Data supervision method and system for Internet of Things

Abstract

The invention relates to the technical field of big data, in particular to a data supervision method and system oriented to the Internet of Things. The invention discloses a data supervision method for the Internet of Things. The method comprises the following steps: S1, constructing a video fence to obtain a multi-level dynamic geofence structure; s2, according to the multi-level dynamic geofence structure, using a fence radius calculation formula to obtain the fence radius of each level; and S3, obtaining a final fence area according to the fence radius of each level, and carrying out differential privacy enhancement processing on supervision data in the final fence area. By constructing a three-level dynamic geofence structure, different levels of differential supervision strategies are realized, and the fence radius of each supervision object is calculated and adjusted in real time by using a time weighting function, a behavior preference function and a privacy data influence function, so that a supervision area is dynamically expanded and shrunk along with object characteristics and risk levels, and the supervision accuracy is improved. And the flexibility and accuracy of supervision are improved.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a data supervision method and system for the Internet of Things. Background Art

[0002] With the widespread deployment of IoT devices in smart transportation, smart cities, smart homes, and industrial control, the demand for real-time collection and analysis of massive amounts of video surveillance and sensor data is rapidly increasing. However, traditional static geofencing technology has the following limitations: fixed boundaries and lack of dynamism: Static geofencing typically uses preset circular or polygonal areas, making it difficult to achieve flexible real-time monitoring of mobile targets and prone to missed detections or false alarms; coarse-grained supervision and insufficient protection: Static geofencing treats all targets equally, failing to differentiate the behavioral preferences and privacy needs of different objects at different times, and failing to strike a balance between data utilization and privacy protection; and a single privacy protection mechanism: Existing differential privacy algorithms are primarily targeted at static queries or batch data processing, making it difficult to implement fine-grained privacy budget allocation and noise addition for data in different areas and at different risk levels in real-time dynamic scenarios. Summary of the Invention

[0003] In order to overcome the shortcomings of lack of fixed boundaries, lack of dynamicity and privacy protection of fences, the present invention provides a data supervision method and system for the Internet of Things.

[0004] The technical implementation scheme of the present invention is: a data supervision method for the Internet of Things, comprising the following steps:

[0005] S1: Construct the video fence to obtain a multi-level dynamic geo-fence structure;

[0006] S2: According to the multi-level dynamic geo-fence structure, using the fence radius calculation formula to obtain the fence radius of each level;

[0007] S3: Obtain a final fence area according to the fence radius of each level, and perform differential privacy enhancement processing on the supervision data within the final fence area.

[0008] Preferably, the video fence is constructed to obtain a multi-level dynamic geo-fence structure, including: the multi-level dynamic geo-fence structure includes a peripheral perception layer, an early warning layer and a core supervision layer.

[0009] Preferably, according to the multi-level dynamic geographic fence structure, the fence radius of each level is obtained using a fence radius calculation formula, including: obtaining personalized time data of the supervised object, and obtaining a time weight value based on the personalized time data using a time weight function; obtaining behavioral preference data of the supervised object, and obtaining a behavioral preference value based on the behavioral preference data using a behavioral preference function; obtaining privacy-related data, and obtaining privacy risks based on the privacy-related data using a privacy data impact function; and obtaining the fence radius of each level using a fence radius calculation formula based on the time weight value, behavioral preference value and privacy risk.

[0010] Preferably, the fence radius of each level is obtained using a fence radius calculation formula according to the time weight value, behavior preference value and privacy risk, including: wherein the fence radius calculation formula is:

[0011] R i =R i,0 *(1+α i *f t (t)+β i *f h (H)―γ i * E);

[0012] Where R i is the fence radius of the i-th fence; R i,0 is the base fence radius; f t (t) is the time weight value; f h (H) is the behavior preference value; E is the privacy risk; α i , β i , γ i is the weight adjustment coefficient.

[0013] Preferably, the obtaining of personalized time data of the supervised object and obtaining a time weight value using a time weight function according to the personalized time data includes: the personalized time data includes the current time, the daytime activity cycle and each high-attention period, and the obtaining of the time weight value using a time weight function according to the personalized time data, wherein the time weight function is:

[0014]

[0015] Where, f t (t) is the time weight value; t is the current moment; ρ d 、 is the diurnal activity cycle adjustment coefficient and phase, and ρ d ∈[0,1]; μ k is the central moment of the kth high attention period; σ kis the variance of the kth high attention period; K is the number of high attention periods; ρ k is the weight coefficient of the kth high attention period.

[0016] Preferably, the step of obtaining the behavior preference data of the supervised object and obtaining the behavior preference value using a behavior preference function according to the behavior preference data includes: the behavior preference data includes the real-time location data, frequently visited places, and visit frequency of the supervised object; obtaining the visit frequency weight of the frequently visited place cluster of the supervised object, the two-dimensional plane coordinates of the supervised object, and the centroid coordinates of the frequently visited place cluster through the behavior preference data; and obtaining the behavior preference value using a behavior preference function, wherein the behavior preference function is:

[0017]

[0018] Where, f h (H) is the behavior preference value; ω m is the visit frequency weight of the mth frequently visited location cluster; [x, y] is the two-dimensional coordinate of the supervised object at the current moment; τ m is the centroid coordinate of the mth frequently visited location cluster; δ m is the variance of the mth frequently visited location cluster; ‖·‖ is the Euclidean norm, which calculates the straight-line distance between plane coordinates.

[0019] Preferably, obtaining privacy-related data and obtaining privacy risk using a privacy data impact function based on the privacy-related data includes: the privacy-related data includes the privacy budget consumed by the supervised object within a preset time window, the total privacy budget allocated to the supervised object, and a query sensitivity factor, wherein the privacy data impact function is:

[0020]

[0021] Where, E is the privacy risk; B current is the privacy budget consumed by the regulated object in the current time window; B total is the total privacy budget allocated to the regulated object; θ is the adjustment coefficient; S sensitivity is the query sensitivity factor.

[0022] Preferably, the final fence area is obtained according to the fence radius of each level, and the supervision data within the final fence area is subjected to differential privacy enhancement processing, including: obtaining the final fence area according to the fence radius of each layer of fence, and obtaining the supervision data and the current fence level of the supervised object according to the final fence area, obtaining the current privacy risk through the privacy data impact function, and performing differential privacy enhancement processing on the supervision data according to the current fence level of the supervised object and the current privacy risk.

[0023] Preferably, performing differentiated privacy enhancement processing on the supervision data according to the current fence level and current privacy risk of the supervised object includes: obtaining the total privacy budget allocated and the consumed privacy budget for this operation according to the current fence level and current privacy risk of the supervised object, and obtaining the noise addition amount using a noise addition formula, wherein the noise addition formula is:

[0024]

[0025] Where R is the amount of noise added; GlobalSensitivity (D) is the global sensitivity of query supervision data; D is the supervision data; ε alloc The total privacy budget allocated for this operation; ε used The privacy budget consumed.

[0026] Preferably, a data supervision system for the Internet of Things further includes:

[0027] An architecture partitioning module is used to construct video fences to obtain a multi-level dynamic geo-fence structure;

[0028] A fence radius calculation module, configured to obtain the fence radius of each level using a fence radius calculation formula according to the multi-level dynamic geo-fence structure;

[0029] a first impact acquisition unit, configured to acquire personalized time data of a supervised object, and obtain a time weight value using a time weight function according to the personalized time data;

[0030] a second influence obtaining unit, configured to obtain behavior preference data of the supervised object, and obtain a behavior preference value using a behavior preference function according to the behavior preference data;

[0031] a third impact acquisition unit, configured to acquire privacy-related data, and acquire privacy risks using a privacy data impact function according to the privacy-related data;

[0032] The privacy differentiation processing module is used to obtain the final fence area according to the fence radius of each level, and perform differential privacy enhancement processing on the supervision data within the final fence area.

[0033] The present invention has the following advantages:

[0034] 1. This invention implements differentiated supervision strategies at different levels by building a three-level dynamic geo-fence structure consisting of a peripheral perception layer, an early warning layer, and a core supervision layer. The high-level layer provides precise monitoring of key areas, while the low-level layer provides macroscopic perception of peripheral areas, thereby reducing missed reports and false alarm rates.

[0035] 2. Using the time weight function, behavior preference function, and privacy data impact function, the fence radius of each regulated object is calculated and adjusted in real time, allowing the regulated area to dynamically expand and contract based on the object's characteristics and risk level, improving the flexibility and accuracy of supervision.

[0036] 3. Based on the allocated budget and consumption, and according to the object's current fence level and privacy risk value, the noise addition formula is used to add appropriate amounts of differential privacy noise to data in different regions and time periods, taking into account both data availability and privacy protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of the data supervision method for the Internet of Things of the present invention;

[0038] Figure 2 This is a schematic diagram of the structure of the data supervision system for the Internet of Things of the present invention. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] Example 1: A data supervision method for the Internet of Things, such as Figure 1 As shown, the following steps are included:

[0041] S1: Construct the video fence to obtain a multi-level dynamic geo-fence structure;

[0042] The multi-level dynamic geo-fence structure includes a peripheral perception layer, an early warning layer and a core supervision layer.

[0043] It needs to be explained that real-time video streams are collected by deploying multiple cameras in the supervision area, and the video images are denoised, motion detected and tracked, the location information and movement trajectory of the supervised objects are extracted, and an initial radius is set around the supervised objects to build a "peripheral perception layer"; with a smaller base radius as the center, a "warning layer" is built; a more compact base radius is set within the warning layer to form a "core supervision layer" to collect basic position and motion information to provide data support for subsequent warnings; behavioral preference and time weight analysis are initiated to make a preliminary judgment on the target's abnormal behavior, and adjust the warning threshold according to the preference model; at the same time, the highest precision data collection and differentiated privacy protection strategies are carried out.

[0044] S2: According to the multi-level dynamic geo-fence structure, using the fence radius calculation formula to obtain the fence radius of each level;

[0045] Obtain the personalized time data of the supervised object, and use the time weight function to obtain the time weight value according to the personalized time data; obtain the behavioral preference data of the supervised object, and use the behavioral preference function to obtain the behavioral preference value according to the behavioral preference data; obtain privacy-related data, and use the privacy data impact function to obtain the privacy risk according to the privacy-related data; use the fence radius calculation formula to obtain the fence radius of each level according to the time weight value, behavioral preference value and privacy risk.

[0046] The formula for calculating the fence radius is:

[0047] R i =R i,0 *(1+α i *f t (t)+β i *f h (H)―γ i *E);

[0048] Where R i is the fence radius of the i-th fence; R i,0 is the base fence radius; f t (t) is the time weight value; f h (H) is the behavior preference value; E is the privacy risk; α i , β i , γ i is the weight adjustment coefficient.

[0049] It should be explained that the calculated radius of each layer is used to update the boundaries of the peripheral perception layer, warning layer, and core supervision layer; when the supervised object moves or the personalized data changes, the above steps are repeated in real time to achieve dynamic and adaptive geo-fence management; R i,0 The baseline fence radius is initially determined by the system administrator or domain expert based on the supervision scenario (such as the visual range of the surveillance camera and the communication coverage radius) and security requirements, or by collecting historical trajectory and warning effect data to calculate the optimal radius when the established missed alarm / false alarm rate threshold is reached, and the mean or median is taken.

[0050] The personalized time data includes the current time, the daytime activity cycle and each high-attention period, and a time weight value is obtained using a time weight function according to the personalized time data, wherein the time weight function is:

[0051]

[0052] Where, f t(t) is the time weight value; t is the current moment; ρ d 、 is the diurnal activity cycle adjustment coefficient and phase, and ρ d ∈[0,1]; μ k is the central moment of the kth high attention period; σ k is the variance of the kth high attention period; K is the number of high attention periods; ρ k is the weight coefficient of the kth high attention period.

[0053] It should be explained that the historical activity (displacement speed and access events) of the supervised object is counted, the normalized amplitude of its all-day activity curve is calculated, and the ratio of the amplitude to the maximum value is taken as ρ d ; According to the historical activity and activity time of the supervised object, the cosine curve is fitted to obtain μ k For the kth high-attention period center moment, historical events (such as abnormal behavior or important access) are clustered into k Gaussian distributions according to timestamps, and the model outputs the corresponding μ k and σ k ρ k is the weight coefficient of the kth high-attention period, which is obtained by counting the number of abnormal events or important visits in this period and normalizing the ratio of the number to the total number of visits for the whole day; is the diurnal cycle term; It is the peak item during the high attention period.

[0054] The behavior preference data includes the real-time location data, frequently visited places, and visit frequencies of the supervised object. The visit frequency weight of the supervised object's frequently visited place cluster, the two-dimensional plane coordinates of the supervised object, and the centroid coordinates of the frequently visited place cluster are obtained through the behavior preference data. The behavior preference value is obtained using the behavior preference function, where the behavior preference function is:

[0055]

[0056] Where, f h (H) is the behavior preference value; ω m is the visit frequency weight of the mth frequently visited location cluster; [x, y] is the two-dimensional coordinate of the supervised object at the current moment; τ m is the centroid coordinate of the mth frequently visited location cluster; δ m is the variance of the mth frequently visited location cluster; ‖·‖ is the Euclidean norm, which calculates the straight-line distance between plane coordinates.

[0057] It is necessary to explain that it is used to quantify the "normality" or "deviance" of the current location relative to the user's historical activity pattern, so as to: reduce the sensitivity of the fence in "frequently visited places"; increase the strictness of the fence in "atypical" or "sudden deviation" locations; δ m is the variance of the mth frequently visited location cluster. In the same clustering result, the standard deviation of the Euclidean distance from all samples to the centroid is calculated or approximated in the form of radius / diameter; ω m is the visit frequency weight of the mth frequently visited location cluster, and the proportion of location points falling in cluster m to all sampling points in the statistical history is normalized; τ m is the centroid coordinate of the mth frequently visited location cluster. Using historical trajectory data, we use the DBSCAN or K-means clustering algorithm to identify the frequently visited location clusters and calculate the average coordinates of the points in each cluster.

[0058] The privacy-related data includes the privacy budget consumed by the supervised object within a preset time window, the total privacy budget allocated to the supervised object, and the query sensitivity factor, where the privacy data impact function is:

[0059]

[0060] Where, E is the privacy risk; B current is the privacy budget consumed by the regulated object in the current time window; B total is the total privacy budget allocated to the regulated object; θ is the adjustment coefficient; S sensitivity is the query sensitivity factor.

[0061] It needs to be explained that B current B is the privacy budget consumed by the regulated object in the current time window. The privacy budget accountor accumulates the privacy budget consumed by all queries in the system in real time. total The total privacy budget allocated to the regulated object is pre-allocated according to the organization or regulations and dynamically adjusted on a daily / weekly basis; S sensitivity The query sensitivity factor is mapped to the corresponding global sensitivity according to the query type during initialization. For custom queries, it is calculated through maximum / minimum value definitions or sensitivity derivation formulas. θ is the adjustment coefficient, which is gradually tuned in a simulated environment through stress testing to select the optimal coefficient when privacy risks can be controlled and the fence failure rate is minimized.

[0062] S3: Obtain a final fence area according to the fence radius of each level, and perform differential privacy enhancement processing on the supervision data within the final fence area.

[0063] The final fence area is obtained based on the fence radius of each layer of fence, and the supervision data and the current fence level of the supervised object are obtained based on the final fence area. The current privacy risk is obtained through the privacy data impact function, and the supervision data is differentially privacy enhanced according to the current fence level of the supervised object and the current privacy risk.

[0064] According to the current fence level and current privacy risk of the supervised object, the total privacy budget allocated for this operation and the consumed privacy budget are obtained, and the noise addition amount is obtained using the noise addition formula, where the noise addition formula is:

[0065]

[0066] Where R is the amount of noise added; GlobalSensitivity (D) is the global sensitivity of query supervision data; D is the supervision data; ε alloc The total privacy budget allocated for this operation; ε used The privacy budget consumed.

[0067] It should be explained that according to the current fence level, the total privacy budget ε is read from the "Differential Privacy Policy Table" alloc , and query the consumed privacy budget ε in real time used ; The differential privacy policy table is a predefined decision matrix / mapping table used to allocate the privacy budget for this query based on the current fence level and privacy risk.

[0068] Example 2: Based on Example 1, a data monitoring system for the Internet of Things, such as Figure 2 As shown, it also includes:

[0069] An architecture partitioning module is used to construct video fences to obtain a multi-level dynamic geo-fence structure;

[0070] A fence radius calculation module, configured to obtain the fence radius of each level using a fence radius calculation formula according to the multi-level dynamic geo-fence structure;

[0071] a first impact acquisition unit, configured to acquire personalized time data of a supervised object, and obtain a time weight value using a time weight function according to the personalized time data;

[0072] a second influence obtaining unit, configured to obtain behavior preference data of the supervised object, and obtain a behavior preference value using a behavior preference function according to the behavior preference data;

[0073] a third impact acquisition unit, configured to acquire privacy-related data, and acquire privacy risks using a privacy data impact function according to the privacy-related data;

[0074] The privacy differentiation processing module is used to obtain the final fence area according to the fence radius of each level, and perform differential privacy enhancement processing on the supervision data within the final fence area.

[0075] It should be understood that this embodiment is only used to illustrate the present invention and is not used to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope limited by the appended claims of the application.

Claims

1. A data supervision method for the Internet of Things, characterized by: The following steps are involved: S1: Construct the video fence to obtain a multi-level dynamic geo-fence structure; S2: According to the multi-level dynamic geo-fence structure, using the fence radius calculation formula to obtain the fence radius of each level; S3: Obtain a final fence area according to the fence radius of each level, and perform differential privacy enhancement processing on the supervision data within the final fence area.

2. The data supervision method for the Internet of Things according to claim 1 is characterized in that: The video fence is constructed to obtain a multi-level dynamic geographic fence structure, including: the multi-level dynamic geographic fence structure includes a peripheral perception layer, an early warning layer and a core supervision layer.

3. The data supervision method for the Internet of Things according to claim 2 is characterized in that: According to the multi-level dynamic geographic fence structure, the fence radius of each level is obtained using a fence radius calculation formula, including: obtaining personalized time data of the supervised object, and obtaining a time weight value based on the personalized time data using a time weight function; obtaining behavioral preference data of the supervised object, and obtaining a behavioral preference value based on the behavioral preference data using a behavioral preference function; obtaining privacy-related data, and obtaining privacy risks based on the privacy-related data using a privacy data impact function; and obtaining the fence radius of each level using a fence radius calculation formula based on the time weight value, behavioral preference value and privacy risk.

4. The data supervision method for the Internet of Things according to claim 3 is characterized in that: The fence radius calculation formula is used to obtain the fence radius of each level according to the time weight value, behavior preference value and privacy risk, including: wherein the fence radius calculation formula is: R i =R i,0 *(1+a i *f t (t)+β i *f h (H)―c i *E); Where R i is the fence radius of the i-th fence; R i,0 is the base fence radius; f t (t) is the time weight value; f h (H) is the behavior preference value; E is the privacy risk; α i , β i , γ i is the weight adjustment coefficient.

5. The data supervision method for the Internet of Things according to claim 3 is characterized in that: The obtaining of personalized time data of the supervised object and obtaining a time weight value using a time weight function according to the personalized time data includes: the personalized time data includes the current time, the daytime activity cycle and each high-attention time period, and the obtaining of the time weight value using a time weight function according to the personalized time data, wherein the time weight function is: Where, f t (t) is the time weight value; t is the current moment; ρ d 、 is the diurnal activity cycle adjustment coefficient and phase, and ρ d ∈[0,1]; μ k is the central moment of the kth high attention period; σ k is the variance of the kth high attention period; K is the number of high attention periods; ρ k is the weight coefficient of the kth high attention period.

6. The data supervision method for the Internet of Things according to claim 3 is characterized in that: The method of obtaining the behavior preference data of the supervised object and obtaining the behavior preference value using a behavior preference function according to the behavior preference data includes: the behavior preference data includes the real-time location data, frequently visited places, and visit frequencies of the supervised object; obtaining the visit frequency weight of the frequently visited place cluster of the supervised object, the two-dimensional plane coordinates of the supervised object, and the centroid coordinates of the frequently visited place cluster through the behavior preference data; and obtaining the behavior preference value using the behavior preference function, wherein the behavior preference function is: Where, f h (H) is the behavior preference value; ω m is the visit frequency weight of the mth frequently visited location cluster; [x, y] is the two-dimensional coordinate of the supervised object at the current moment; τ m is the centroid coordinate of the mth frequently visited location cluster; δ m is the variance of the mth frequently visited location cluster; ‖·‖ is the Euclidean norm, which calculates the straight-line distance between plane coordinates.

7. The data supervision method for the Internet of Things according to claim 3 is characterized in that: Obtaining privacy-related data and obtaining privacy risk using a privacy data impact function based on the privacy-related data includes: the privacy-related data includes a privacy budget consumed by a supervised object within a preset time window, a total privacy budget allocated to the supervised object, and a query sensitivity factor, wherein the privacy data impact function is: Where E is the privacy risk; B current is the privacy budget consumed by the regulated object in the current time window; B total is the total privacy budget allocated to the regulated object; θ is the adjustment coefficient; S sensitivity is the query sensitivity factor.

8. The data supervision method for the Internet of Things according to claim 1 is characterized in that: The method of obtaining a final fence area according to the fence radius of each level and performing differential privacy enhancement processing on the supervision data within the final fence area includes: obtaining a final fence area according to the fence radius of each layer of fence, obtaining the supervision data and the current fence level of the supervised object according to the final fence area, obtaining the current privacy risk through the privacy data impact function, and performing differential privacy enhancement processing on the supervision data according to the current fence level of the supervised object and the current privacy risk.

9. The data supervision method for the Internet of Things according to claim 8 is characterized in that: The differentiated privacy enhancement processing of the supervision data according to the current fence level and current privacy risk of the supervised object includes: obtaining the total privacy budget allocated and the consumed privacy budget of this operation according to the current fence level and current privacy risk of the supervised object, and obtaining the noise addition amount using the noise addition formula, wherein the noise addition formula is: Where R is the amount of noise added; GlobalSensitivity (D) is the global sensitivity of query supervision data; D is the supervision data; ε alloc The total privacy budget allocated for this operation; ε used The privacy budget consumed.

10. A data supervision system for the Internet of Things, according to a data supervision method for the Internet of Things according to any one of claims 1 to 9, characterized in that: include: An architecture partitioning module is used to construct video fences to obtain a multi-level dynamic geo-fence structure; A fence radius calculation module, configured to obtain the fence radius of each level using a fence radius calculation formula according to the multi-level dynamic geo-fence structure; a first impact acquisition unit, configured to acquire personalized time data of a supervised object and obtain a time weight value using a time weight function according to the personalized time data; a second influence obtaining unit, configured to obtain behavior preference data of the supervised object, and obtain a behavior preference value using a behavior preference function according to the behavior preference data; a third impact acquisition unit, configured to acquire privacy-related data, and acquire privacy risks using a privacy data impact function according to the privacy-related data; The privacy differentiation processing module is used to obtain the final fence area according to the fence radius of each level, and perform differential privacy enhancement processing on the supervision data within the final fence area.

Images