Multi-user joint steady-state distribution tensor determination method and related equipment

Abstract

The invention provides a multi-user joint steady-state distribution tensor determination method and related equipment, which effectively protect the privacy of data. The method comprises the following steps: constructing a user frequency tensor for each user in a plurality of users, and generating a noise tensor; determining a synthetic tensor with a differential privacy protection characteristic according to the noise tensor and the user frequency tensor; performing normalization processing on the user frequency tensor and the synthetic tensor; constructing a DP-2M Markov model based on the normalized user frequency tensor and the synthesized tensor, and solving the DP-2M Markov model to obtain an SJE of each user and a DP-SJE of each user; determining a space-time distance and a semantic distance between the SJE and the DP-SJE; the DP-SJE of each user is optimized according to the space-time distance and the semantic distance; and sending the synthetic tensor and the optimized DP-SJE to a cloud device, so that the cloud device determines a user influence coefficient according to the synthetic tensor, and determines a multi-user DP-SJE according to the user influence coefficient and the optimized DP-SJE.

Description

Technical Field

[0001] This invention relates to the field of the Internet of Things, and in particular to a method and related equipment for determining a multi-user joint steady-state distribution tensor. Background Technology

[0002] With the rapid development of Internet of Things (IoT) devices and their increasing reliance on cloud computing, mobile devices are now continuously and massively collecting user trajectory data. This data not only contains rich personal information but also exhibits characteristics of being multi-source, heterogeneous, and highly complex. This trajectory data reflects users' geographical location and movement patterns, and may even involve sensitive information such as behavioral habits and social relationships. In intelligent transportation systems, utilizing edge computing and cloud computing to collaboratively predict user trajectory patterns has become an important means of improving traffic efficiency and providing personalized services. However, when transmitting this sensitive trajectory data from edge devices to the cloud, there is a risk of it being intercepted, stolen, or tampered with by malicious attackers, which could lead to serious breaches of user privacy. Therefore, how to effectively protect user privacy during data transmission and processing while ensuring the accuracy of trajectory prediction has become a critical challenge that urgently needs to be addressed.

[0003] As data is transmitted from edge devices to the cloud, the potential risk of data breaches is becoming increasingly severe. This risk stems not only from possible interception and theft during transmission but also from security vulnerabilities and internal abuse during cloud storage and processing. While traditional privacy protection measures, such as asymmetric encryption, k-anonymization, generalization, anonymization, data suppression, and federated learning, offer some level of data security, they face numerous challenges in practical applications. First, these methods typically involve significant computational overhead, especially asymmetric encryption, whose complex encryption and decryption processes place high demands on device processing power and energy consumption. Second, these methods are inefficient and struggle to meet the demands of real-time data processing and high-frequency communication. Furthermore, many traditional methods rely on assumptions about the attacker's background knowledge, such as their understanding of the data or their attack capabilities. However, these assumptions often fail when facing complex and ever-changing attack methods, leading to reduced privacy protection effectiveness. Finally, in large-scale data environments, the scalability and applicability of these methods are limited, making it difficult to process massive amounts of data in a timely manner and thus failing to meet the data transmission and analysis needs of the Internet of Things era. Therefore, achieving high efficiency while effectively protecting the privacy of large-scale data has become a significant challenge. Summary of the Invention

[0004] This invention provides a method and related equipment for determining a multi-user joint steady-state distribution tensor, ensuring privacy when data is transmitted to the cloud. In the cloud, a user influence coefficient is constructed based on a single-user synthetic tensor, and a multi-purpose DP-SJE is generated based on the user influence coefficient and the single-user DP-SJE, effectively protecting data privacy.

[0005] The first aspect of this invention provides a method for determining a multi-user joint steady-state distribution tensor, comprising: The edge device constructs a user frequency tensor for each of the multiple users and generates a noise tensor corresponding to the user frequency tensor; The edge device determines a synthetic tensor with differential privacy protection features based on the noise tensor and the user frequency tensor. The edge device normalizes the user frequency tensor and the composite tensor. The edge device constructs a DP-2M Markov model based on the normalized user frequency tensor and the synthetic tensor, and solves the DP-2M Markov model to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features. The edge device determines the spatiotemporal distance and semantic distance between each user's SJE and each user's DP-SJE; The edge device optimizes the DP-SJE of each user based on the spatiotemporal distance and the semantic distance; The edge device sends the synthesized tensor and the optimized DP-SJE to the cloud device, so that the cloud device determines the user influence coefficient corresponding to each user based on the synthesized tensor, and determines the multi-user DP-SJE based on the user influence coefficient and the optimized DP-SJE of each user.

[0006] A second aspect of the present invention provides a method for determining a multi-user joint steady-state distribution tensor, comprising: The cloud device receives a synthetic tensor and an optimized joint steady-state distribution tensor (DP-SJE) with privacy protection features for each user, sent by the edge device. The synthetic tensor is determined by the edge device based on the noise tensor and the user frequency tensor of each user. The optimized DP-SJE for each user is obtained by the edge device by optimizing the DP-SJE of each user based on spatiotemporal distance and semantic distance, where the spatiotemporal distance and semantic distance are the spatiotemporal distance and semantic distance between the SJE and the DP-SJE of each user. The SP-SJE for each user is obtained by the edge device constructing a DP-2M Markov model based on the normalized user frequency tensor and the normalized synthetic tensor, and solving the DP-2M Markov model. The cloud device determines the user influence coefficient corresponding to each user based on the synthetic tensor. The cloud device determines the multi-user DP-SJE based on the optimized DP-SJE of each user and the normalized user influence coefficient.

[0007] A third aspect of the present invention provides an edge device comprising: A construction module is used to construct a user frequency tensor for each of the multiple users and generate a noise tensor corresponding to the user frequency tensor; The first determining module is used to determine a synthetic tensor with differential privacy protection characteristics based on the noise tensor and the user frequency tensor. The normalization processing module is used to normalize the user frequency tensor and the synthetic tensor. The processing module is used to construct a DP-2M Markov model based on the normalized user frequency tensor and the synthetic tensor, and solve the DP-2M Markov model to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features. The second determining module is used to determine the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user; The optimization module is used to optimize the DP-SJE of each user based on the spatiotemporal distance and the semantic distance; The sending module is used to send the synthesized tensor and the optimized DP-SJE to the cloud device, so that the cloud device can determine the user influence coefficient corresponding to each user based on the synthesized tensor, and determine the multi-user DP-SJE based on the user influence coefficient and the optimized DP-SJE of each user.

[0008] A fourth aspect of the present invention provides a cloud device, comprising: The receiving module is used to receive the synthetic tensor and the optimized joint steady-state distribution tensor (DP-SJE) with privacy protection features for each user sent by the edge device. The synthetic tensor is determined by the edge device based on the noise tensor and the user frequency tensor of each user. The optimized DP-SJE for each user is obtained by the edge device by optimizing the DP-SJE of each user based on the spatiotemporal distance and the semantic distance. The spatiotemporal distance and the semantic distance are the spatiotemporal distance and the semantic distance between the SJE of each user and the DP-SJE of each user. The SP-SJE for each user is obtained by the edge device by constructing a DP-2M Markov model based on the normalized user frequency tensor and the normalized synthetic tensor, and solving the DP-2M Markov model. The first determining module is used to determine the user influence coefficient corresponding to each user based on the synthesized tensor; The second determining module is used to determine the multi-user DP-SJE based on the optimized DP-SJE of each user and the normalized user influence coefficient.

[0009] The fifth aspect of the present invention provides an electronic device, including a memory and a processor, wherein the processor is configured to execute a computer management program stored in the memory to implement the steps of the method for determining a multi-user joint steady-state distribution tensor as described in any of the preceding aspects.

[0010] A sixth aspect of the present invention provides a computer-readable storage medium having a computer management program stored thereon, which, when executed by a processor, performs the steps of determining the multi-user joint steady-state distribution tensor as described in any of the preceding aspects.

[0011] In summary, it can be seen that the embodiments provided by this invention balance the relationship between data privacy and prediction accuracy by introducing two measurement methods: tensor-based spatiotemporal distance and tensor-based semantic distance. A DP-2M Markov model is implemented at the edge to generate single-user SJE and single-user DP-SJE, ensuring their privacy when transmitted to the cloud. At the same time, a user influence coefficient is constructed in the cloud based on the single-user synthetic tensor, and a multi-purpose DP-SJE is generated based on the user influence coefficient and the single-user DP-SJE, effectively protecting data privacy. Attached Figure Description

[0012] Figure 1 A flowchart illustrating a method for determining a multi-user joint steady-state distribution tensor provided in an embodiment of the present invention; Figure 2A schematic diagram of DP-2M provided in an embodiment of the present invention; Figure 3 This is another flowchart illustrating the method for determining the multi-user joint steady-state distribution tensor provided in an embodiment of the present invention; Figure 4 A schematic diagram of DP-3M provided in an embodiment of the present invention; Figure 5 This is another flowchart illustrating the method for determining the multi-user joint steady-state distribution tensor provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the virtual structure of an edge device provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the virtual structure of a cloud device provided in an embodiment of the present invention; Figure 8 A schematic diagram of the hardware structure of the device for determining the multi-user joint steady-state distribution tensor provided in an embodiment of the present invention; Figure 9 A schematic diagram of an embodiment of the electronic device provided in this invention; Figure 10 A schematic diagram illustrating an embodiment of a computer-readable storage medium provided in this invention. Detailed Implementation

[0013] 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 embodiments of the present invention, and not all embodiments. 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.

[0014] The following section explains the method for determining the joint steady-state distribution tensor of multiple users from the perspective of edge devices. Please refer to [link / reference needed]. Figure 1 , Figure 1 A flowchart illustrating the method for determining the joint steady-state distribution tensor of multiple users provided in this embodiment of the invention includes: 101. The edge device constructs a user frequency tensor for each of the multiple users and generates a noise tensor corresponding to the user frequency tensor.

[0015] In this embodiment, the edge device can uniformly divide the target spatiotemporal region to obtain a set of basic grid cells. The basic grid cells in this set of basic grid cells are of the same size and have a regular structure. At the same time, the edge device can establish an initial spatiotemporal grid structure system based on the basic grid cells. Edge devices can determine the cumulative frequency of user orientation for each basic grid cell in the basic grid cell set based on the pre-collected user historical movement trajectory dataset. The cumulative frequency is then normalized to obtain a standard access probability value, which reflects the historical access density of each basic grid cell. Subsequently, the edge device can determine the adaptive spatiotemporal grid structure based on the access probability value of each basic grid cell. Specifically, all grid cells can be strictly sorted from high to low access probability values ​​to obtain a sorting result. The top 80% of high access probability grid cells in the sorting result are identified and selected. These high access probability grid cells are then subdivided using a bisection method to divide them into two equal parts in each dimension. The access probability values ​​of all subdivided grid cells are then recalculated and sorted again from high to low. The top 80% of high access probability grid cells in the re-sorting result are selected for the next round of subdivision. This subdivision-sorting-re-subdivision process is iterated three times to form an adaptive spatiotemporal grid structure with a hierarchical precision distribution. Finally, the edge device constructs a user frequency tensor for each user based on the adaptive spatiotemporal grid structure. The dimensional structure of the user frequency tensor is consistent with the adaptive spatiotemporal grid structure, and each element value in the user frequency tensor represents the historical access frequency statistics of the user in the corresponding spatiotemporal grid cell.

[0016] Edge devices can also use a Laplace distributed random function to generate a noise tensor that is completely corresponding to the user frequency tensor structure. The intensity of this noise is strictly controlled by a preset privacy protection parameter ε. A smaller ε value provides stronger privacy protection, but correspondingly reduces data availability; a larger ε value has the opposite effect. In this invention, the value of ε ranges from 0.1 to 100.

[0017] It should be noted that the values ​​mentioned above are for illustrative purposes only, and can be adjusted to other values ​​according to the actual situation. No specific restrictions are imposed.

[0018] 102. Edge devices determine a synthetic tensor with differential privacy protection features based on the noise tensor and the user frequency tensor.

[0019] In this embodiment, after determining the noise tensor and the user frequency tensor, the edge device can perform element-wise precise addition on the user frequency tensor and the noise tensor to obtain a synthetic tensor with differential privacy protection characteristics. This synthetic tensor effectively preserves the main features of the user's activity pattern, while introducing random noise to effectively prevent precise inference of the user's sensitive individual information, thereby achieving privacy data protection.

[0020] 103. Edge devices normalize user frequency tensors and composite tensors.

[0021] In this embodiment, the edge device can normalize the user frequency tensor and the composite tensor respectively. The normalization process determines that the value of each element in the tensor is combined to 1, and converts it into a probability distribution tensor. The normalization process strictly follows the differential privacy protection mechanism to maintain data privacy.

[0022] 104. The edge device constructs a DP-2M Markov model based on the normalized user frequency tensor and the synthetic tensor, and solves the DP-2M Markov model to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features.

[0023] In this embodiment, the edge device performs Markov model construction and solution based on the normalized tensor. The following section combines... Figure 2 To explain, Figure 2 This is a schematic diagram of the construction process of the DP-2M model provided in this embodiment of the invention. Specifically, the edge device uses the DP-2M Markov model to construct the user state transition tensor, representing the model as a state transition matrix in tensor form. An initial state distribution tensor is set to represent the user's initial probability distribution. Through the tensor unified multiplication operation, the initial state distribution tensor and the transition tensor are iteratively multiplied. This iterative multiplication operation continues until the state distribution converges to a stable value. The converged distribution is the joint stable distribution tensor (SJE) of a single user. Correspondingly, the same method is used to generate a joint stable distribution tensor (DP-SJE) with privacy protection characteristics based on the differential privacy-preserving transition tensor.

[0024] 105. Edge devices determine the spatiotemporal distance and semantic distance between each user's SJE and each user's DP-SJE.

[0025] In this embodiment, to measure the difference between the original steady-state joint tensor and the steady-state joint tensor under differential privacy, we need to perform quantitative analysis based on the spatiotemporal distance and semantic distance of the tensors. The distributions of spatiotemporal distance and semantic distance reflect the similarity of two tensors in the spatiotemporal and semantic dimensions. These two distances preserve the statistical characteristics of the original steady-state joint distribution tensor as much as possible, thereby ensuring the accuracy of the prediction task. Thus, the edge device can determine the spatiotemporal distance and semantic distance between each user's SJE and each user's DP-SJE.

[0026] In one embodiment, the edge device determines the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user, including: The edge device vectorizes the SJE and DP-SJE of each user to obtain vectors x and y; The edge device sorts the elements in the vectors x and y in descending order; The edge device calculates the target distance between the coordinates of two elements before and after sorting; The edge device calculates the target difference between elements before and after sorting. The edge device determines the spatiotemporal distance based on the target distance and the target difference using the following formula: ; in, The spatiotemporal distance is... The target distance is... For the target difference; The edge device determines the semantic distance using the following formula: ; in, The semantic distance, and For each user, the element at the corresponding position in the SJE is... The element at the corresponding position of DP-SJE for each user.

[0027] In this embodiment, the spatiotemporal distance is calculated as follows: First, given two tensors... and Let represent the SJE and DP-SJE of each user, respectively, and let the tensor and Vectorization, resulting in a vector and Then sort the elements of the two vectors in descending order. The vector elements before sorting are represented as follows: and The sorted vector elements are represented as and Secondly, calculate the distance between the coordinates of two elements before and after sorting. , This represents the difference in the position of the l-th element in the two tensors after sorting. Then, the difference in the elements before and after sorting is calculated. Finally, the tensor is calculated. and The spacetime distance is .

[0028] Semantic distance calculation: First, given two tensors respectively and , representing each user's SJE and each user's DP-SJE respectively. and These are the two elements at corresponding positions in the tensor. It is the average of these two elements, that is Edge devices use JSD to calculate the semantic distance between these two spatiotemporal locations, i.e. .in, , Similarly, the calculation of tensors is performed. and tensor The formula for calculating semantic distance is as follows: .

[0029] 106. Edge devices optimize DP-SJE for each user based on spatiotemporal distance and semantic distance.

[0030] In this embodiment, after determining the spatiotemporal distance and semantic distance, the cloud device can optimize the DP-SJE of each user based on the spatiotemporal distance and semantic distance, that is, minimize the spatiotemporal distance and semantic distance, and determine the minimized DP-SJE of each user as the optimized DP-SJE of each user.

[0031] 107. The edge device sends the synthesized tensor and the optimized DP-SJE for each user to the cloud device.

[0032] In this embodiment, the edge device can send the synthesized tensor and the optimized DP-SJE of each user to the cloud device. The cloud device determines the user influence coefficient corresponding to each user based on the synthesized tensor, and determines the multi-user DP-SJE based on the optimized DP-SJE of each user and the user influence coefficient.

[0033] In summary, it can be seen that the embodiments provided by this invention balance the relationship between data privacy and prediction accuracy by introducing two measurement methods: tensor-based spatiotemporal distance and tensor-based semantic distance. A DP-2M Markov model is implemented at the edge to generate single-user SJE and single-user DP-SJE, ensuring their privacy when transmitted to the cloud. At the same time, a user influence coefficient is constructed in the cloud based on the single-user synthetic tensor, and a multi-purpose DP-SJE is generated based on the user influence coefficient and the single-user DP-SJE, effectively protecting data privacy.

[0034] The following provides a detailed explanation of the method for determining the multi-user joint steady-state distribution tensor provided in this invention from the perspective of cloud devices. Please refer to [link to relevant documentation]. Figure 3 , Figure 3 Another flowchart illustrating the method for determining the multi-user joint steady-state distribution tensor provided in this embodiment of the invention includes: 301. The cloud device receives the synthesized tensor sent by the edge device and the optimized joint steady-state distribution tensor DP-SJE with privacy protection features for each user.

[0035] In this embodiment, the cloud device can receive the synthesized tensor and the optimized DP-SJE for each user sent by the edge device. Figure 1 The article has already provided a detailed explanation of how edge devices generate synthetic tensors and the optimized DP-SJE for each user, so the details will not be repeated here.

[0036] 302. The cloud device determines the user influence coefficient corresponding to each user based on the composite tensor.

[0037] In this embodiment, the cloud device determines the influence coefficient using the following formula: ; in, The influence coefficient of each user on each spatiotemporal location of the target user. For the first user's synthesized tensor, Let the second user be the composite tensor. The first and second users are any two distinct users from a pool of users. The influence coefficient of this user is expressed by the following formula: Specifically, the cloud-based device first uses the Pearson correlation coefficient (PCC) to calculate the influence coefficient of each user on the target user at each spatiotemporal location. The calculation formula is: Tensor The differential privacy-based transition tensor for user 1, the tensor This represents the differential privacy-based transition tensor for target user 2. and represent tensors and tensor The average value of each element. User A's influence coefficient on the target user can be expressed as: Then, the influence coefficients of all users related to the target user are calculated and normalized. The calculation process is as follows: .

[0038] 303. The cloud device determines the multi-user DP-SJE based on the optimized DP-SJE of each user and the normalized user influence coefficient.

[0039] In this embodiment, the cloud device multiplies the optimized DP-SJE of each user with its corresponding normalized user influence coefficient, and then aggregates these results to obtain the multi-user DP-SJE. The formula is as follows: ; Where t represents a certain moment, m represents several past moments, n represents the number of users affecting the target user, and u represents the target user.

[0040] The following is combined Figure 4 For further explanation, please refer to [link / reference]. Figure 4 , Figure 4 A schematic diagram of the DP-3M Markov model provided in an embodiment of the present invention, as shown below. Figure 4 As shown, a DP-3M Markov prediction model is constructed based on multi-user DP-SJE to predict the movement trajectory of target users while protecting data privacy (since multi-user DP-SJE is transmitted to the cloud after adding differential privacy at the edge of single-user SJE, user privacy is protected while predicting the movement trajectory of target users). For different scenarios, prediction targets are extracted and ranked, and the target value with the highest probability is selected for prediction, achieving accurate multimodal prediction. For example, based on multi-user DP-SJE, different predictions can be provided when different state attributes are specified for different values. Assume that the state of a suitable prediction system is... Decision, when When specified by different conditions, we can use... The prediction of clothing is achieved by using values ​​(i.e., sorted by probability from highest to lowest, for example, w=1, the clothing with the highest probability is the predicted value for that scenario), such as... wait.

[0041] In summary, it can be seen that the embodiments provided by this invention balance the relationship between data privacy and prediction accuracy by introducing two measurement methods: tensor-based spatiotemporal distance and tensor-based semantic distance. A DP-2M Markov model is implemented at the edge to generate single-user SJE and single-user DP-SJE, ensuring their privacy when transmitted to the cloud. At the same time, a user influence coefficient is constructed in the cloud based on the single-user synthetic tensor, and a multi-purpose DP-SJE is generated based on the user influence coefficient and the single-user DP-SJE, effectively protecting data privacy.

[0042] The above describes the method for determining the multi-user joint steady-state distribution tensor provided by the embodiments of the present invention from the perspectives of edge devices and cloud devices, respectively. The following describes the embodiments of the present invention from the perspective of the interaction between edge devices and cloud devices.

[0043] Please see Figure 5 , Figure 5 Another flowchart illustrating the method for determining the multi-user joint steady-state distribution tensor provided in this embodiment of the invention includes: 501. The edge device constructs a user frequency tensor for each of the multiple users and generates a noise tensor corresponding to the user frequency tensor.

[0044] 502. Edge devices determine a synthetic tensor with differential privacy protection features based on the noise tensor and the user frequency tensor.

[0045] 503. Edge devices normalize user frequency tensors and composite tensors.

[0046] 504. The edge device constructs a DP-2M Markov model based on the normalized user frequency tensor and the synthetic tensor, and solves the DP-2M Markov model to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features.

[0047] 505. Edge devices determine the spatiotemporal distance and semantic distance between each user's SJE and each user's DP-SJE.

[0048] 506. Edge devices optimize DP-SJE for each user based on spatiotemporal distance and semantic distance.

[0049] 507. The edge device sends the synthesized tensor and the optimized DP-SJE for each user to the cloud device.

[0050] It should be noted that steps 501 to 507 are related to... Figure 1 Steps 101 to 107 are similar to those described above. Figure 1 The details have already been explained in detail, and will not be repeated here.

[0051] 508. The cloud-based device determines the user influence coefficient corresponding to each user based on the composite tensor; 509. The cloud device determines the multi-user DP-SJE based on the optimized DP-SJE of each user and the normalized user influence coefficient.

[0052] It should be noted that steps 508 to 509 are related to... Figure 3 Steps 302 to 303 are similar to those described above. Figure 3 The details have already been explained in detail, and will not be repeated here.

[0053] In summary, it can be seen that the embodiments provided by this invention balance the relationship between data privacy and prediction accuracy by introducing two measurement methods: tensor-based spatiotemporal distance and tensor-based semantic distance. A DP-2M Markov model is implemented at the edge to generate single-user SJE and single-user DP-SJE, ensuring their privacy when transmitted to the cloud. At the same time, a user influence coefficient is constructed in the cloud based on the single-user synthetic tensor, and a multi-purpose DP-SJE is generated based on the user influence coefficient and the single-user DP-SJE, effectively protecting data privacy.

[0054] The embodiments of the present invention have been described above from the perspective of determining the multi-user joint steady-state distribution tensor. The embodiments of the present invention will be described below from the perspective of edge devices and cloud devices.

[0055] Please see Figure 6 , Figure 6 A schematic diagram of the virtual structure of an edge device in this embodiment of the invention. The edge device 600 includes: The construction module 601 is used to construct a user frequency tensor for each of the multiple users and generate a noise tensor corresponding to the user frequency tensor. The first determining module 602 is used to determine a synthetic tensor with differential privacy protection characteristics based on the noise tensor and the user frequency tensor. The normalization processing module 603 is used to normalize the user frequency tensor and the synthetic tensor; The processing module 604 is used to construct a DP-2M Markov model based on the normalized user frequency tensor and the synthetic tensor, and solve the DP-2M Markov model to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features. The second determining module 605 is used to determine the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user; Optimization module 606 is used to optimize the DP-SJE of each user based on the spatiotemporal distance and the semantic distance; The sending module 607 is used to send the synthesized tensor and the optimized DP-SJE to the cloud device, so that the cloud device can determine the user influence coefficient corresponding to each user based on the synthesized tensor, and determine the multi-user DP-SJE based on the user influence coefficient and the optimized DP-SJE of each user.

[0056] In one possible design, the second determining module 605 is specifically used for: The SJE and DP-SJE of each user are vectorized to obtain vectors x and y; Sort the elements in the vectors x and y in descending order; Calculate the target distance between the coordinates of two elements before and after sorting; Calculate the target difference between elements before and after sorting; Based on the target distance and the target difference, the spatiotemporal distance is determined using the following formula: ; in, The spatiotemporal distance is... The target distance is... For the target difference; The semantic distance is determined by the following formula: ; in, The semantic distance, and For each user, the element at the corresponding position in the SJE is... The element at the corresponding position of DP-SJE for each user.

[0057] In one possible design, the building module 601 is specifically used for: The target spatiotemporal region is uniformly divided to obtain a set of basic grid cells; The access probability value of each basic grid cell in the basic grid cell set is determined based on the user's historical movement trajectory dataset; The adaptive spatiotemporal grid structure is determined based on the access probability value of each basic grid cell; The user frequency tensor is constructed for each user based on the adaptive spatiotemporal grid structure.

[0058] In one possible design, the construction module 601 determines the access probability value of each basic grid cell in the basic grid cell set based on the user's historical movement trajectory dataset, including: Calculate the cumulative frequency of user access to each basic grid cell in the aforementioned basic grid cell set; The cumulative frequency is normalized to obtain the access probability value of each basic grid cell.

[0059] Please see Figure 7 , Figure 7 This is a schematic diagram of the virtual structure of a cloud device provided in an embodiment of the present invention. The cloud device 700 includes: The receiving module 701 is used to receive the synthetic tensor and the optimized joint steady-state distribution tensor DP-SJE with privacy protection features for each user sent by the edge device. The synthetic tensor is determined by the edge device based on the noise tensor and the user frequency tensor of each user. The optimized DP-SJE of each user is obtained by the edge device by optimizing the DP-SJE of each user based on the spatiotemporal distance and the semantic distance. The spatiotemporal distance and the semantic distance are the spatiotemporal distance and the semantic distance between the SJE of each user and the DP-SJE of each user. The SP-SJE of each user is obtained by the edge device by constructing a DP-2M Markov model based on the normalized user frequency tensor and the normalized synthetic tensor, and solving the DP-2M Markov model. The first determining module 702 is used to determine the user influence coefficient corresponding to each user based on the synthesized tensor; The second determining module 703 is used to determine the multi-user DP-SJE based on the optimized DP-SJE of each user and the normalized user influence coefficient.

[0060] In one possible design, the first determining module 702 is specifically used for: The influence coefficient of each user on each spatiotemporal location of the target user is determined by the following formula: ; in, For the first user's synthesized tensor, The second user is a composite tensor, where the first user and the second user are any two different users among the plurality of users; The user influence coefficient is determined using the following formula: .

[0061] In one possible design, the second determining module 703 is specifically used for: The multi-user DP-SJE is determined by the following formula: in, Let t represent a certain moment, m represent several past moments, n represent the number of users affecting the target user, and u represent the target user.

[0062] In one possible design, the first determining module 702 is further configured to: The user influence coefficient is normalized using the following formula: ; in, The user influence coefficient is the normalized value. The user influence coefficient, Let be the influence coefficient of each user on each spatiotemporal location of the target user.

[0063] above Figure 6 and Figure 7 The present invention has been described from the perspective of modular functional entities. The following is a detailed description of the device for determining the multi-user joint steady-state distribution tensor in the embodiments of the present invention from the perspective of hardware processing. Please refer to Figure 800, which is a schematic diagram of an embodiment of the device 800 for determining the multi-user joint steady-state distribution tensor in the present invention. The device 800 includes: Input device 801, output device 802, processor 803, and memory 804 (where the number of processors 803 can be one or more). Figure 8 (Taking a processor 803 as an example). In some embodiments of the present invention, the input device 801, the output device 802, the processor 803, and the memory 804 may be connected via a communication bus or other means, wherein... Figure 8 Take the China-Israel communication bus connection as an example.

[0064] Specifically, by calling the operation instructions stored in memory 804, processor 803 performs the following steps: For each of the multiple users, construct a user frequency tensor and generate a noise tensor corresponding to the user frequency tensor; A synthetic tensor with differential privacy protection characteristics is determined based on the noise tensor and the user frequency tensor. The user frequency tensor and the composite tensor are normalized. Based on the normalized user frequency tensor and the synthetic tensor, a DP-2M Markov model is constructed, and the DP-2M Markov model is solved to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features. Determine the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user; The DP-SJE of each user is optimized based on the spatiotemporal distance and the semantic distance; The synthesized tensor and the optimized DP-SJE are sent to the cloud device so that the cloud device can determine the user influence coefficient corresponding to each user based on the synthesized tensor, and determine the multi-user DP-SJE based on the user influence coefficient and the optimized DP-SJE of each user.

[0065] Specifically, by calling the operation instructions stored in memory 804, processor 803 performs the following steps: The system receives a synthetic tensor and an optimized joint steady-state distribution tensor (DP-SJE) with privacy protection features for each user, sent by an edge device. The synthetic tensor is determined by the edge device based on a noise tensor and the user frequency tensor of each user. The optimized DP-SJE for each user is obtained by the edge device by optimizing the DP-SJE of each user based on spatiotemporal distance and semantic distance. The spatiotemporal distance and semantic distance are the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user. The SP-SJE for each user is obtained by the edge device by constructing a DP-2M Markov model based on the normalized user frequency tensor and the normalized synthetic tensor, and solving the DP-2M Markov model. The user influence coefficient corresponding to each user is determined based on the composite tensor; The multi-user DP-SJE is determined based on the optimized DP-SJE for each user and the normalized user influence coefficient.

[0066] By calling the operation instructions stored in memory 804, processor 803 is also used to execute... Figure 1 or Figure 3 Any of the methods in the corresponding embodiments.

[0067] Please see Figure 9 , Figure 9 A schematic diagram of an embodiment of the electronic device provided in this invention.

[0068] like Figure 9 As shown, this embodiment of the invention provides an electronic device, including a memory 910, a processor 920, and a computer program 911 stored in the memory 910 and executable on the processor 920. When the processor 920 executes the computer program 911, it performs the following steps: For each of the multiple users, construct a user frequency tensor and generate a noise tensor corresponding to the user frequency tensor; A synthetic tensor with differential privacy protection characteristics is determined based on the noise tensor and the user frequency tensor. The user frequency tensor and the composite tensor are normalized. Based on the normalized user frequency tensor and the synthetic tensor, a DP-2M Markov model is constructed, and the DP-2M Markov model is solved to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features. Determine the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user; The DP-SJE of each user is optimized based on the spatiotemporal distance and the semantic distance; The synthesized tensor and the optimized DP-SJE are sent to the cloud device so that the cloud device can determine the user influence coefficient corresponding to each user based on the synthesized tensor, and determine the multi-user DP-SJE based on the user influence coefficient and the optimized DP-SJE of each user.

[0069] When processor 920 executes computer program 911, it also performs the following steps: The system receives a synthetic tensor and an optimized joint steady-state distribution tensor (DP-SJE) with privacy protection features for each user, sent by an edge device. The synthetic tensor is determined by the edge device based on a noise tensor and the user frequency tensor of each user. The optimized DP-SJE for each user is obtained by the edge device by optimizing the DP-SJE of each user based on spatiotemporal distance and semantic distance. The spatiotemporal distance and semantic distance are the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user. The SP-SJE for each user is obtained by the edge device by constructing a DP-2M Markov model based on the normalized user frequency tensor and the normalized synthetic tensor, and solving the DP-2M Markov model. The user influence coefficient corresponding to each user is determined based on the composite tensor; The multi-user DP-SJE is determined based on the optimized DP-SJE for each user and the normalized user influence coefficient.

[0070] In practical implementation, when the processor 920 executes the computer program 911, it can achieve... Figure 1 or Figure 3 Any of the corresponding implementation methods in the embodiments.

[0071] Please refer to Figure 1000, which is a schematic diagram of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention.

[0072] As shown in Figure 1000, this embodiment of the invention also provides a computer-readable storage medium 1000, on which a computer program 1011 is stored. When the computer program 1011 is executed by a processor, it performs the following steps: For each of the multiple users, construct a user frequency tensor and generate a noise tensor corresponding to the user frequency tensor; A synthetic tensor with differential privacy protection characteristics is determined based on the noise tensor and the user frequency tensor. The user frequency tensor and the composite tensor are normalized. Based on the normalized user frequency tensor and the synthetic tensor, a DP-2M Markov model is constructed, and the DP-2M Markov model is solved to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features. Determine the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user; The DP-SJE of each user is optimized based on the spatiotemporal distance and the semantic distance; The synthesized tensor and the optimized DP-SJE are sent to the cloud device so that the cloud device can determine the user influence coefficient corresponding to each user based on the synthesized tensor, and determine the multi-user DP-SJE based on the user influence coefficient and the optimized DP-SJE of each user.

[0073] When the computer program 1011 is executed by the processor, it also performs the following steps: The system receives a synthetic tensor and an optimized joint steady-state distribution tensor (DP-SJE) with privacy protection features for each user, sent by an edge device. The synthetic tensor is determined by the edge device based on a noise tensor and the user frequency tensor of each user. The optimized DP-SJE for each user is obtained by the edge device by optimizing the DP-SJE of each user based on spatiotemporal distance and semantic distance. The spatiotemporal distance and semantic distance are the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user. The SP-SJE for each user is obtained by the edge device by constructing a DP-2M Markov model based on the normalized user frequency tensor and the normalized synthetic tensor, and solving the DP-2M Markov model. The user influence coefficient corresponding to each user is determined based on the composite tensor; The multi-user DP-SJE is determined based on the optimized DP-SJE for each user and the normalized user influence coefficient.

[0074] In the specific implementation process, the computer program 1011 is executed by the processor to achieve... Figure 1 or Figure 3 Any of the corresponding implementation methods in the embodiments.

[0075] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0076] This invention also provides a computer program product comprising computer software instructions that, when executed on a processing device, cause the processing device to perform actions such as... Figure 1 The process in the corresponding embodiment.

[0077] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining a multi-user joint steady-state distribution tensor, characterized in that, include: The edge device constructs a user frequency tensor for each of the multiple users and generates a noise tensor corresponding to the user frequency tensor; The edge device determines a synthetic tensor with differential privacy protection features based on the noise tensor and the user frequency tensor. The edge device normalizes the user frequency tensor and the composite tensor. The edge device constructs a DP-2M Markov model based on the normalized user frequency tensor and the synthetic tensor, and solves the DP-2M Markov model to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features. The edge device determines the spatiotemporal distance and semantic distance between each user's SJE and each user's DP-SJE; The edge device optimizes the DP-SJE of each user based on the spatiotemporal distance and the semantic distance; The edge device sends the synthesized tensor and the optimized DP-SJE to the cloud device, so that the cloud device determines the user influence coefficient corresponding to each user based on the synthesized tensor, and determines the multi-user DP-SJE based on the user influence coefficient and the optimized DP-SJE of each user.

2. The method according to claim 1, characterized in that, The edge device determines the spatiotemporal distance and semantic distance between each user's SJE and each user's DP-SJE, including: The edge device vectorizes the SJE and DP-SJE of each user to obtain vectors x and y; The edge device sorts the elements in the vectors x and y in descending order; The edge device calculates the target distance between the coordinates of two elements before and after sorting; The edge device calculates the target difference between elements before and after sorting. The edge device determines the spatiotemporal distance based on the target distance and the target difference using the following formula: ； in, The spatiotemporal distance is... The target distance is... For the target difference; The edge device determines the semantic distance using the following formula: ； in, The semantic distance, and For each user, the element at the corresponding position in the SJE is... The element at the corresponding position of DP-SJE for each user.

3. The method according to claim 1, characterized in that, The edge device constructs a user frequency tensor for each user, including: The edge device uniformly divides the target spatiotemporal region to obtain a set of basic grid cells; The edge device determines the access probability value of each basic grid cell in the basic grid cell set based on the user's historical movement trajectory dataset; The edge device determines the adaptive spatiotemporal grid structure based on the access probability value of each basic grid cell; The edge device constructs the user frequency tensor for each user based on the adaptive spatiotemporal grid structure.

4. The method according to claim 3, characterized in that, The edge device determines the access probability value of each basic grid cell in the basic grid cell set based on the user's historical movement trajectory dataset, including: The edge device counts the cumulative frequency of user access to each basic grid cell in the basic grid cell set; The edge device normalizes the cumulative frequency to obtain the access probability value of each basic grid cell.

5. A method for determining a multi-user joint steady-state distribution tensor, characterized in that, include: The cloud device receives a synthetic tensor and an optimized joint steady-state distribution tensor (DP-SJE) with privacy protection features for each user sent by the edge device. The synthetic tensor is determined by the edge device based on the noise tensor and the user frequency tensor of each user. The optimized DP-SJE for each user is obtained by the edge device by optimizing the DP-SJE of each user based on the spatiotemporal distance and semantic distance. The spatiotemporal distance and the semantic distance are the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user. The SP-SJE for each user is obtained by the edge device by constructing a DP-2M Markov model based on the normalized user frequency tensor and the normalized synthetic tensor, and solving the DP-2M Markov model. The cloud device determines the user influence coefficient corresponding to each user based on the composite tensor; The cloud device determines the multi-user DP-SJE based on the optimized DP-SJE of each user and the normalized user influence coefficient.

6. The method according to claim 5, characterized in that, The cloud device determines the user influence coefficient corresponding to each user based on the composite tensor, including: The cloud device determines the influence coefficient of each user on each spatiotemporal location of the target user using the following formula: ； in, For the first user's synthesized tensor, The second user is a composite tensor, where the first user and the second user are any two different users among the plurality of users; The cloud device determines the user influence coefficient using the following formula: 。 7. The method according to claim 6, characterized in that, The cloud device determines the multi-user DP-SJE based on the optimized DP-SJE for each user and the normalized user influence coefficient, including: The cloud device determines the multi-user DP-SJE using the following formula: in, Let t represent a certain moment, m represent several past moments, n represent the number of users affecting the target user, and u represent the target user.

8. The method according to claim 7, characterized in that, The method further includes: The cloud device normalizes the user influence coefficient using the following formula: ； in, The user influence coefficient is the normalized value. The user influence coefficient, The influence coefficient of each user on each spatiotemporal location of the target user.

9. An edge device, characterized in that, include: A construction module is used to construct a user frequency tensor for each of the multiple users and generate a noise tensor corresponding to the user frequency tensor; The first determining module is used to determine a synthetic tensor with differential privacy protection characteristics based on the noise tensor and the user frequency tensor. The normalization processing module is used to normalize the user frequency tensor and the synthetic tensor. The processing module is used to construct a DP-2M Markov model based on the normalized user frequency tensor and the synthetic tensor, and solve the DP-2M Markov model to obtain the joint stable distribution tensor SJE of each user and the joint stable distribution tensor DP-SJE of each user with privacy protection features. The second determining module is used to determine the spatiotemporal distance and semantic distance between the SJE of each user and the DP-SJE of each user; The optimization module is used to optimize the DP-SJE of each user based on the spatiotemporal distance and the semantic distance; The sending module is used to send the synthesized tensor and the optimized DP-SJE to the cloud device, so that the cloud device can determine the user influence coefficient corresponding to each user based on the synthesized tensor, and determine the multi-user DP-SJE based on the user influence coefficient and the optimized DP-SJE of each user.

10. A cloud device, characterized in that, include: The receiving module is used to receive the synthetic tensor and the optimized joint steady-state distribution tensor (DP-SJE) with privacy protection features for each user sent by the edge device. The synthetic tensor is determined by the edge device based on the noise tensor and the user frequency tensor of each user. The optimized DP-SJE for each user is obtained by the edge device by optimizing the DP-SJE of each user based on the spatiotemporal distance and the semantic distance. The spatiotemporal distance and the semantic distance are the spatiotemporal distance and the semantic distance between the SJE of each user and the DP-SJE of each user. The SP-SJE for each user is obtained by the edge device by constructing a DP-2M Markov model based on the normalized user frequency tensor and the normalized synthetic tensor, and solving the DP-2M Markov model. The first determining module is used to determine the user influence coefficient corresponding to each user based on the synthesized tensor; The second determining module is used to determine the multi-user DP-SJE based on the optimized DP-SJE of each user and the normalized user influence coefficient.

Images