Intelligent service strategy pushing method and system based on privacy protection

Abstract

The invention discloses an intelligent service strategy pushing method and system based on privacy protection, the system comprises a local sensing terminal cluster and a server cluster, the local sensing terminal cluster comprises a plurality of intelligent sensing devices, and the plurality of intelligent sensing devices are in communication connection with the server cluster. According to the embodiment of the invention, each intelligent sensing device can send the current sensing data to the server cluster in a private data encryption mode, performs encrypted data classification to obtain the current user classification result, and obtains the current service recommendation strategy corresponding to the current user classification result based on the personalized service recommendation engine; the hidden danger of original data leakage of the current sensing data is avoided, and the data processing efficiency is improved through the personalized service recommendation engine in the server cluster, which is specially used for the data processing of the service recommendation strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of privacy protection, in particular to a smart service policy pushing method and system based on privacy protection. BACKGROUND

[0002] At present, smart devices such as smart phones, smart watches, smart homes, etc. have been widely used in users' daily life. The above-mentioned smart devices can actively perceive the operation behavior (such as screen click operation, current screen stay time, device on-off operation, etc.) or the physical data (such as heart rate data, etc.) of the user. These user operation device behavior data or detected physical data are generally uploaded to the cloud by the smart device in plaintext data to obtain corresponding feedback data, for example, the feedback data is the recommended information (such as video data, webpage link, etc.) corresponding to the user operation device behavior data. However, the above-mentioned interaction mode based on plaintext data not only has the risk of data leakage, but also the cloud center does not divide the multi-functional nodes, but processes a large amount of original plaintext data in the same cloud server or multiple server nodes, which may exist the problem of slow processing rate and response delay. SUMMARY

[0003] The embodiments of the present application provide a smart service policy pushing method and system based on privacy protection, which aims to solve the problem that the related data actively perceived by the smart device is uploaded to the cloud in plaintext data to obtain corresponding feedback data in the prior art, which not only has the risk of data leakage, but also the cloud center processes a large amount of original plaintext data, which may exist the problem of slow processing rate and response delay.

[0004] In a first aspect, the embodiments of the present application provide a smart service policy pushing method based on privacy protection, which includes a local perception terminal cluster and a server cluster. The local perception terminal cluster includes a plurality of smart perception devices, and the plurality of smart perception devices are in communication connection with the server cluster. The plurality of smart perception devices obtain current perception data collected by user authorization according to the corresponding preset data dimension in a plurality of preset data dimensions, and send the encrypted privacy data corresponding to each current perception data to the server cluster. The server cluster inputs the encrypted privacy data set composed of the encrypted privacy data uploaded by the plurality of smart perception devices respectively into a classification model to obtain a current user classification result. The server cluster inputs the current user classification result into a pre-constructed personalized service recommendation engine to obtain a current service recommendation strategy, and sends the current service recommendation strategy to the plurality of smart perception devices. Each of the plurality of intelligent sensing devices obtains a corresponding current service recommendation sub-strategy from the current service recommendation strategy, and executes the current service recommendation sub-strategy to perform device action control.

[0005] In some possible embodiments, the plurality of intelligent sensing devices at least includes a smartphone, a smartwatch, and a smart home; The plurality of intelligent sensing devices obtains current sensing data authorized to be collected by a user in a corresponding preset data dimension of a plurality of preset data dimensions, and sends corresponding encrypted privacy data of each current sensing data to the server cluster, including: The smartphone in the plurality of intelligent sensing devices obtains a user device operation behavior parameter sequence authorized to be collected by a user, encrypts sequence features corresponding to the user device operation behavior parameter sequence based on a locally deployed homomorphic encryption model to obtain first encrypted privacy data, and sends the first encrypted privacy data to the server cluster; The smartwatch in the plurality of intelligent sensing devices obtains a user vital sign parameter sequence authorized to be collected by a user, encrypts sequence features corresponding to the user vital sign parameter sequence based on a locally deployed homomorphic encryption model to obtain second encrypted privacy data, and sends the second encrypted privacy data to the server cluster; The smart home in the plurality of intelligent sensing devices obtains an environment parameter sequence authorized to be collected by a user and in a current stay space of the user, encrypts sequence features corresponding to the environment parameter sequence based on a locally deployed homomorphic encryption model to obtain third encrypted privacy data, and sends the third encrypted privacy data to the server cluster.

[0006] In some possible embodiments, the smartphone in the plurality of intelligent sensing devices obtains a user device operation behavior parameter sequence authorized to be collected by a user, encrypts sequence features corresponding to the user device operation behavior parameter sequence based on a locally deployed homomorphic encryption model to obtain first encrypted privacy data; The smartphone in the plurality of intelligent sensing devices obtains a user device operation behavior parameter sequence authorized to be collected by a user, extracts features of the user device operation behavior parameter sequence based on a local behavior feature extractor to obtain user behavior features; The user behavior features are encrypted based on the homomorphic encryption model to obtain the first encrypted privacy data.

[0007] In some possible embodiments, the smartwatch in the plurality of intelligent sensing devices obtains a user vital sign parameter sequence authorized to be collected by a user, encrypts sequence features corresponding to the user vital sign parameter sequence based on a locally deployed homomorphic encryption model to obtain second encrypted privacy data, including: The smartwatches among the multiple smart sensing devices acquire the heart rate variability parameter sequence authorized by the user and use it as the user's vital sign parameter sequence. Based on the local physiological feature extractor, the user's vital sign parameter sequence is used to extract features to obtain the user's physiological features. The user's physiological characteristics are classified based on a local lightweight classification model to obtain the current user's stress level parameters. The second encrypted privacy data is obtained by encrypting the current user stress level parameters based on the homomorphic encryption model.

[0008] In some possible embodiments, the smart home includes at least smart curtains; the smart home among the plurality of smart sensing devices acquires a sequence of environmental parameters collected with user authorization, which is the user's current location, and encrypts the sequence features corresponding to the environmental parameter sequence based on a locally deployed homomorphic encryption model to obtain third encrypted privacy data, including: The smart curtains among the multiple smart sensing devices acquire the current environmental sound intensity sequence within the user's current space based on the built-in sound sensor, which is authorized by the user and serves as the environmental parameter sequence. The environmental feature is then extracted from the environmental parameter sequence using a local environmental feature extractor to obtain environmental features. The third encrypted privacy data is obtained by encrypting the environmental features based on the homomorphic encryption model.

[0009] In some possible embodiments, the server cluster inputs an encrypted privacy dataset, composed of encrypted privacy data uploaded by the multiple smart sensing devices, into a classification model to obtain the current user classification result, including: The encrypted privacy dataset is obtained from the plurality of smart sensing devices, including the first encrypted privacy data corresponding to the smartphone, the second encrypted privacy data corresponding to the smartwatch, and the third encrypted privacy data corresponding to the smart home. The encrypted privacy dataset is securely aggregated to obtain encrypted set features, which are then input into the classification model to obtain the current user classification result.

[0010] In some possible embodiments, the server cluster inputs the current user classification result into a pre-built personalized service recommendation engine to obtain the current service recommendation strategy, including: Obtain multiple server recommendation strategies in the personalized service recommendation engine, and the user classification result identifier value corresponding to each server recommendation strategy; If a user classification result identifier value corresponding to a server recommendation strategy is the same as the current user classification result, then the corresponding server recommendation strategy is obtained as the current service recommendation strategy.

[0011] In some possible embodiments, each of the plurality of intelligent sensing devices obtains a corresponding current service recommendation sub-policy from the current service recommendation policy and executes the current service recommendation sub-policy to perform device action control, including: The smartphones among the multiple intelligent sensing devices obtain a first current service recommendation sub-strategy from the current service recommendation strategy, and obtain corresponding recommendation learning data based on the first current service recommendation sub-strategy and display it locally; The smartwatches among the multiple smart sensing devices obtain a second current service recommendation sub-strategy from the current service recommendation strategy, and obtain corresponding recommendation behavior data according to the second current service recommendation sub-strategy and provide local prompts; The smart home devices among the multiple smart sensing devices obtain a third current service recommendation sub-strategy from the current service recommendation strategy, and obtain the corresponding recommended execution action according to the third current service recommendation sub-strategy and perform the corresponding operation accordingly.

[0012] In some possible embodiments, after each of the plurality of intelligent sensing devices obtains a corresponding current service recommendation sub-policy from the current service recommendation policy and executes the current service recommendation sub-policy to perform device action control, the method further includes: The smartphones or smartwatches among the multiple smart sensing devices send rating and evaluation information for the current service recommendation strategy to the server cluster.

[0013] Secondly, embodiments of the present invention provide a privacy-preserving intelligent service policy push system, which includes a local sensing terminal cluster and a server cluster. The local sensing terminal cluster includes multiple intelligent sensing devices, and all of the multiple intelligent sensing devices are communicatively connected to the server cluster. The privacy-preserving intelligent service policy push system is used to execute the privacy-preserving intelligent service policy push method as described in the first aspect and any possible embodiments thereunder.

[0014] This invention provides a method and system for pushing intelligent service policies based on privacy protection, including a local sensing terminal cluster and a server cluster. The local sensing terminal cluster includes multiple intelligent sensing devices, all of which are communicatively connected to the server cluster. The multiple intelligent sensing devices acquire user-authorized current sensing data according to a set of preset data dimensions, and send the encrypted privacy data corresponding to each set of current sensing data to the server cluster. The server cluster inputs the encrypted privacy dataset composed of the encrypted privacy data uploaded by the multiple intelligent sensing devices into a classification model to obtain the current user classification result. The server cluster inputs the current user classification result into a pre-built personalized service recommendation engine to obtain a current service recommendation policy, and sends the current service recommendation policy to the multiple intelligent sensing devices. Each of the multiple intelligent sensing devices obtains a corresponding current service recommendation sub-policy from the current service recommendation policy and executes the current service recommendation sub-policy to control device actions. The embodiments of the present invention enable each intelligent sensing device to send the current sensing data to the server cluster in an encrypted and privacy-preserving manner, and perform encrypted data classification to obtain the current user classification result. Based on the personalized service recommendation engine, the current service recommendation strategy corresponding to the current user classification result is obtained and sent to each intelligent sensing device in a timely manner, avoiding the risk of leakage of the original sensing data. Furthermore, the personalized service recommendation engine in the server cluster is dedicated to the data processing of service recommendation strategies, which improves the data processing efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram illustrating an application scenario of the privacy-protected intelligent service strategy push system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the privacy-protected intelligent service strategy push method provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of a sub-process of the privacy-protected intelligent service strategy push method provided in an embodiment of the present invention; Figure 4 This is another sub-process diagram of the privacy-protected intelligent service strategy push method provided in the embodiments of the present invention; Figure 5 This is another sub-process diagram of the privacy-protected intelligent service strategy push method provided in the embodiments of the present invention; Figure 6 This is another sub-process diagram of the privacy-protected intelligent service strategy push method provided in the embodiments of the present invention; Figure 7 This is a schematic block diagram of a privacy-protected intelligent service strategy push system provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0019] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0020] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0021] Please also refer to Figure 1 and Figure 2 ,in Figure 1 This is a schematic diagram illustrating an application scenario of the privacy-protected intelligent service strategy push method according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the privacy-preserving intelligent service policy push method provided in an embodiment of the present invention. Figure 1 As shown, the privacy-preserving intelligent service policy push method provided in this embodiment of the invention specifically applies a privacy-preserving intelligent service policy push system, which includes a local sensing terminal cluster 10 and a server cluster 20. The local sensing terminal cluster 10 is communicatively connected to the server cluster 20. The local sensing terminal cluster 10 includes multiple intelligent sensing devices, and all of the multiple intelligent sensing devices are communicatively connected to the server cluster 20. Figure 2As shown, the privacy-preserving intelligent service policy push method includes steps S110-S140.

[0022] S110. The multiple intelligent sensing devices acquire the current sensing data authorized by the user according to the corresponding preset data dimension among multiple preset data dimensions, and send the encrypted privacy data corresponding to each current sensing data to the server cluster.

[0023] In this embodiment, the multiple intelligent sensing devices serve as data acquisition terminals. Each intelligent sensing device acquires the current sensing data authorized by the user in a corresponding preset data dimension, encrypts it locally, and then uploads it to the server cluster. This avoids sending the current sensing data of each intelligent sensing device to the server cluster for data processing in plaintext.

[0024] In one embodiment, the plurality of smart sensing devices includes at least smartphones, smartwatches, and smart home devices; such as Figure 3 As shown, step S110 includes: S111. The smartphone among the plurality of intelligent sensing devices acquires the user device operation behavior parameter sequence authorized by the user, encrypts the sequence features corresponding to the user device operation behavior parameter sequence based on the locally deployed homomorphic encryption model to obtain the first encrypted privacy data, and sends the first encrypted privacy data to the server cluster. S112. The smartwatch among the multiple smart sensing devices acquires the user's vital sign parameter sequence authorized by the user, encrypts the sequence features corresponding to the user's vital sign parameter sequence based on the locally deployed homomorphic encryption model to obtain the second encrypted privacy data, and sends the second encrypted privacy data to the server cluster. S113. The smart home system among the multiple smart sensing devices acquires the environmental parameter sequence collected with user authorization, which is the user's current residence space. Based on the locally deployed homomorphic encryption model, the sequence features corresponding to the environmental parameter sequence are encrypted to obtain third encrypted privacy data, and the third encrypted privacy data is sent to the server cluster.

[0025] In this embodiment, taking the application of a local sensing terminal cluster in a private, enclosed environment (such as a user's residence) as an example, the user can perform corresponding actions based on their own preferences or arrangements in this private, enclosed environment, such as reading books, watching TV, using a smartphone to read news, watch videos, or listen to music. At this time, if the user specifically uses a smartphone to perform the relevant operations and is wearing a smartwatch, and one of the multiple smart sensing devices detects that the user's current location data falls within the previously preset location range for the private, enclosed environment, then each of the multiple smart sensing devices can continuously collect parameter sequences for a preset collection duration (such as 1 minute, 5 minutes, 10 minutes, etc.).

[0026] For example, taking a preset collection time of 5 minutes as an example, when a user performs relevant operations using a smartphone and wears a smartwatch, and the user's current location data falls within a preset privacy-protected environment location range, a data collection start point is first generated. The smartphone then acquires the user's authorized user device operation behavior parameter sequence within the time interval [data collection start point, data collection start point + 5 minutes]. Based on a locally deployed homomorphic encryption model, the sequence features corresponding to the user device operation behavior parameter sequence are encrypted to obtain first encrypted privacy data, which is then sent to the server cluster. Similarly, the parameter sequence collection and encryption processes for smartwatches and smart home devices follow the same procedure. After the smartphone acquires the first encrypted privacy data, the smartwatch acquires the second encrypted privacy data, and the smart home device acquires the third encrypted privacy data, they are sent to the server cluster for subsequent encrypted data analysis and processing.

[0027] In one embodiment, step S111 includes: The smartphone among the multiple intelligent sensing devices acquires the user device operation behavior parameter sequence authorized by the user, and extracts features from the user device operation behavior parameter sequence based on the local behavior feature extractor to obtain user behavior features; The first encrypted privacy data is obtained by encrypting the user behavior features based on the homomorphic encryption model.

[0028] In this embodiment, referring to the example above, the smartphone can acquire a sequence of user device operation parameters within the time interval of [data acquisition time start point, data acquisition time start point + 5 minutes], such as [(APP ID1, APP ID1 operation behavior 1, operation time point of APP ID1 operation behavior 1), (APP ID1, APP ID1 operation behavior 2, operation time point of APP ID1 operation behavior 2), ..., (APP IDN, APP IDN operation behavior 1, APPIDN operation behavior 1 operation time point)], etc. A first lightweight classification model (such as an LSTM model, i.e., a Long Short-Term Memory network model) can also be deployed on the smartphone. Then, based on the behavior feature extractor in the LSTM model, the above user device operation parameter sequence is analyzed to obtain user behavior features. This user behavior feature is then encrypted by the smartphone's local homomorphic encryption model to obtain the first encrypted privacy data. Through the above method, the original plaintext data of the smartphone can be kept within the domain, and only the first encrypted privacy data is uploaded, achieving data usability but not visibility while protecting the user's original data from leakage.

[0029] In one embodiment, step S112 includes: The smartwatches among the multiple smart sensing devices acquire the heart rate variability parameter sequence authorized by the user and use it as the user's vital sign parameter sequence. Based on the local physiological feature extractor, the user's vital sign parameter sequence is used to extract features to obtain the user's physiological features. The user's physiological characteristics are classified based on a local lightweight classification model to obtain the current user's stress level parameters. The second encrypted privacy data is obtained by encrypting the current user stress level parameters based on the homomorphic encryption model.

[0030] In this embodiment, referring to the example above, the smartwatch's heart rate sensor can acquire a sequence of heart rate values ​​(multiple heart rate values ​​arranged chronologically) from the user's smartphone within the time interval of [data acquisition time start point, data acquisition time start point + 5 minutes]. Then, relevant calculations are performed based on this heart rate value sequence to convert it into a heart rate variability parameter sequence. A second lightweight classification model (such as an LSTM model) can also be deployed within the smartwatch's processor. The physiological feature extractor in the LSTM model analyzes the aforementioned heart rate variability parameter sequence to obtain the user's physiological characteristics. The classifier in the LSTM model then classifies these physiological characteristics to obtain the current user stress level parameter (which can be represented by an exponential value between 0 and 100). This current user stress level parameter is then encrypted by the smartwatch's local homomorphic encryption model to obtain the second encrypted privacy data. Similarly, through the above method, the smartwatch's original plaintext data can remain within its domain, with only the second encrypted privacy data uploaded, achieving data usability without visibility while protecting the user's original data from leakage.

[0031] In one embodiment, the smart home includes at least smart curtains, and step S113 includes: The smart curtains among the multiple smart sensing devices acquire the current environmental sound intensity sequence within the user's current space based on the built-in sound sensor, which is authorized by the user and serves as the environmental parameter sequence. The environmental feature is then extracted from the environmental parameter sequence using a local environmental feature extractor to obtain environmental features. The third encrypted privacy data is obtained by encrypting the environmental features based on the homomorphic encryption model.

[0032] In this embodiment, referring to the example above, if the smart curtain in the smart home has a built-in sound sensor, it can acquire multiple ambient sound intensities within the user's current space within the time interval of [data acquisition time start point, data acquisition time start point + 5 minutes]. A third lightweight classification model (such as an LSTM model) can also be deployed within the smart curtain's processor. Then, based on the environmental feature extractor in the LSTM model, the aforementioned ambient sound intensity sequence is analyzed to obtain environmental features. This environmental feature is then encrypted by the smart curtain's local homomorphic encryption model to obtain third-encrypted privacy data. Similarly, through the above method, the original plaintext data of the smart curtain can remain within the domain, and only the third-encrypted privacy data can be uploaded, achieving data usability without visibility while protecting the user's original data from leakage.

[0033] S120. The server cluster inputs the encrypted privacy dataset, composed of encrypted privacy data uploaded by the multiple smart sensing devices, into the classification model to obtain the current user classification result.

[0034] In this embodiment, the server cluster may include at least edge servers and cloud servers, and a personalized service recommendation engine is pre-built in the cloud server or edge server. When the edge server or cloud server in the server cluster obtains the encrypted privacy data uploaded by the multiple smart sensing devices, they form a multi-dimensional encrypted privacy dataset and input it into the classification model to obtain the current user classification result.

[0035] In one embodiment, such as Figure 4 As shown, step S120 includes: S121. Obtain the first encrypted privacy data corresponding to the smartphone, the second encrypted privacy data corresponding to the smartwatch, and the third encrypted privacy data corresponding to the smart home from the plurality of smart sensing devices, and form the encrypted privacy dataset by the first encrypted privacy data, the second encrypted privacy data, and the third encrypted privacy data; S122. Securely aggregate the encrypted privacy dataset to obtain encrypted set features, and input them into the classification model to obtain the current user classification result.

[0036] In this embodiment, when the edge server or cloud server in the server cluster obtains the first encrypted privacy data corresponding to the smartphone, the second encrypted privacy data corresponding to the smartwatch, and the third encrypted privacy data corresponding to the smart home, and the first, second, and third encrypted privacy data are combined to form the encrypted privacy dataset, a multi-dimensional encrypted privacy dataset is created. Then, based on the security aggregation strategy in the edge server or cloud server of the server cluster, the data in the encrypted privacy dataset are encrypted and aggregated to obtain encrypted set features, which are then input into the classification model to obtain the current user classification result.

[0037] Specifically, when using a secure aggregation strategy on edge servers or cloud servers in a server cluster to perform secure aggregation on each data point in the encrypted privacy dataset to obtain encrypted set features, the secure aggregation strategy can specifically adopt a differential privacy aggregation strategy. The first encrypted privacy data, the second encrypted privacy data, and the third encrypted privacy data are differentially and securely aggregated using the differential privacy aggregation strategy to obtain encrypted set features, which are then input into the classification model to obtain the current user classification result.

[0038] Furthermore, the classification model deployed in the server cluster can be obtained through federated learning with multiple intelligent sensing devices in the local sensing terminal cluster. The lightweight classification model in each intelligent sensing device can be regarded as a simplified version of the parameters in the classification model deployed in the server cluster. During federated learning, the server cluster first sends the initial model parameters to each intelligent sensing device; each intelligent sensing device trains the model with local data, obtains the local parameter update, and adds noise to the local parameter update accordingly based on Gaussian noise to update the corresponding local parameter update; each intelligent sensing device sends its local parameter update to the server cluster; the server cluster aggregates the local parameter update of each device and calculates the average update; the server cluster then performs a global model update on the classification model based on the average update.

[0039] When using a differential privacy aggregation strategy to perform aggregation calculations (such as summation, mean, count, and median) on the first, second, and third encrypted privacy data in the above example, controllable noise is introduced to ensure that the presence or absence of a single data subject will not significantly affect the final aggregation result, thereby preserving the overall statistical availability of the data while protecting individual privacy.

[0040] S130, The server cluster inputs the current user classification result into the pre-built personalized service recommendation engine to obtain the current service recommendation strategy, and sends the current service recommendation strategy to the multiple intelligent sensing devices.

[0041] In this embodiment, in order to improve data processing efficiency, a personalized service recommendation engine can be built in a distributed processing system composed of multiple servers in the server cluster. By inputting the current user classification result corresponding to a certain local sensing terminal cluster into the personalized service recommendation engine, the current service recommendation strategy is obtained, and the current service recommendation strategy is sent to multiple intelligent sensing devices in the local sensing terminal cluster.

[0042] In one embodiment, such as Figure 5 As shown, step S130 includes: S131. Obtain multiple server recommendation strategies in the personalized service recommendation engine, and the user classification result identifier value corresponding to each server recommendation strategy; S132. If the user classification result identifier value corresponding to a server recommendation strategy is the same as the current user classification result, then the corresponding server recommendation strategy is obtained as the current service recommendation strategy.

[0043] In this embodiment, taking the personalized service recommendation engine in one of the servers of the aforementioned distributed processing system as an example, the process first involves obtaining multiple server recommendation strategies included in the personalized service recommendation engine, and the user classification result identifier value corresponding to each server recommendation strategy. Next, it is determined whether any server recommendation strategy has a user classification result identifier value that is the same as the current user classification result. If it is determined that a server recommendation strategy has a user classification result identifier value that is the same as the current user classification result, then the corresponding server recommendation strategy is obtained as the current service recommendation strategy. Through the matching process between the user classification result identifier value and the current user classification result, the current service recommendation strategy can be quickly obtained from the personalized service recommendation engine.

[0044] S140. Each of the plurality of intelligent sensing devices obtains the corresponding current service recommendation sub-strategy from the current service recommendation strategy and executes the current service recommendation sub-strategy to perform device action control.

[0045] In this embodiment, after the server cluster sends the current service recommendation policy to multiple intelligent sensing devices in the local sensing terminal cluster, each intelligent sensing device obtains the corresponding current service recommendation sub-policy from the current service recommendation policy and executes the current service recommendation sub-policy to control device actions, such as smartphones displaying relevant content, smartwatches providing both vibration and message notifications, and smart home devices performing corresponding operations to reduce environmental noise. It is evident that through the above method, the server cluster can quickly feed back the current service recommendation sub-policy to each of the multiple intelligent sensing devices, and execute the current service recommendation sub-policy to provide timely responses and feedback to user operations within the current private and enclosed environment, thereby achieving intelligent interaction.

[0046] In one embodiment, such as Figure 6 As shown, step S140 includes: S141. The smartphone among the plurality of intelligent sensing devices obtains a first current service recommendation sub-strategy from the current service recommendation strategy, and obtains corresponding recommendation learning data according to the first current service recommendation sub-strategy and displays it locally. S142. The smartwatch among the plurality of smart sensing devices obtains a second current service recommendation sub-strategy from the current service recommendation strategy, and obtains corresponding recommendation behavior data according to the second current service recommendation sub-strategy and provides local prompts; S143. The smart home devices among the plurality of smart sensing devices obtain a third current service recommendation sub-strategy from the current service recommendation strategy, and obtain the corresponding recommendation execution action according to the third current service recommendation sub-strategy and execute the corresponding operation.

[0047] In this embodiment, a smartphone can obtain a first current service recommendation sub-strategy from the current service recommendation strategy, and acquire corresponding recommendation learning data (such as text data, audio data, video data, image data, etc.) based on the first current service recommendation sub-strategy and display it locally. A smartwatch can obtain a second current service recommendation sub-strategy from the current service recommendation strategy, and acquire corresponding recommendation behavior data (such as taking a deep breath for 10 seconds, stopping the current user operation and resting for 30 seconds, etc.) based on the second current service recommendation sub-strategy and provide local prompts. In a smart home, a third current service recommendation sub-strategy can be obtained from the current service recommendation strategy, and acquire corresponding recommendation execution actions (such as turning on the indoor noise reduction function of smart curtains, etc.) based on the third current service recommendation sub-strategy and execute the corresponding operation. It can be seen that by executing the corresponding current service recommendation sub-strategies through the above-mentioned smart sensing devices, a timely response to the current service recommendation strategy sent by the server cluster is achieved.

[0048] In one embodiment, the method further includes the following after step S140: The smartphones or smartwatches among the multiple smart sensing devices send rating and evaluation information for the current service recommendation strategy to the server cluster.

[0049] In this embodiment, after the current round of privacy-protected intelligent service strategy has completed the rapid push from the server cluster to multiple intelligent sensing devices and the execution of the current service recommendation sub-strategy locally on the multiple intelligent sensing devices, the user can rate and evaluate the environmental changes and user experience caused by the current service recommendation strategy in the user's privacy-protected environment, and obtain rating and evaluation information. Then, the average user evaluation score for each service recommendation strategy in the personalized service recommendation engine can be obtained, and high-rated service recommendation strategies can be selected based on the average user evaluation score (e.g., selecting the top 3, top 5, or top 10 service recommendation strategies based on the average user evaluation score).

[0050] As can be seen, the implementation of this method enables each intelligent sensing device to send the current sensing data to the server cluster in an encrypted privacy manner, and perform encrypted data classification to obtain the current user classification result. Based on the personalized service recommendation engine, the current service recommendation strategy corresponding to the current user classification result is obtained and sent to each intelligent sensing device in a timely manner, avoiding the risk of leakage of the original sensing data. Furthermore, the personalized service recommendation engine in the server cluster is dedicated to the data processing of service recommendation strategies, which improves the data processing efficiency.

[0051] This application also provides a privacy-preserving intelligent service policy push system, such as... Figure 1 andFigure 7 As shown, the privacy-preserving intelligent service policy push method provided in this embodiment of the invention specifically applies a privacy-preserving intelligent service policy push system, which includes a local sensing terminal cluster 10 and a server cluster 20. The local sensing terminal cluster 10 is communicatively connected to the server cluster 20. The local sensing terminal cluster 10 includes multiple intelligent sensing devices, and all of the multiple intelligent sensing devices are communicatively connected to the server cluster 20.

[0052] The plurality of intelligent sensing devices are used to acquire the current sensing data authorized by the user according to the corresponding preset data dimension among a plurality of preset data dimensions, and send the encrypted privacy data corresponding to each current sensing data to the server cluster.

[0053] In this embodiment, the multiple intelligent sensing devices serve as data acquisition terminals. Each intelligent sensing device acquires the current sensing data authorized by the user in a corresponding preset data dimension, encrypts it locally, and then uploads it to the server cluster. This avoids sending the current sensing data of each intelligent sensing device to the server cluster for data processing in plaintext.

[0054] In one embodiment, such as Figure 7 As shown, the local sensing terminal cluster 10 includes at least a smartphone 11, a smartwatch 12, and a smart home device 13 among its multiple intelligent sensing devices; The smartphone 11 among the multiple intelligent sensing devices is used to acquire the user device operation behavior parameter sequence authorized by the user, encrypt the sequence features corresponding to the user device operation behavior parameter sequence based on the locally deployed homomorphic encryption model to obtain the first encrypted privacy data, and send the first encrypted privacy data to the server cluster. The smartwatch 12 among the multiple smart sensing devices is used to acquire the user's authorized collection of user vital sign parameter sequences, encrypt the sequence features corresponding to the user's vital sign parameter sequences based on a locally deployed homomorphic encryption model to obtain second encrypted privacy data, and send the second encrypted privacy data to the server cluster. The smart home device 13 among the multiple smart sensing devices is used to acquire the environmental parameter sequence that is authorized by the user and is currently in the user's space, encrypt the sequence features corresponding to the environmental parameter sequence based on the locally deployed homomorphic encryption model to obtain third encrypted privacy data, and send the third encrypted privacy data to the server cluster.

[0055] In this embodiment, taking the application of a local sensing terminal cluster in a private, enclosed environment (such as a user's residence) as an example, the user can perform corresponding actions based on their own preferences or arrangements in this private, enclosed environment, such as reading books, watching TV, using a smartphone to read news, watch videos, or listen to music. At this time, if the user specifically uses a smartphone to perform the relevant operations and is wearing a smartwatch, and one of the multiple smart sensing devices detects that the user's current location data falls within the previously preset location range for the private, enclosed environment, then each of the multiple smart sensing devices can continuously collect parameter sequences for a preset collection duration (such as 1 minute, 5 minutes, 10 minutes, etc.).

[0056] For example, taking a preset collection time of 5 minutes as an example, when a user performs relevant operations using a smartphone and wears a smartwatch, and the user's current location data falls within a preset privacy-protected environment location range, a data collection start point is first generated. The smartphone then acquires the user's authorized user device operation behavior parameter sequence within the time interval [data collection start point, data collection start point + 5 minutes]. Based on a locally deployed homomorphic encryption model, the sequence features corresponding to the user device operation behavior parameter sequence are encrypted to obtain first encrypted privacy data, which is then sent to the server cluster. Similarly, the parameter sequence collection and encryption processes for smartwatches and smart home devices follow the same procedure. After the smartphone acquires the first encrypted privacy data, the smartwatch acquires the second encrypted privacy data, and the smart home device acquires the third encrypted privacy data, they are sent to the server cluster for subsequent encrypted data analysis and processing.

[0057] In one embodiment, the smartphone 11 is specifically used for: Obtain the sequence of user device operation behavior parameters authorized by the user, and extract features from the sequence of user device operation behavior parameters based on the local behavior feature extractor to obtain user behavior features; The first encrypted privacy data is obtained by encrypting the user behavior features based on the homomorphic encryption model.

[0058] In this embodiment, referring to the example above, the smartphone can acquire a sequence of user device operation parameters within the time interval of [data acquisition time start point, data acquisition time start point + 5 minutes], such as [(APP ID1, APP ID1 operation behavior 1, operation time point of APP ID1 operation behavior 1), (APP ID1, APP ID1 operation behavior 2, operation time point of APP ID1 operation behavior 2), ..., (APP IDN, APP IDN operation behavior 1, APPIDN operation behavior 1 operation time point)], etc. A first lightweight classification model (such as an LSTM model, i.e., a Long Short-Term Memory network model) can also be deployed on the smartphone. Then, based on the behavior feature extractor in the LSTM model, the above user device operation parameter sequence is analyzed to obtain user behavior features. This user behavior feature is then encrypted by the smartphone's local homomorphic encryption model to obtain the first encrypted privacy data. Through the above method, the original plaintext data of the smartphone can be kept within the domain, and only the first encrypted privacy data is uploaded, achieving data usability but not visibility while protecting the user's original data from leakage.

[0059] In one embodiment, the smartwatch 12 is specifically used for: The heart rate variability parameter sequence authorized by the user is obtained and used as the user's vital sign parameter sequence. Based on the local physiological feature extractor, the user's vital sign parameter sequence is used to extract features to obtain the user's physiological features. The user's physiological characteristics are classified based on a local lightweight classification model to obtain the current user's stress level parameters. The second encrypted privacy data is obtained by encrypting the current user stress level parameters based on the homomorphic encryption model.

[0060] In this embodiment, referring to the example above, the smartwatch's heart rate sensor can acquire a sequence of heart rate values ​​(multiple heart rate values ​​arranged chronologically) from the user's smartphone within the time interval of [data acquisition time start point, data acquisition time start point + 5 minutes]. Then, relevant calculations are performed based on this heart rate value sequence to convert it into a heart rate variability parameter sequence. A second lightweight classification model (such as an LSTM model) can also be deployed within the smartwatch's processor. The physiological feature extractor in the LSTM model analyzes the aforementioned heart rate variability parameter sequence to obtain the user's physiological characteristics. The classifier in the LSTM model then classifies these physiological characteristics to obtain the current user stress level parameter (which can be represented by an exponential value between 0 and 100). This current user stress level parameter is then encrypted by the smartwatch's local homomorphic encryption model to obtain the second encrypted privacy data. Similarly, through the above method, the smartwatch's original plaintext data can remain within its domain, with only the second encrypted privacy data uploaded, achieving data usability without visibility while protecting the user's original data from leakage.

[0061] In one embodiment, the smart home 13 includes at least a smart curtain, which is specifically used for: The system acquires the current ambient sound intensity sequence within the user's current location space based on the built-in sound sensor, and uses this sequence as the environmental parameter sequence. The system then extracts features from the environmental parameter sequence using a local environmental feature extractor to obtain the environmental features. The third encrypted privacy data is obtained by encrypting the environmental features based on the homomorphic encryption model.

[0062] In this embodiment, referring to the example above, if the smart curtain in the smart home has a built-in sound sensor, it can acquire multiple ambient sound intensities within the user's current space within the time interval of [data acquisition time start point, data acquisition time start point + 5 minutes]. A third lightweight classification model (such as an LSTM model) can also be deployed within the smart curtain's processor. Then, based on the environmental feature extractor in the LSTM model, the aforementioned ambient sound intensity sequence is analyzed to obtain environmental features. This environmental feature is then encrypted by the smart curtain's local homomorphic encryption model to obtain third-encrypted privacy data. Similarly, through the above method, the original plaintext data of the smart curtain can remain within the domain, and only the third-encrypted privacy data can be uploaded, achieving data usability without visibility while protecting the user's original data from leakage.

[0063] The server cluster 20 is used to input the encrypted privacy dataset, composed of encrypted privacy data uploaded by the multiple smart sensing devices, into the classification model to obtain the current user classification result.

[0064] In this embodiment, the server cluster may include at least edge servers and cloud servers, and a personalized service recommendation engine is pre-built in the cloud server or edge server. When the edge server or cloud server in the server cluster obtains the encrypted privacy data uploaded by the multiple smart sensing devices, they form a multi-dimensional encrypted privacy dataset and input it into the classification model to obtain the current user classification result.

[0065] In one embodiment, the server cluster 20 is specifically used for: The encrypted privacy dataset is obtained from the plurality of smart sensing devices, including the first encrypted privacy data corresponding to the smartphone, the second encrypted privacy data corresponding to the smartwatch, and the third encrypted privacy data corresponding to the smart home. The encrypted privacy dataset is securely aggregated to obtain encrypted set features, which are then input into the classification model to obtain the current user classification result.

[0066] In this embodiment, when the edge server or cloud server in the server cluster obtains the first encrypted privacy data corresponding to the smartphone, the second encrypted privacy data corresponding to the smartwatch, and the third encrypted privacy data corresponding to the smart home, and the first, second, and third encrypted privacy data are combined to form the encrypted privacy dataset, a multi-dimensional encrypted privacy dataset is created. Then, based on the security aggregation strategy in the edge server or cloud server of the server cluster, the data in the encrypted privacy dataset are encrypted and aggregated to obtain encrypted set features, which are then input into the classification model to obtain the current user classification result.

[0067] Specifically, when using a secure aggregation strategy on edge servers or cloud servers in a server cluster to perform secure aggregation on each data point in the encrypted privacy dataset to obtain encrypted set features, the secure aggregation strategy can specifically adopt a differential privacy aggregation strategy. The first encrypted privacy data, the second encrypted privacy data, and the third encrypted privacy data are differentially and securely aggregated using the differential privacy aggregation strategy to obtain encrypted set features, which are then input into the classification model to obtain the current user classification result.

[0068] Furthermore, the classification model deployed in the server cluster can be obtained through federated learning with multiple intelligent sensing devices in the local sensing terminal cluster. The lightweight classification model in each intelligent sensing device can be regarded as a simplified version of the parameters in the classification model deployed in the server cluster. During federated learning, the server cluster first sends the initial model parameters to each intelligent sensing device; each intelligent sensing device trains the model with local data, obtains the local parameter update, and adds noise to the local parameter update accordingly based on Gaussian noise to update the corresponding local parameter update; each intelligent sensing device sends its local parameter update to the server cluster; the server cluster aggregates the local parameter update of each device and calculates the average update; the server cluster then performs a global model update on the classification model based on the average update.

[0069] When using a differential privacy aggregation strategy to perform aggregation calculations (such as summation, mean, count, and median) on the first, second, and third encrypted privacy data in the above example, controllable noise is introduced to ensure that the presence or absence of a single data subject will not significantly affect the final aggregation result, thereby preserving the overall statistical availability of the data while protecting individual privacy.

[0070] The server cluster 20 is also used to input the current user classification result into a pre-built personalized service recommendation engine to obtain a current service recommendation strategy, and send the current service recommendation strategy to the multiple intelligent sensing devices.

[0071] In this embodiment, in order to improve data processing efficiency, a personalized service recommendation engine can be built in a distributed processing system composed of multiple servers in the server cluster. By inputting the current user classification result corresponding to a certain local sensing terminal cluster into the personalized service recommendation engine, the current service recommendation strategy is obtained, and the current service recommendation strategy is sent to multiple intelligent sensing devices in the local sensing terminal cluster.

[0072] In one embodiment, the server cluster 20 is further specifically used for: Obtain multiple server recommendation strategies in the personalized service recommendation engine, and the user classification result identifier value corresponding to each server recommendation strategy; If a user classification result identifier value corresponding to a server recommendation strategy is the same as the current user classification result, then the corresponding server recommendation strategy is obtained as the current service recommendation strategy.

[0073] In this embodiment, taking the personalized service recommendation engine in one of the servers of the aforementioned distributed processing system as an example, the process first involves obtaining multiple server recommendation strategies included in the personalized service recommendation engine, and the user classification result identifier value corresponding to each server recommendation strategy. Next, it is determined whether any server recommendation strategy has a user classification result identifier value that is the same as the current user classification result. If it is determined that a server recommendation strategy has a user classification result identifier value that is the same as the current user classification result, then the corresponding server recommendation strategy is obtained as the current service recommendation strategy. Through the matching process between the user classification result identifier value and the current user classification result, the current service recommendation strategy can be quickly obtained from the personalized service recommendation engine.

[0074] Each of the plurality of intelligent sensing devices is used to obtain the corresponding current service recommendation sub-strategy from the current service recommendation strategy, and execute the current service recommendation sub-strategy to perform device action control.

[0075] In this embodiment, after the server cluster sends the current service recommendation policy to multiple intelligent sensing devices in the local sensing terminal cluster, each intelligent sensing device obtains the corresponding current service recommendation sub-policy from the current service recommendation policy and executes the current service recommendation sub-policy to control device actions, such as smartphones displaying relevant content, smartwatches providing both vibration and message notifications, and smart home devices performing corresponding operations to reduce environmental noise. It is evident that through the above method, the server cluster can quickly feed back the current service recommendation sub-policy to each of the multiple intelligent sensing devices, and execute the current service recommendation sub-policy to provide timely responses and feedback to user operations within the current private and enclosed environment, thereby achieving intelligent interaction.

[0076] In one embodiment, among the plurality of intelligent sensing devices: The smartphones among the multiple intelligent sensing devices obtain a first current service recommendation sub-strategy from the current service recommendation strategy, and obtain corresponding recommendation learning data based on the first current service recommendation sub-strategy and display it locally; The smartwatches among the multiple smart sensing devices obtain a second current service recommendation sub-strategy from the current service recommendation strategy, and obtain corresponding recommendation behavior data according to the second current service recommendation sub-strategy and provide local prompts; The smart home devices among the multiple smart sensing devices obtain a third current service recommendation sub-strategy from the current service recommendation strategy, and obtain the corresponding recommended execution action according to the third current service recommendation sub-strategy and perform the corresponding operation accordingly.

[0077] In this embodiment, a smartphone can obtain a first current service recommendation sub-strategy from the current service recommendation strategy, and acquire corresponding recommendation learning data (such as text data, audio data, video data, image data, etc.) based on the first current service recommendation sub-strategy and display it locally. A smartwatch can obtain a second current service recommendation sub-strategy from the current service recommendation strategy, and acquire corresponding recommendation behavior data (such as taking a deep breath for 10 seconds, stopping the current user operation and resting for 30 seconds, etc.) based on the second current service recommendation sub-strategy and provide local prompts. In a smart home, a third current service recommendation sub-strategy can be obtained from the current service recommendation strategy, and acquire corresponding recommendation execution actions (such as turning on the indoor noise reduction function of smart curtains, etc.) based on the third current service recommendation sub-strategy and execute the corresponding operation. It can be seen that by executing the corresponding current service recommendation sub-strategies through the above-mentioned smart sensing devices, a timely response to the current service recommendation strategy sent by the server cluster is achieved.

[0078] In one embodiment, it is also used for: The smartphones or smartwatches among the multiple smart sensing devices send rating and evaluation information for the current service recommendation strategy to the server cluster.

[0079] In this embodiment, after the current round of privacy-protected intelligent service strategy has completed the rapid push from the server cluster to multiple intelligent sensing devices and the execution of the current service recommendation sub-strategy locally on the multiple intelligent sensing devices, the user can rate and evaluate the environmental changes and user experience caused by the current service recommendation strategy in the user's privacy-protected environment, and obtain rating and evaluation information. Then, the average user evaluation score for each service recommendation strategy in the personalized service recommendation engine can be obtained, and high-rated service recommendation strategies can be selected based on the average user evaluation score (e.g., selecting the top 3, top 5, or top 10 service recommendation strategies based on the average user evaluation score).

[0080] As can be seen, the implementation of this system enables each intelligent sensing device to send the current sensing data to the server cluster in an encrypted and privacy-preserving manner, and to perform encrypted data classification to obtain the current user classification result. Based on the personalized service recommendation engine, the system obtains the current service recommendation strategy corresponding to the current user classification result and sends it to each intelligent sensing device in a timely manner. This avoids the risk of leakage of the original sensing data and improves data processing efficiency by using the personalized service recommendation engine in the server cluster to process the service recommendation strategy data.

[0081] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0082] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0083] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0084] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A privacy protection based intelligent service policy push method, characterized in that, The local perception terminal cluster comprises a plurality of intelligent perception devices, and the plurality of intelligent perception devices are in communication connection with the server cluster; The plurality of intelligent perception devices obtain current perception data collected by user authorization in a corresponding preset data dimension of a plurality of preset data dimensions, and send encrypted privacy data corresponding to each current perception data to the server cluster; The server cluster inputs an encrypted privacy data set composed of the encrypted privacy data uploaded by the plurality of intelligent perception devices into a classification model to obtain a current user classification result; The server cluster inputs the current user classification result into a pre-constructed personalized service recommendation engine to obtain a current service recommendation strategy, and sends the current service recommendation strategy to the plurality of intelligent perception devices; Each intelligent perception device in the plurality of intelligent perception devices obtains a corresponding current service recommendation sub-strategy from the current service recommendation strategy, and executes the current service recommendation sub-strategy to control device action. 2.The privacy protection based intelligent service policy pushing method according to claim 1, characterized in that, The plurality of intelligent perception devices at least include a smartphone, a smartwatch, and a smart home; The plurality of intelligent perception devices obtain current perception data collected by user authorization in a corresponding preset data dimension of a plurality of preset data dimensions, and send encrypted privacy data corresponding to each current perception data to the server cluster, comprising: The smartphone in the plurality of intelligent perception devices obtains a user device operation behavior parameter sequence collected by user authorization, encrypts sequence features corresponding to the user device operation behavior parameter sequence based on a locally deployed homomorphic encryption model to obtain first encrypted privacy data, and sends the first encrypted privacy data to the server cluster; The smartwatch in the plurality of intelligent perception devices obtains a user vital sign parameter sequence collected by user authorization, encrypts sequence features corresponding to the user vital sign parameter sequence based on a locally deployed homomorphic encryption model to obtain second encrypted privacy data, and sends the second encrypted privacy data to the server cluster; The smart home in the plurality of intelligent perception devices obtains an environment parameter sequence collected by user authorization and in a current stay space of the user, encrypts sequence features corresponding to the environment parameter sequence based on a locally deployed homomorphic encryption model to obtain third encrypted privacy data, and sends the third encrypted privacy data to the server cluster. 3.The privacy protection based intelligent service policy pushing method according to claim 2, characterized in that, The smartphone in the plurality of intelligent perception devices obtains a user device operation behavior parameter sequence collected by user authorization, encrypts sequence features corresponding to the user device operation behavior parameter sequence based on a locally deployed homomorphic encryption model to obtain first encrypted privacy data; The smartphone in the plurality of intelligent perception devices obtains a user device operation behavior parameter sequence collected by user authorization, extracts features of the user device operation behavior parameter sequence based on a local behavior feature extractor to obtain user behavior features; The user behavior features are encrypted based on the homomorphic encryption model to obtain the first encrypted privacy data. 4.The privacy protection based intelligent service policy pushing method according to claim 2, characterized in that, The smart watch in the plurality of intelligent sensing devices acquires a user-authorized collected user vital parameter sequence, and encrypts sequence features corresponding to the user vital parameter sequence based on a locally deployed homomorphic encryption model to obtain second encrypted privacy data, including: The smart watch in the plurality of intelligent sensing devices acquires a user-authorized collected heart rate variability parameter sequence as the user vital parameter sequence, and extracts features of the user vital parameter sequence based on a local physiological feature extractor to obtain user physiological features; Classify the user physiological features based on a local lightweight classification model to obtain a current user stress classification parameter; Encrypt the current user stress classification parameter based on the homomorphic encryption model to obtain the second encrypted privacy data. 5.The privacy protection based intelligent service policy pushing method according to claim 2, characterized in that, The smart home at least includes a smart curtain; the smart home in the plurality of intelligent sensing devices acquires an environment parameter sequence authorized by the user to be collected and in the current stay space of the user, and encrypts sequence features corresponding to the environment parameter sequence based on a locally deployed homomorphic encryption model to obtain third encrypted privacy data, including: The smart curtain in the plurality of intelligent sensing devices acquires a current environment sound intensity sequence authorized by the user to be collected and in the current stay space of the user based on a built-in acoustic sensor, and takes the current environment sound intensity sequence as the environment parameter sequence, and extracts features of the environment parameter sequence based on a local environment feature extractor to obtain environment features; Encrypt the environment features based on the homomorphic encryption model to obtain the third encrypted privacy data. 6.The privacy protection based intelligent service policy pushing method according to claim 2, characterized in that, The server cluster inputs an encrypted privacy data set composed of encrypted privacy data uploaded by the plurality of intelligent sensing devices respectively into a classification model to obtain a current user classification result, including: Obtain first encrypted privacy data corresponding to the smart phone, second encrypted privacy data corresponding to the smart watch, and third encrypted privacy data corresponding to the smart home in the plurality of intelligent sensing devices, and compose the encrypted privacy data set from the first encrypted privacy data, the second encrypted privacy data, and the third encrypted privacy data; Securely aggregate the encrypted privacy data set to obtain an encrypted set feature, and input the encrypted set feature into the classification model to obtain the current user classification result. 7.The privacy protection based intelligent service policy pushing method according to claim 1, characterized in that, The server cluster inputs the current user classification result into a pre-constructed personalized service recommendation engine to obtain a current service recommendation strategy, including: Obtain a plurality of server recommendation strategies in the personalized service recommendation engine, and user classification result identifier values corresponding to each server recommendation strategy; If a user classification result identifier value corresponding to a server recommendation strategy is the same as the current user classification result, obtain the corresponding server recommendation strategy as the current service recommendation strategy. 8.The privacy protection based intelligent service policy pushing method according to claim 2, characterized in that, Each intelligent sensing device in the plurality of intelligent sensing devices acquires a corresponding current service recommendation sub-strategy from the current service recommendation strategy, and executes the current service recommendation sub-strategy to control device actions, including: The smartphone in the plurality of intelligent sensing devices acquires a first current service recommendation sub-strategy from the current service recommendation strategy, and acquires corresponding recommendation learning data and performs local display according to the first current service recommendation sub-strategy; The smart watch in the plurality of intelligent sensing devices acquires a second current service recommendation sub-strategy from the current service recommendation strategy, and acquires corresponding recommendation behavior data and performs local prompting according to the second current service recommendation sub-strategy; The smart home in the plurality of intelligent sensing devices acquires a third current service recommendation sub-strategy from the current service recommendation strategy, and acquires corresponding recommendation execution action and performs corresponding operation according to the third current service recommendation sub-strategy. 9.The privacy protection based intelligent service policy pushing method according to claim 1, characterized in that, After the step of each intelligent sensing device in the plurality of intelligent sensing devices acquiring corresponding current service recommendation sub-strategy from the current service recommendation strategy and performing the current service recommendation sub-strategy to control device action, the method further comprises: The smartphone or smart watch in the plurality of intelligent sensing devices sends score evaluation information for the current service recommendation strategy to the server cluster.

10. A privacy protection based intelligent service policy push system, characterized in that, The system comprises a local sensing terminal cluster and a server cluster, the local sensing terminal cluster comprises a plurality of intelligent sensing devices, and the plurality of intelligent sensing devices are in communication connection with the server cluster; the privacy protection based intelligent service strategy pushing system is used to execute the privacy protection based intelligent service strategy pushing method according to any one of claims 1-9.

Images