes:

[0027] The fourth key is updated according to the last M values ​​of the third model parameter, where M is the length value of the fourth key and M is equal to N.

[0028] Optionally, updating the second key according to the last N values ​​of the second model parameter includes:

[0029] updating the second key by adding the last N values ​​of the second model parameter to the corresponding N values ​​of the second key;

[0030] Updating the fourth key according to the last M values ​​of the third model parameter includes:

[0031] The fourth key is updated by adding the last M values ​​of the third model parameter to the corresponding M values ​​of the fourth key.

[0032] Optionally, the terminal device sends data encrypted according to the second combined key to the server, and the server decrypts the encrypted data according to the first combined key.

[0033] A second aspect of the present invention provides a data transmission system for artificial intelligence, comprising a terminal device and a server, wherein the server stores a trained first neural network model, and the terminal device stores a trained second neural network model, wherein the first neural network model is identical to the second neural network model;

[0034] The server includes:

[0035] A first parameter updating module, configured to update a first model parameter of the first neural network model;

[0036] A first transceiver module is configured to send the first model parameter encrypted according to a first combined key to a terminal device, wherein the first combined key includes a combined first key and a second key;

[0037] a first key updating module, configured to obtain corresponding second model parameters in the first neural network model according to a sending time of the first model parameters, and update a second key of the server according to the second model parameters;

[0038] The terminal device includes:

[0039] a second transceiver module, configured to receive the first model parameter and decrypt the first model parameter according to a second combined key, where the second combined key includes a third key and a fourth key combined, where the third key is the same as the first key, and the fourth key is the same as the second key;

[0040] A second parameter updating module, configured to update the model parameters of the second neural network model according to the first model parameters;

[0041] The second key updating module is used to obtain the corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, and update the fourth key of the terminal device according to the third model parameters.

[0042] Optionally, the first key updating module is configured to obtain the corresponding second model parameter in the first neural network model according to the sending time of the first model parameter, specifically to:

[0043] The server obtains corresponding model parameters from the first neural network model as the second model parameters according to the time value of the sending time;

[0044] The second key updating module is configured to obtain the corresponding third model parameter in the second neural network model according to the reception time of the first model parameter, specifically to:

[0045] According to the receiving time, corresponding model parameters are obtained from the second neural network model as the third model parameters.

[0046] Optionally, the second key update module is also used to request the sending time of the first model parameter from the server when the difference between the receiving time and the whole time is less than or equal to 1 minute, and obtain the corresponding model parameter from the second neural network model as the third model parameter according to the time value of the sending time of the first model parameter.

[0047] One or more technical solutions provided in the embodiments of the present application can improve the security of data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings illustrate exemplary embodiments of the present invention and together with the description serve to explain the principles of the present invention. These drawings are included to provide a further understanding of the present invention and are incorporated in and constitute a part of this specification.

[0049] Figure 1 A schematic diagram illustrating an example system in which the various methods described herein may be implemented according to an exemplary embodiment of the present invention;

[0050] Figure 2 A flow chart of a data transmission method for artificial intelligence according to an exemplary embodiment of the present invention is shown;

[0051] Figure 3 A schematic block diagram of a data transmission system for artificial intelligence according to an exemplary embodiment of the present invention is shown. DETAILED DESCRIPTION

[0052] Embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0053] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0054] As used herein, the term "including" and its variations are open-ended, meaning "including but not limited to." The term "based on" means "based, at least in part, on." It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used solely to distinguish between different devices, modules, or units, and are not intended to limit the order or interdependence of the functions performed by these devices, modules, or units.

[0055] It should be noted that the modifications of "one" and "multiple" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0056] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0057] The image recognition method based on deep learning of the present invention can be applied to Figure 1 See the implementation environment diagram shown in Figure 1 The implementation environment includes a terminal device 101 and a server 102, which are connected via a network 103. The server stores a trained first neural network model, and the terminal device stores a second neural network model that is identical to the first neural network model. The terminal device can be a robot, a mobile phone, a tablet, a computer, etc.

[0058] Hereinafter, aspects of the present invention will be described with reference to the accompanying drawings.

[0059] See also Figure 2 A data transmission method for artificial intelligence, wherein a server stores a trained first neural network model, and a terminal device stores a trained second neural network model, wherein the first neural network model is identical to the second neural network model, and the method comprises:

[0060] S201, the server updates the first model parameters of the first neural network model.

[0061] In one embodiment, step S202 may obtain a training sample and train a first neural network model based on the training sample to update the first model parameters of the first neural network model. The training sample may be obtained from a terminal device or acquired in other ways. The training sample herein is a new training sample used to train and update the model parameters of the first neural network model. It is understood that when training the first neural network model, its model parameters may be updated in whole or in part. The first model parameters are the updated model parameters.

[0062] It can be known that the first neural network model can be a neural network model such as an image recognition model, a speech recognition model, an object classification model, etc., wherein the first neural network model is the same as the second neural network model, which means that the first neural network model is exactly the same as the second neural network model, that is, the model structure is the same and the model parameters are also the same.

[0063] S202: The server sends a first model parameter encrypted according to a first combined key to the terminal device, where the first combined key includes a combined first key and a second key.

[0064] In step S202, the first key combination includes a first key and a second key. The first key and the second key can be combined in series. For example, if the first key is A234 and the second key is 123456, the first key combination can be A234123456. Of course, other combinations are also possible, as long as the first key combination and the second key combination are the same.

[0065] In step S202, an existing symmetric encryption algorithm, such as an RC encryption algorithm, may be used to encrypt the first model parameter according to the first combined key.

[0066] S203, the server obtains the corresponding second model parameters in the first neural network model according to the sending time of the first model parameters, and updates the second key of the server according to the second model parameters.

[0067] In one embodiment, when obtaining the corresponding second model parameter in the first neural network model according to the sending time of the first model parameter, the corresponding model parameter can be obtained from the first neural network model as the second model parameter according to the time value of the sending time. For example, when the first model parameter is sent at 14:00 on February 12, 2022, the time value of the sending time is 14, and the 14th model parameter in the first neural network model can be obtained at this time. The order of the model parameters in the first neural network model can be pre-set. Of course, according to actual conditions, the corresponding model parameter can be obtained from the first neural network model as the second model parameter according to the month value, day value, and time value of the sending time. For example, when the first model parameter is sent at 14:00 on February 12, 2022, the 2nd, 12th, and 14th model parameters in the first neural network model can be obtained as model parameters.

[0068] In one embodiment, when the second key of the server is updated according to the second model parameter, the second key can be updated specifically according to the last N values ​​of the second model parameter, where N is the length value of the second key. Exemplarily, the second key can be updated by adding the last N values ​​of the second model parameter to the corresponding N values ​​of the second key. For example, when the second key is 123456, N is 6, and when the second model parameter is 0.0000186123123, the updated second key can be 123456+123123, that is, the updated second key is 248579. It can be known that when the added value exceeds the limit, the portion exceeding the limit can be deleted. It should be noted that the above-mentioned first key and second key are only for exemplary purposes.

[0069] S204, the terminal device receives the first model parameter and decrypts the first model parameter according to the second combined key, where the second combined key includes a combined third key and a fourth key, where the third key is the same as the first key, and the fourth key is the same as the second key.

[0070] The second combined key is the same as the first combined key. When the server updates the second key, the terminal device also updates the fourth key, so that the fourth key is always the same as the second key, that is, the second combined key is always the same as the first combined key. The first key can remain unchanged.

[0071] S205: The terminal device updates the model parameters of the second neural network model according to the first model parameters.

[0072] In step S205, the model parameters of the second neural network model are updated according to the first model parameters. Since the first model parameters are the updated parameters of the first neural network model, and the second neural network model is the same as the first neural network model, after the model parameters of the second neural network model are updated according to the first model parameters, the second neural network model can be kept the same as the first neural network model.

[0073] S206: The terminal device obtains corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, and updates the fourth key of the terminal device according to the third model parameters.

[0074] In step S206, when obtaining the corresponding third model parameters in the second neural network model based on the reception time of the first model parameters, the acquisition method is the same as the acquisition method of "obtaining the corresponding second model parameters in the first neural network model based on the transmission time of the first model parameters" in step S203. Because the second neural network model is identical to the first neural network model, their transmission and reception times are also substantially identical. Therefore, when obtaining the corresponding third model parameters in the second neural network model based on the reception time of the first model parameters, the obtained third model parameters are identical to the second model parameters.

[0075] In one embodiment, step 206 may obtain corresponding model parameters from the second neural network model as third model parameters based on the time value of the reception time.

[0076] Since there may be a time difference between the sending time of the server and the receiving time of the terminal device during data transmission, the time difference may cause the third model parameters to be different from the second model parameters. Therefore, in order to prevent this problem, optionally, when the difference between the receiving time value and the whole time is less than or equal to 1 minute, the terminal device requests the sending time of the first model parameter from the server, and obtains the corresponding model parameters from the second neural network model as the third model parameters based on the time value of the sending time of the first model parameter. In this way, it is possible to prevent the third model parameters from being inconsistent with the second model parameters due to the time difference between the sending time of the server and the receiving time of the terminal device, thereby preventing the fourth key from being inconsistent with the second key. Moreover, since the difference between most time values ​​and the whole time is greater than 1, there is no need to request the sending time of the first model parameter in most cases.

[0077] In one embodiment, when updating the fourth key of the terminal device based on the third model parameter, the fourth key may be updated based on the last M values ​​of the third model parameter, where M is the length of the fourth key and M is equal to N. Exemplarily, the fourth key is updated by adding the last M values ​​of the third model parameter to the corresponding M values ​​of the fourth key. This step corresponds to "updating the second key by adding the last N values ​​of the second model parameter to the corresponding N values ​​of the second key" in step 203. This step can ensure that the fourth key of the terminal device is consistent with the second key of the server, and will not be repeated here.

[0078] The present invention provides a data transmission method for artificial intelligence. In this method, a server, acting as a model update training server, updates the first model parameters of the first neural network based on new training samples after the first neural network model has been trained. The server then sends the updated first model parameters of the first neural network model to a terminal device, allowing the terminal device to update its second neural network model based on the first model parameters, making the second neural network model consistent with the first neural network model. Because the data is transmitted with the first model parameters, and when the first neural network model is trained and updated, only a portion of the model parameters are updated, even if the first model parameters are stolen during transmission, it is difficult for the thief to obtain the corresponding combined key. However, in this method, when the server sends the updated first model parameters to the terminal device, the second key of the first combined key is updated, and the corresponding fourth key of the second combined key of the terminal device is also updated synchronously. This allows the terminal device to synchronize the key update with the server without the server sending the updated key. This makes it difficult for the thief to crack the key even if he obtains the first model parameters and the original first key, first combined key, or second combined key, thereby improving the security of the data transmission. Furthermore, this data transmission method requires virtually no additional data transmission, ensuring transmission efficiency.

[0079] In one embodiment, the terminal device sends data encrypted according to the second combined key to the server, and the server decrypts the encrypted data according to the first combined key.

[0080] See also Figure 3 The present invention also provides a data transmission system for artificial intelligence, comprising a terminal device and a server, wherein the server stores a trained first neural network model, and the terminal device stores a trained second neural network model, wherein the first neural network model is identical to the second neural network model;

[0081] The server includes:

[0082] A first parameter updating module 301 is used to update a first model parameter of the first neural network model;

[0083] A first transceiver module 302 is configured to send a first model parameter encrypted according to a first combined key to a terminal device, wherein the first combined key includes a combined first key and a second key;

[0084] A first key updating module 303 is configured to obtain corresponding second model parameters in the first neural network model according to the sending time of the first model parameters, and update the second key of the server according to the second model parameters;

[0085] Terminal equipment includes:

[0086] A second transceiver module 304 is configured to receive the first model parameter and decrypt the first model parameter according to a second combined key, where the second combined key includes a third key and a fourth key, where the third key is the same as the first key and the fourth key is the same as the second key;

[0087] A second parameter updating module 305 is configured to update the model parameters of the second neural network model according to the first model parameters;

[0088] The second key updating module 306 is used to obtain the corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, and update the fourth key of the terminal device according to the third model parameters.

[0089] In one embodiment, the first key updating module is configured to obtain the corresponding second model parameters in the first neural network model according to the sending time of the first model parameters, specifically configured to:

[0090] The server obtains corresponding model parameters from the first neural network model as second model parameters according to the time value of the sending time;

[0091] The second key updating module is configured to obtain corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, specifically for:

[0092] According to the time value of the receiving time, corresponding model parameters are obtained from the second neural network model as third model parameters.

[0093] In one embodiment, the second key update module is also used to request the sending time of the first model parameter from the server when the difference between the receiving time and the whole time is less than or equal to 1 minute, and obtain the corresponding model parameter from the second neural network model as the third model parameter according to the time value of the sending time of the first model parameter.

[0094] An exemplary embodiment of the present invention further provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, which, when executed by the at least one processor, causes the electronic device to perform a method according to an embodiment of the present invention.

[0095] The structural block diagram of an electronic device that can serve as a server or client of the present disclosure will now be described, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0096] An electronic device includes a computing unit that can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) or a computer program loaded from a storage unit into a random access memory (RAM). Various programs and data required for device operation can also be stored in the RAM. The computing unit, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0097] Exemplary embodiments of the present invention further provide a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is used to cause the computer to perform a method according to an embodiment of the present invention.

[0098] The above are only specific embodiments of the present invention, but the technical features of the present invention are not limited thereto. Any changes or modifications made by any technician in this field within the scope of the present invention are included in the patent scope of the present invention.

Claims

1. A data transmission method for artificial intelligence, characterized in that: The server stores a trained first neural network model, and the terminal device stores a trained second neural network model, wherein the first neural network model is identical to the second neural network model, and the method includes: The server updates the first model parameter of the first neural network model; The server sends the first model parameter encrypted according to the first combined key to the terminal device, wherein the first combined key includes the combined first key and the second key; The server obtains corresponding second model parameters in the first neural network model according to the sending time of the first model parameters, and updates the second key of the server according to the second model parameters; The terminal device receives the first model parameter and decrypts the first model parameter according to a second combined key, where the second combined key includes a third key and a fourth key, where the third key is the same as the first key, and the fourth key is the same as the second key; The terminal device updates the model parameters of the second neural network model according to the first model parameters; The terminal device obtains corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, and updates the fourth key of the terminal device according to the third model parameters; The server obtains corresponding second model parameters in the first neural network model according to the sending time of the first model parameters, including: The server obtains corresponding model parameters from the first neural network model as the second model parameters according to the time value of the sending time; The terminal device obtains corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, including: The terminal device obtains corresponding model parameters from the second neural network model as the third model parameters based on the time value of the receiving time. When the difference between the receiving time and the whole time is less than or equal to 1 minute, the terminal device requests the sending time of the first model parameters from the server, and obtains corresponding model parameters from the second neural network model as the third model parameters based on the time value of the sending time of the first model parameters.

2. The data transmission method for artificial intelligence according to claim 1, characterized in that: The server updates the first model parameter of the first neural network model, including: Get training samples; The first neural network model is trained according to the training sample to update first model parameters of the first neural network model.

3. The data transmission method for artificial intelligence according to claim 1 or 2, characterized in that: Updating the second key of the server according to the second model parameter includes: Updating the second key according to the last N values ​​of the second model parameter, where N is the length of the second key; Updating the fourth key of the terminal device according to the third model parameter includes: The fourth key is updated according to the last M values ​​of the third model parameter, where M is the length value of the fourth key and M is equal to N.

4. The data transmission method for artificial intelligence according to claim 3, characterized in that: Updating the second key according to the last N values ​​of the second model parameter includes: updating the second key by adding the last N values ​​of the second model parameter to the corresponding N values ​​of the second key; Updating the fourth key according to the last M values ​​of the third model parameter includes: The fourth key is updated by adding the last M values ​​of the third model parameter to the corresponding M values ​​of the fourth key.

5. The data transmission method for artificial intelligence according to claim 1, characterized in that: The terminal device sends data encrypted according to the second combined key to the server, and the server decrypts the encrypted data according to the first combined key.

6. A data transmission system for artificial intelligence, characterized in that: The method comprises a terminal device and a server, wherein the server stores a trained first neural network model, and the terminal device stores a trained second neural network model, wherein the first neural network model is identical to the second neural network model; The server includes: A first parameter updating module, configured to update a first model parameter of the first neural network model; A first transceiver module is configured to send the first model parameter encrypted according to a first combined key to a terminal device, wherein the first combined key includes a combined first key and a second key; a first key updating module, configured to obtain corresponding second model parameters in the first neural network model according to a sending time of the first model parameters, and update a second key of the server according to the second model parameters; The terminal device includes: a second transceiver module, configured to receive the first model parameter and decrypt the first model parameter according to a second combined key, where the second combined key includes a third key and a fourth key combined, where the third key is the same as the first key, and the fourth key is the same as the second key; A second parameter updating module, configured to update the model parameters of the second neural network model according to the first model parameters; a second key updating module, configured to obtain corresponding third model parameters in the second neural network model according to the reception time of the first model parameters, and update the fourth key of the terminal device according to the third model parameters; The first key updating module is configured to obtain the corresponding second model parameters in the first neural network model according to the sending time of the first model parameters, specifically to: The server obtains corresponding model parameters from the first neural network model as the second model parameters according to the time value of the sending time; The second key updating module is configured to obtain the corresponding third model parameter in the second neural network model according to the reception time of the first model parameter, specifically to: According to the receiving time, corresponding model parameters are obtained from the second neural network model as the third model parameters, wherein when the difference between the receiving time and the whole time is less than or equal to 1 minute, the sending time of the first model parameters is requested from the server, and according to the time value of the sending time of the first model parameters, corresponding model parameters are obtained from the second neural network model as the third model parameters.

Images