An encryption protection-based internet of things collaborative sensing method

Abstract

The application relates to an encryption protection-based Internet of Things collaborative sensing method and belongs to the technical field of the Internet of Things.The method uses a full homomorphic encryption algorithm to ensure the safe protection of sensitive data in the process of communication transmission and cloud computing, combines multi-source factor data, and designs a graph convolutional neural network model framework to perform data prediction and abnormal source identification.The model fuses spatial environment topological structure and content features, and is beneficial to improving the scientific nature of collaborative sensing decisions in the Internet of Things scene.The application combines full homomorphic encryption technology and a graph convolutional neural network, solves the data security and sharing problems in the process of Internet of Things data integration processing, realizes the safe sharing application of cross-region sensing information, and improves the deep decision level of Internet of Things monitoring management.

Description

Technical Field

[0001] This invention belongs to the field of Internet of Things (IoT) technology and relates to an IoT collaborative sensing method based on encryption protection. Background Technology

[0002] With the deepening application of IoT technology in various sensing scenarios and professional fields, various data collection methods have gradually achieved comprehensive and real-time coverage. Due to the diverse, massive, and frequently updated nature of data, and the fact that it often involves cross-departmental application needs, data sharing and protection face significant challenges. The main problems with current monitoring data include: insufficient protection of sensitive information, high difficulty in data utilization, and inadequate understanding of data value. These problems limit the in-depth utilization and maximization of the value of monitoring data.

[0003] Currently, basic sensing and monitoring networks at all levels have been established. For example, a certain city has over 3,000 water conservancy monitoring points, generating terabytes of data daily. This data contains information on the layout and parameters of critical infrastructure; leaks could severely impact corporate interests. Furthermore, because the analysis of monitoring data often involves multiple units and regions, data sharing and utilization are difficult due to technological limitations, hindering further improvements in management efficiency.

[0004] Fully homomorphic encryption technology enables direct computation on encrypted data while ensuring data privacy. In continuous monitoring, collaborative analysis services can effectively address data sharing challenges. However, the application of homomorphic encryption in the monitoring field currently requires overcoming challenges related to algorithm efficiency and the customization of deep learning models. Different types of variables exhibit variations in their encrypted representation and computational patterns, necessitating further in-depth research.

[0005] In the field of sensing and monitoring, the dynamic changes of multi-source data exhibit multivariate and nonlinear characteristics, and artificial neural networks have good predictive capabilities for nonlinear problems. Although the deployment of automatic monitoring stations provides a large amount of high-frequency data, which reflects the dynamic changes of variables better than traditional sampling frequencies, neural network monitoring data analysis models based on spatial correlation are still relatively lacking.

[0006] Therefore, a new collaborative sensing method for the Internet of Things is urgently needed to solve the above problems. Summary of the Invention

[0007] In view of this, the purpose of this invention is to provide an IoT collaborative sensing method based on encryption protection. By combining fully homomorphic encryption technology with graph convolutional neural networks, this method solves the data security and sharing problems in the IoT data integration and processing process, realizes the secure sharing and application of cross-regional sensing information, and improves the level of in-depth decision-making in IoT monitoring and management.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] An IoT collaborative sensing method based on encryption protection is proposed. This method utilizes the fully homomorphic encryption (CKKS) algorithm to ensure the security of sensitive data during communication transmission and cloud computing. Simultaneously, it combines multi-source factor data and designs a graph convolutional neural network (GCNN) model framework for data prediction and anomaly source identification. This model integrates spatial environmental topology and content features, which is beneficial for improving the scientific rigor of collaborative sensing decisions in IoT scenarios.

[0010] The method specifically includes the following steps:

[0011] S1: On the remote terminal control system equipment (RTU) of the IoT monitoring station, it communicates with various types of sensors and collects monitoring data in the current scenario at a preset frequency. For example, monitoring data in the water conservancy environment includes water level, flow rate, water quality and seepage pressure parameters, etc. It also performs time-series synchronization and data fusion of monitoring data from multiple sensors to generate standardized time-point monitoring data.

[0012] S2: The collected raw monitoring data is format-converted and standardized, timestamps are unified and missing values ​​are filled, the monitoring data is aggregated by time series, a multi-dimensional monitoring dataset is constructed, and the dataset is downsampled and compressed to reduce the data volume. Then the processed data is deterministically verified for subsequent homomorphic encryption and calculation.

[0013] S3: Based on the preset storage strategy, the preprocessed monitoring dataset is sorted and divided by time, and stored in the local cache space of the RTU;

[0014] S4: Generate a public-private key pair using the CKKS algorithm, and use the public key to homomorphically encrypt the cached data to generate encrypted ciphertext data. The public-private key pair needs to be updated regularly, and the private key is stored on a cloud server for subsequent decryption operations. At the same time, perform modulus management and expansion on the ciphertext data to ensure that the encrypted data meets the numerical range requirements of homomorphic computation. Here, the CKKS algorithm represents a fully homomorphic encryption algorithm.

[0015] S5: Transmits homomorphically encrypted ciphertext data to the cloud server via a wireless network, encapsulates and encodes it according to the communication protocol, and transmits the ciphertext data in real time, ensuring that real-time data is transmitted first, and ensuring the reliability of transmission through a disconnection and reconnection mechanism in case of an anomaly.

[0016] S6: Receive the uploaded homomorphic encrypted ciphertext on the cloud server, and use the GCNN model to perform forward and backward propagation calculations on the monitoring data in the ciphertext domain to complete model training or inference, and finally generate the encrypted ciphertext of the prediction result; where GCNN stands for Graph Convolutional Neural Network.

[0017] S7: Transmits the ciphertext result obtained from cloud computing to the terminal via the network, decrypts the ciphertext using the private key, restores the decrypted data to the plaintext result, displays the decryption result report on the front-end page, records the decryption log, and monitors potential improper use.

[0018] Furthermore, in step S3, the cache space uses a circular buffer or a fixed cache structure to manage the data for each time slice, and regularly cleans up outdated data to ensure the utilization rate of the cache resources of the embedded device.

[0019] Furthermore, in step S4, the specific steps for encryption and decryption using the CKKS algorithm are as follows:

[0020] (1) Initialize basic parameters: given basic integers and initial modulus Initialize security parameters Power of the modulus of a polynomial and noise distribution parameters Perform data feature analysis on different types of data sources, including statistics such as mean, variance, maximum and minimum values, in order to select a suitable ciphertext modulus for each data type;

[0021] (2) Define adaptive parameters : Set adaptive parameters based on the characteristics of different data types This allows for a balance between noise growth and data accuracy during encryption; it can be defined based on factors such as variance and rate of change of different data types. ,For example: ;

[0022] in, It is an adjustment parameter used to control the overall range of the ciphertext modulus size. It is a data source The larger the standard deviation and variance, the better. The larger the value, the higher the modulus can be to provide a higher modulus for data with rapidly increasing noise.

[0023] (3) Define the adaptive modulus Based on adaptive parameters and number of encryption layers Set corresponding modulus sequences for different data sources. , It represents the current computational depth of a specific data source, and the modulus is dynamically adjusted based on its data characteristics and noise accumulation. For example, location data. : ,in This represents adaptive parameters for location data; , Indicates the maximum number of encryption layers globally;

[0024] This adaptive modulus setting ensures that a larger modulus is assigned to data that varies greatly (such as location data). This is to control the rapid increase of noise during the calculation process; for data with small variations, such as temperature, a smaller modulus is assigned. This reduces the computational overhead of encryption.

[0025] (4) Generate common parameters Based on the adaptive modulus sequence of each data source Construct adaptive common parameters It includes each adaptive modulus and basic encryption parameters; it assumes the polynomial modulus of the encryption ring. and ring size The generated common parameters include: ,in, Indicates the maximum number of encryption layers globally; The modulus sequence representing different data sources is used for noise control and adaptive parameter selection during subsequent encryption and computation processes;

[0026] (5) Perform key generation, encryption, and decryption operations: using predefined adaptive common parameters Generate keys, encrypt and decrypt data respectively:

[0027] Key generation: via Generate public key Private key and evaluation key :

[0028]

[0029] Encryption: Using public keys Encrypt different data sources to generate adaptive modulus ciphertext. :

[0030]

[0031] in, To represent different types of sensor data (such as temperature, humidity, location, device status, etc.), the appropriate modulus is selected for each type of data based on its characteristics. ;

[0032] Decryption: Using the private key For ciphertext Decrypt to obtain the original data :

[0033]

[0034] (6) Adaptive modulus adjustment in addition and multiplication operations: When performing addition and multiplication operations between different data sources, a suitable modulus is selected based on the adaptive modulus of the data. To control noise growth. For example, for addition operations on temperature and humidity data, choose a larger modulus for the humidity data to suppress noise growth:

[0035]

[0036] For multiplication operations, a data modulus with slower noise growth can be selected, or the size of the modulus can be dynamically adjusted to balance the problem of calculation accuracy and noise accumulation.

[0037] An adaptive polynomial parameter selection is proposed to improve the ciphertext modulus. It can dynamically adapt to the characteristics of different data sources in IoT scenarios to improve encrypted computing efficiency and data security.

[0038] Furthermore, in step S4, the CKKS algorithm, for information data collected in different IoT scenarios (e.g., water level, water quality, rainfall, and seepage pressure in a water conservancy scenario), requires functionalization representation of the data before encryption, mapping the numerical monitoring data to a finite-dimensional complex number domain. Using CKKS's default polynomial representation can more conveniently represent the changing patterns of various monitoring data, while also providing a more compact representation that easily regenerates the original curve, facilitating decryption, restoration, and multi-source data integration and analysis.

[0039] Furthermore, in step S6, the GCNN model based on the topological relationship of multiple monitoring points is constructed within the homomorphic ciphertext space, specifically including the following steps:

[0040] (1) Distribute IoT monitoring points in different locations in the collection area, collect geographic information data of different monitoring point locations, and construct a graph structure between monitoring points based on geographic information, where nodes represent monitoring points, edges represent the relationship between them, and the weight of the edges can reflect the similarity or distance between nodes.

[0041] Given Each monitoring point provides a graphical representation of the entire topological information. It means that among them It is a collection of monitoring points. This represents the set of edges between monitoring points;

[0042] (2) The data uploaded by monitoring points to the cloud server includes time-series data of important information indicators, such as dissolved oxygen, ammonia nitrogen, and pH value, which are water quality indicators in the water conservancy environment. A node feature matrix is ​​constructed from the information data indicator values ​​obtained from the most recent monitoring at each monitoring point. The adjacency matrix represents the topological relationships; the characteristic matrix... The representation is:

[0043]

[0044] in, Indicates monitoring point The feature vector contains data values ​​of different monitoring indicators;

[0045] (3) Since the Internet of Things (IoT) scenario is a real-time, dynamically changing system, according to the first law of geography, monitoring points have varying degrees of influence on each other. The following formula can be used to calculate the distance between two monitoring points. The reciprocal is used as the edge weight value and stored in the adjacency matrix. middle;

[0046]

[0047] in, , , ; This represents the height difference between two monitoring points. This is the vertical distance weighting factor, reflecting the degree of influence of vertical distance in the calculation; and These represent the difference in latitude and the difference in longitude, respectively. This represents a weighting factor used to dynamically adjust the importance of distance in specific IoT scenarios. It can be set according to the environmental characteristics of the monitoring point or the priority of the device. The radius of the Earth;

[0048] (4) The GCNN model utilizes the geographical location information and topological structure of the monitoring points. and characteristic matrix To train and learn a complex mapping function, thereby extracting spatial features;

[0049] In the GCNN model, node features are aggregated from the features of neighboring nodes and convolutional to generate higher-order feature representations; in graph convolution, nodes... The feature aggregation is represented as:

[0050]

[0051] in, It is a node In the Layer feature representation; It is a node ; and These are nodes and The degree; It is the first The layer's weight matrix is ​​used to learn feature maps; It is an activation function used to add non-linearity; It is the edge weight, representing the node weight. and The similarity relationship is the distance. The reciprocal of the distance is used to indicate that the smaller the distance, the greater the edge weight.

[0052] Graph convolution operations utilize the adjacency matrix. node feature matrix The convolution calculation is represented as follows:

[0053]

[0054] in, That is, adding self-loops to the adjacency matrix. It is the identity matrix; yes The degree matrix; It is the first The feature matrix of the layer; This indicates the number of long positions, i.e., the number of long positions aggregated. and They represent the first The adjacency matrix and adaptive edge weight coefficients of each head; Indicates the first Layer Each head uses an independent weight matrix to capture different feature representations. This formula achieves smooth propagation of features from each layer of nodes and extracts spatial features. At the same time, it uses a multi-head aggregation mechanism to learn and aggregate neighborhood information from multiple subspace perspectives, thereby improving the richness of feature representation and forming an enhanced feature representation.

[0055] Furthermore, in step S6, the GCNN model training is completed in the homomorphic encryption space, specifically including: calculating the loss function in the ciphertext space and maintaining the privacy of the loss value through homomorphic encryption; and performing model weighting through gradient descent in the ciphertext space. Update:

[0056]

[0057] in, For learning rate, This is the encrypted loss function.

[0058] The beneficial effects of this invention are as follows:

[0059] (1) The research on fully homomorphic encryption in this invention enables computation without exposing the original data, thereby effectively protecting the security of monitoring data. Storing monitoring data in the cloud and performing computations there has many advantages, such as resource sharing and elastic scaling. However, transmitting sensitive monitoring data to the cloud may pose a risk of information leakage. The fully homomorphic encryption in this invention provides a secure protection method for computation in the cloud, enabling useful computations without exposing plaintext data.

[0060] (2) The topological relationship graph of monitoring points constructed in this invention can more comprehensively capture the complexity and interrelationships of the sensing system. Training the graph convolutional neural network in the encrypted domain, and utilizing its ability on graph data, can comprehensively consider the topological structure between different monitoring points, which helps to more accurately understand the overall characteristics of the sensing system. In collaborative management, data cooperation between different units is involved. The research on fully homomorphic encryption in this invention helps to realize cross-unit data computation in an encrypted state, promote the sharing and cooperation of monitoring data, and protect the privacy of each unit.

[0061] (3) By performing calculations in an encrypted state, this invention enables more in-depth analysis of monitoring data in the cloud, providing more accurate decision support. The collaborative analysis of the spatiotemporal attribute relationships between different monitoring points is of great significance for the planning, management, and prediction of Internet of Things (IoT) technology and engineering.

[0062] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0063] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0064] Figure 1 The overall flowchart of the IoT collaborative sensing method based on encryption protection provided by the present invention;

[0065] Figure 2 The following is an application flowchart of the CKKS fully homomorphic encryption algorithm provided by this invention;

[0066] Figure 3The diagram illustrates the Laplacian matrix generation process of the GCNN model provided in this invention. Detailed Implementation

[0067] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0068] This invention will describe the implementation steps for water quality environmental monitoring in a water conservancy IoT data collection scenario. The objects to be encrypted include various water conservancy data, such as water level, water quality, rainfall, and seepage pressure. Before encryption, the data needs to be functionalized to map the numerical monitoring data to a finite-dimensional complex number domain. Using CKKS's default polynomial representation can more conveniently represent the changing patterns of various monitoring data, while also providing a more compact representation that easily regenerates the original curves, facilitating decryption, restoration, and multi-source data integration and analysis.

[0069] Please see Figures 1-2 This invention provides a system for fully homomorphic encryption using the CKKS algorithm. The system is primarily based on an edge computing acquisition terminal, deploying the encryption algorithm and model framework, and combining cloud computing to complete training and prediction of IoT sensing data under multi-monitoring point collaboration. It includes a data acquisition module, a data processing module, a homomorphic encryption module, and a homomorphic computing module, divided into edge computing and cloud computing aspects. It forms a feature representation method for the sensing scene in a homomorphic space, applies the fully homomorphic encryption algorithm (CKKS) suitable for IoT data, and simultaneously extracts geographical information and data features from multiple IoT monitoring points to construct topological structure and parameter attributes, training a graph convolutional neural network model (GCNN) based on the monitoring point topological spatial graph.

[0070] The process steps of the above system are mainly divided into: encoding, encryption, homomorphic computation, decryption and decoding.

[0071] S1: First, define the order of the polynomial. This means that the polynomial only contains elements from... Step to The coefficients of each term of the order. For each monitored value. Calculate its value in the standard basis polynomial. The coefficients in the set are combined to form a coefficient vector and a coefficient matrix. As the input quantization matrix of the encryption system.

[0072] S2: Given security parameters Choose two powers of integers , respectively in The above shows the distribution of keys, error learning, and encryption settings. For a fundamental integer and number of levels and a basic module Set the modulus of the ciphertext For the ring This place makes Then, parameters are randomly generated. Output: .

[0073] The selection of security parameters affects the cost and capability of homomorphic computation, among which the number of levels... In other words, `setLevel` controls the number of bits in the encrypted number. A larger value results in stronger homomorphic computation capabilities, but also increases computational costs. It's typically chosen to be between 30 and 60 bits, defining the maximum precision depth supported by the system. Each level... Generated modulus ModulusQ describes the magnitude of the operand modulus; a larger value results in smaller computational errors but also higher computational costs. Lower values ​​for setLevel and modulus can accelerate computation, but homomorphic operations may fail prematurely due to overmodulation. Higher parameters provide stronger homomorphic computational power and deeper computation, but the cost increases exponentially. Therefore, in practice, these two core parameters, along with other indicators, need to be dynamically adjusted based on model complexity and data scale to balance safety, correctness, and computational cost.

[0074] S3: Key Generation Specifically, it includes:

[0075] Randomly generated and set the private key .

[0076] Randomly generated and errors Set the public key: .

[0077] Randomly generated and Set the evaluation key: .

[0078] S4: Encryption: Specifically, it includes:

[0079] Randomly generated as well as Output ciphertext ,and :

[0080]

[0081] S5: Decryption: Specifically, it includes:

[0082] For the same level The ciphertext is decrypted to obtain the plaintext result:

[0083]

[0084] S6: Addition: Specifically, it includes:

[0085] For ciphertext The result of adding their ciphertexts is

[0086]

[0087] S7: Multiplication: Specifically, it includes:

[0088] For ciphertext :

[0089]

[0090]

[0091] The result of multiplying the ciphertext is

[0092]

[0093] S8: The plaintext functional coefficient matrix is ​​encrypted into a ciphertext tensor form supported by the CKKS system using the public key. Then, the CKKS.Modulo operation in the CKKS system is used to perform modular truncation on the encrypted tensor. The main purpose is to control the accumulation of floating-point errors and maintain the reversibility of the calculation. Repeating this operation is an important means of controlling floating-point errors.

[0094] S9: In the key management section, the public and private key pairs are generated by the acquisition device through a random algorithm. The public key is used by the acquisition device to encrypt the raw data to be uploaded, while the private key is securely stored by the cloud computing analysis end. The corresponding private key is selected according to the device identification information and decrypted when needed. To ensure the long-term security of the key, the key can be updated and regenerated periodically.

[0095] Please see Figure 3Water monitoring stations are distributed across different river basins. The water data uploaded from these stations to the cloud includes time-series data of important water quality indicators, such as dissolved oxygen, ammonia nitrogen, and pH. Geographic information data of the different monitoring stations' basins is collected, and a graph structure is constructed based on this geographic information. Nodes represent monitoring stations, edges represent relationships between them, and edge weights reflect the similarity or distance between nodes. The water quality indicator values ​​obtained from the most recent monitoring at each station constitute a node feature vector, and an adjacency matrix represents the topological relationships.

[0096] Define a signal map for each monitoring point. The signal is represented by a vector and includes characteristic indicators related to water quality prediction at that monitoring point, using a feature matrix. This represents the signal at each monitoring point. The historical time series length is determined by... Provide, that is, when hour, Indicates the selected monitoring watershed Each monitoring point at time A set of current values ​​for different water quality indicators at a given location.

[0097] Graph Convolutional Neural Network (GCNN) models utilize the geographical location information and topological structure of monitoring points. and characteristic matrix The main steps for training a complex mapping function to extract spatial features are as follows:

[0098] M1. Construct a topology map based on the geographical locations of the monitoring points. and adjacency matrix .

[0099] M2. Construct a feature matrix using water quality index values ​​from water monitoring data at various points. .

[0100] M3, based on adjacency matrix Perform a Laplace transform to generate the Laplace transform matrix. , can be represented as:

[0101]

[0102] in, It is a self-connected adjacency matrix. It is the identity matrix. It is a degree matrix.

[0103] M4. Based on graph convolution operations, the model uses encrypted data of the adjacency matrix and node feature matrix for convolution calculations and extracts water quality features. The convolution operation formula in the encrypted space is as follows:

[0104]

[0105] in, This is the activation function, used to implement nonlinear transformations. Indicates the first The weight matrix of the layer, Indicates the first The activation value of the layer, and In the homomorphic encryption space, the adjacency matrix and characteristic matrix All are encrypted, among which This represents the encrypted matrix or vector. An approximate activation function in the ciphertext space is used to achieve nonlinear feature transformation, which is then used for nonlinear processing of the encrypted data. GCNN then uses the Laplacian matrix to dynamically capture the influence weights of different monitoring points, thereby obtaining spatial features.

[0106] Based on the above four steps, a topology map is constructed using the geographical location information of monitoring points in the target watershed. Then, the corresponding adjacency matrix, degree matrix, and Laplace matrix are calculated and generated based on the topology map, such as... Figure 3 As shown.

[0107] M5. Model training is performed in a homomorphic encrypted space to ensure that raw data is not leaked when water quality predictions are executed in the cloud. The loss function is calculated in the encrypted space, and the privacy of the loss value is maintained through homomorphic encryption. Model weights are then determined using gradient descent in the encrypted space. Update:

[0108]

[0109] in, For learning rate, This is the encrypted loss function. After the model is trained, predictions are made on new encrypted water quality data to generate predicted values ​​for future water quality conditions. After decryption, the prediction results are used for analysis and early warning.

[0110] The graph incorporates link predictions to analyze possible water quality propagation paths between different monitoring points, detect abnormal changes, identify propagation and correlation between nodes, and identify monitoring points that may be pollution sources. For example, if the predicted values ​​of individual monitoring points or areas deviate significantly, they may be located at the source of water pollution or within the affected area.

[0111] Specifically, in terms of implementation, CrypTen is a privacy-preserving machine learning framework based on PyTorch. Its goal is to allow machine learning practitioners to use secure computing technologies, significantly simplifying the development of homomorphic computation using CKKS, while also meeting the needs of building neural networks and applying deep learning within the homomorphic space on cloud servers. CrypTen provides an object-oriented programming model, using Modules and Tensors to define model structures, and includes built-in implementations of common layers and operators, such as convolutional layers and fully connected layers. Combining CrypTen with CKKS enables convenient and efficient deep learning prediction of water quality under homomorphic encryption.

[0112] This embodiment provides an IoT collaborative sensing method based on encryption protection. By combining fully homomorphic encryption technology and graph convolutional neural networks, this method realizes the propagation and fusion of hydrological features throughout the entire spatial region, and delves deeper into the inherent laws of water quality evolution from a global perspective. While protecting data security, it can also predict the dynamic changes in water quality and the sources of water pollution, thus promoting the realization of hydrological security sharing.

[0113] Finally, it should be noted that the above 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A collaborative sensing method for the Internet of Things (IoT) based on encryption protection, characterized in that, The method specifically includes the following steps: S1: On the remote terminal control system equipment of the IoT monitoring station, it communicates with various types of sensors, collects monitoring data under the current scenario at a preset frequency, and performs time-series synchronization and data fusion of the monitoring data of multiple sensors to generate standardized time-point monitoring data. S2: The collected raw monitoring data is format-converted and standardized, timestamps are unified and missing values ​​are filled, the monitoring data is aggregated by time series, a multi-dimensional monitoring dataset is constructed, the dataset is downsampled and compressed, and then the processed data is deterministically verified. S3: Based on the preset storage strategy, the preprocessed monitoring dataset is sorted and divided by time, and stored in the local cache space of the remote terminal control system equipment; S4: Generate a public-private key pair using the CKKS algorithm, and use the public key to homomorphically encrypt the cached data to generate encrypted ciphertext data. The public-private key pair needs to be updated regularly, and the private key is stored on a cloud server for subsequent decryption operations. At the same time, perform modulus management and expansion on the ciphertext data to ensure that the encrypted data meets the numerical range requirements of homomorphic computation. Here, the CKKS algorithm represents a fully homomorphic encryption algorithm. S5: Transmits homomorphically encrypted ciphertext data to the cloud server via a wireless network, encapsulates and encodes it according to the communication protocol, and transmits the ciphertext data in real time, ensuring that real-time data is transmitted first, and ensuring the reliability of transmission through a disconnection and reconnection mechanism in case of an anomaly. S6: Receive the uploaded homomorphic encrypted ciphertext on the cloud server, and use the GCNN model to perform forward and backward propagation calculations on the monitoring data in the ciphertext domain to complete model training or inference, and finally generate the encrypted ciphertext of the prediction result; where GCNN stands for Graph Convolutional Neural Network. S7: Transmits the ciphertext result obtained from cloud computing to the terminal via the network, decrypts the ciphertext using the private key, restores the decrypted data to the plaintext result, displays the decryption result report on the front-end page, records the decryption log, and monitors potential improper use.

2. The IoT collaborative sensing method based on encryption protection according to claim 1, characterized in that, In step S3, the cache space uses a circular buffer or a fixed cache structure to manage the data for each time slice, and regularly cleans up old data to ensure the utilization rate of the cache resources of the embedded device.

3. The IoT collaborative sensing method based on encryption protection according to claim 1, characterized in that, In step S4, the specific steps for encryption and decryption using the CKKS algorithm are as follows: (1) Initialize basic parameters: given basic integers and initial modulus Initialize security parameters Power of the modulus of a polynomial and noise distribution parameters Perform data feature analysis on different types of data sources, including mean, variance, maximum and minimum values, in order to select a suitable ciphertext modulus for each data type; (2) Define adaptive parameters : Set adaptive parameters based on the characteristics of different data types In order to balance noise growth and data accuracy during the encryption process; (3) Define the adaptive modulus Based on adaptive parameters and number of encryption layers Set corresponding modulus sequences for different data sources. , It represents the current computation depth of a data source, and the modulus is dynamically adjusted based on its data characteristics and noise accumulation. (4) Generate common parameters Based on the adaptive modulus sequence of each data source Construct adaptive common parameters It includes each adaptive modulus and basic encryption parameters; it assumes the polynomial modulus of the encryption ring. and ring size The generated common parameters include: ,in, Indicates the maximum number of encryption layers globally; The modulus sequence representing different data sources is used for noise control and adaptive parameter selection during subsequent encryption and computation processes; (5) Perform key generation, encryption, and decryption operations: using predefined adaptive common parameters Generate keys, encrypt and decrypt data respectively: Key generation: via Generate public key Private key and evaluation key : Encryption: Using public keys Encrypt different data sources to generate adaptive modulus ciphertext. : in, This represents different types of sensor data, with each type of data having a corresponding modulus selected based on its characteristics. ; Decryption: Using the private key For ciphertext Decrypt to obtain the original data : (6) Adaptive modulus adjustment in addition and multiplication operations: When performing addition and multiplication operations between different data sources, a suitable modulus is selected based on the adaptive modulus of the data. To control noise growth.

4. A method for collaborative sensing of the Internet of Things based on encryption protection according to claim 1 or 3, characterized in that, In step S4, the CKKS algorithm requires that the data to be encrypted be collected from different IoT scenarios. Before encryption, the data needs to be represented by a functional, which maps the numerical monitoring data to a finite-dimensional complex number domain.

5. The IoT collaborative sensing method based on encryption protection according to claim 1, characterized in that, In step S6, the GCNN model based on the topological relationship of multiple monitoring points is constructed in the homomorphic ciphertext space, which specifically includes the following steps: (1) Distribute IoT monitoring points in different locations in the collection area, collect geographic information data of different monitoring point locations, and construct a graph structure between monitoring points based on geographic information, where nodes represent monitoring points, edges represent the relationship between them, and the weight of the edges reflects the similarity or distance between nodes. Given Each monitoring point, the entire topological information is presented graphically. It means that among them It is a collection of monitoring points. This represents the set of edges between monitoring points; (2) The node feature matrix is ​​composed of the information data index values ​​obtained from the most recent monitoring of each monitoring point. The adjacency matrix represents the topological relationships; the characteristic matrix... The representation is: in, Indicates monitoring point The feature vector contains data values ​​of different monitoring indicators; (3) Use the following formula to calculate the distance between the two monitoring points. The reciprocal is used as the edge weight value and stored in the adjacency matrix. middle; in, , , ; This represents the height difference between two monitoring points. This is the vertical distance weighting factor, reflecting the degree of influence of vertical distance in the calculation; and These represent the difference in latitude and the difference in longitude, respectively. This represents a weighting factor used to dynamically adjust the importance of distance in specific IoT scenarios, and is set according to the environmental characteristics of the monitoring point or the priority of the device. The radius of the Earth; (4) The GCNN model utilizes the geographical location information and topological structure of the monitoring points. and characteristic matrix To train and learn a complex mapping function, thereby extracting spatial features; In the GCNN model, node features are aggregated from the features of neighboring nodes and convolutional to generate higher-order feature representations; in graph convolution, nodes... The feature aggregation is represented as: in, It is a node In the Layer feature representation; It is a node ; and These are nodes and The degree; It is the first The layer's weight matrix is ​​used to learn feature maps; It is an activation function used to add non-linearity; It is the edge weight, representing the node weight. and The similarity relationship is the distance. The reciprocal of the distance is used to indicate that the smaller the distance, the larger the edge weight. Graph convolution operations utilize the adjacency matrix. node feature matrix The convolution calculation is represented as follows: in, That is, adding self-loops to the adjacency matrix. It is the identity matrix; yes The degree matrix; It is the first The feature matrix of the layer; This indicates the number of long positions, i.e., the number of long positions aggregated. and They represent the first The adjacency matrix and adaptive edge weight coefficients of each head; Indicates the first Layer Each head has a weight matrix, and each head captures a different feature representation through an independent weight matrix.

6. The IoT collaborative sensing method based on encryption protection according to claim 5, characterized in that, In step S6, the GCNN model training is completed in the homomorphic encryption space, specifically including: calculating the loss function in the ciphertext space and maintaining the privacy of the loss value through homomorphic encryption; and performing model weighting through gradient descent in the ciphertext space. Update: in, For learning rate, This is the encrypted loss function.

Images