Method, system, medium and server for protecting data security of underwater internet of things facility

Abstract

The application provides an underwater Internet of Things facility data security protection method, system, medium and server, relates to the technical field of underwater Internet of Things data security, and the method comprises the following steps: deploying corresponding edge nodes on underwater Internet of Things facilities, collecting environmental data of the facilities by using the edge nodes, calculating the underwater environmental risk degree of all underwater Internet of Things facilities based on the environmental data, judging whether the kth underwater environmental risk degree is greater than a preset warning threshold, if yes, acquiring running data stored in the kth underwater Internet of Things facility and encrypting, and transmitting the encrypted data to a server, if no, updating the k+1th underwater environmental risk degree to the kth underwater environmental risk degree, and continuing to perform the risk assessment step until all underwater environmental risk degrees are traversed. The application can improve the protection degree of the data stored in the underwater Internet of Things facility.

Description

Technical Field

[0001] This application relates to the technical field of underwater Internet of Things (IoT) data security, and in particular to methods, systems, media, and servers for protecting the data security of underwater IoT facilities. Background Technology

[0002] Against the backdrop of a growing global reliance on marine resources, underwater Internet of Things (IoT) technology has entered a golden age of rapid development. Numerous sensing devices are being precisely deployed across the vast underwater space, collectively constructing a massive underwater IoT system. These underwater IoT facilities act as intelligent sentinels in the ocean, playing an irreplaceable and crucial role in several key areas, including marine resource monitoring and management, and marine economic development.

[0003] In practical applications, some underwater IoT facilities incorporate underwater drifting observers into their components to acquire broader and more dynamic underwater environmental data. These observers act like drifting messengers in the ocean, cleverly using the power of water currents to move freely underwater, continuously collecting valuable information from different locations in the deep ocean, providing solid data support for in-depth scientific research and efficient development of marine resources.

[0004] These underwater IoT facilities operate in extremely harsh environments, where pressure, temperature, and depth data are constantly changing. Water pressure increases with depth, exerting immense stress on the physical structure and electronic components of the equipment, potentially causing deformation or even damage to the casing, affecting the accuracy of data acquisition and the normal operation of the equipment. Significant fluctuations in seawater temperature, from the cold deep sea to the warm shallow sea, can interfere with the stability of the equipment's circuitry and the accuracy of its sensors. Furthermore, complex hydrological conditions at different depths, such as currents and tides, pose significant challenges to the stable operation of underwater IoT facilities. If the harsh environmental conditions damage the underwater IoT facilities, the data stored within them may be lost, causing incalculable losses to related work. Summary of the Invention

[0005] To improve the protection of data stored internally in underwater IoT facilities, this application provides a method, system, medium, and server for data security protection of underwater IoT facilities.

[0006] Firstly, this application provides a method for data security protection of underwater Internet of Things (IoT) facilities, employing the following technical solution:

[0007] A method for protecting the data security of underwater IoT facilities, which includes multiple underwater IoT facilities, comprising the following steps:

[0008] Data Acquisition: Deploy corresponding edge nodes on underwater IoT facilities, use edge nodes to collect environmental data of the location of the i-th underwater IoT facility, calculate the underwater environmental risk level of the i-th underwater IoT facility based on the environmental data, and obtain the underwater environmental risk level of all underwater IoT facilities. The environmental data includes pressure data, temperature data and depth data.

[0009] Risk assessment: Determine whether the risk level of the k-th underwater environment exceeds the preset warning threshold.

[0010] If so, then obtain the operation data and target data stored in the kth underwater IoT facility, encrypt the operation data and target data, and transmit the encrypted operation data and target data to the server;

[0011] If not, update the (k+1)th underwater environmental risk level to the kth underwater environmental risk level, and perform the risk assessment steps until all underwater environmental risk levels have been traversed.

[0012] This application first deploys edge nodes on underwater IoT facilities and directly collects environmental data from these facilities through these edge nodes, achieving localized environmental data collection, which helps reduce data transmission latency and improve the real-time performance of data collection. Each edge node in this application is only responsible for collecting environmental data from its corresponding underwater IoT facility. This one-to-one collection mode improves the relevance of the data and facilitates independent risk assessments for each underwater IoT facility.

[0013] This application calculates the underwater environmental risk level for each underwater IoT facility and determines whether the risk level exceeds a preset warning threshold, which distinguishes whether a facility is in a risky state. When the underwater environmental risk level exceeds the preset warning threshold, the application acquires the operational data and target data stored within that underwater IoT facility. The operational data includes various operational parameters and status information of the underwater IoT facility, indicating its operational status. The target data refers to confidential data related to the tasks performed by the underwater IoT facility that needs to be transmitted back. Subsequently, this application encrypts the operational data and target data before transmitting the encrypted data to a server. Encryption improves data security during transmission and reduces the risk of data leakage or tampering. When the underwater environmental risk level does not exceed the preset warning threshold, the application proceeds to assess the next underwater environmental risk level until all underwater environmental risk levels have been traversed.

[0014] By adopting the above-mentioned scheme, this application can respond to changes in the underwater environment in real time, promptly detect underwater IoT facilities with risks, and encrypt the transmission of operational data and target data in the event of risks, thereby reducing the risk of the operational data and target data being stolen or tampered with during transmission and improving data security.

[0015] Optionally, the method further includes:

[0016] Collect real-time operational data: Record the operational data stored in each underwater IoT facility as the historical operational data of each underwater IoT facility, and collect the real-time operational data of each underwater IoT facility.

[0017] Modeling and Training: Based on the federated learning framework, local models of various underwater IoT facilities and global models of servers are constructed. The local models are trained using labeled historical operating data to obtain the trained local models.

[0018] Update parameters: Obtain the parameters of the trained local model, denoted as local parameters, encrypt the local parameters, and upload the encrypted local parameters to the server. The server aggregates the encrypted local parameters to generate global parameters for the global model. The server encrypts the global parameters and distributes them to each underwater IoT facility. Each underwater IoT facility decrypts the encrypted global parameters and updates the local parameters in the trained local model according to the decrypted global parameters to obtain the updated local model.

[0019] Classification: Input the real-time running data into the updated local model to obtain the classification results.

[0020] This application records the operational data stored within the underwater IoT facility as historical operational data and collects real-time operational data. By simultaneously acquiring both historical and real-time operational data, the operational status of the underwater IoT facility can be comprehensively reflected from multiple dimensions. Historical operational data helps analyze the long-term operational trends and potential problems of the underwater IoT facility, while real-time operational data can capture instantaneous operational changes of the underwater IoT facility, providing support for the detection of anomalies and malfunctions.

[0021] Subsequently, this application constructs local models for each underwater IoT facility and a global model for the server based on a federated learning framework, fully leveraging the advantages of federated learning in protecting data privacy and enabling distributed learning. Each underwater IoT facility can train its model locally using its internally stored operational data, eliminating the need to repeatedly upload internally stored operational and target data to the server. By reducing the number of times internally stored data is transmitted externally, data security and privacy are effectively protected. Simultaneously, by aggregating and updating the global model through the server, collaborative learning among the various local models is achieved, improving the overall model performance.

[0022] Subsequently, this application trains the local model using labeled historical running data. The labels help the local model learn classification features under different running states, thereby improving the local model's accuracy in classifying running data. Through training, the local model is able to better classify real-time running data.

[0023] Subsequently, this application obtains the parameters of the trained local model (i.e., local parameters) and encrypts them. The encrypted local parameters are then uploaded to the server, improving the security of the local parameters during transmission. Encryption effectively reduces the risk of local parameters being stolen or tampered with, thus enhancing security. The server then aggregates the encrypted local parameters to generate global parameters for the global model, thereby achieving effective integration of the various local parameters.

[0024] Subsequently, the server encrypts the global parameters and distributes them to each underwater IoT facility. Each underwater IoT facility decrypts the encrypted global parameters and updates its local parameters in its trained local model based on the decrypted global parameters, resulting in an updated local model. This allows each local model to promptly absorb the knowledge and experience of the global model while maintaining its own unique characteristics (i.e., the local model only processes the real-time operational data of its corresponding underwater IoT facility). The updated local model can better classify real-time operational data, improving the accuracy of the overall classification of operational data for the underwater IoT facilities.

[0025] Finally, this application inputs real-time operational data into the updated local model to obtain classification results, achieving rapid classification and analysis of real-time operational data of underwater IoT facilities. This provides timely and accurate information support for the operation management and decision-making of underwater IoT facilities. While protecting the data privacy of each underwater IoT facility, this application utilizes a federated learning framework to train and update a distributed model (i.e., a local model), further enhancing the security of data stored internally within underwater IoT facilities.

[0026] Optionally, the method further includes:

[0027] Similarity assessment: Calculate the similarity between the real-time operational data of the i-th underwater IoT facility and each historical operational data point of the i-th underwater IoT facility, and determine whether the similarity exceeds a preset similarity threshold.

[0028] If so, the real-time running data will be input into the trained local model to obtain the output result of the trained local model;

[0029] If not, then proceed with the parameter update step.

[0030] This application calculates the similarity between the real-time operational data and each historical operational data point of the i-th underwater IoT facility, obtaining multiple similarity scores. Then, it determines whether each similarity score exceeds a preset similarity threshold, thus achieving preliminary processing of the real-time operational data. When the similarity score exceeds the threshold, it indicates that the real-time operational data and historical operational data are similar in features and are in similar operational states or scenarios. The real-time operational data can then be input into a trained local model, leveraging the model's ability to process known similar data to quickly obtain output results, improving processing efficiency and reducing computational and parameter update operations.

[0031] When the similarity is not greater than the threshold, it indicates that there is a significant difference between the real-time operating data and the historical operating data, and the data may be in a new operating state or facing new environmental conditions. At this time, the step of updating parameters can update the local model to adapt to the new data characteristics, so that the local model can maintain good performance and accuracy when facing various situations, thereby improving the stability and reliability of the entire underwater IoT facility operation and management.

[0032] Optionally, before performing the similarity determination step, the method further includes:

[0033] Clustering algorithms are used to cluster the historical operational data of each underwater IoT facility to obtain normal clusters and abnormal clusters. The centroid of the normal cluster is recorded as the first data and the centroid of the abnormal cluster is recorded as the second data. The Euclidean distance between the real-time operational data and the first data is calculated and recorded as the third data. The Euclidean distance between the real-time operational data and the second data is calculated and recorded as the fourth data.

[0034] Determine if the third data is less than the fourth data.

[0035] If so, delete the historical runtime data contained within the abnormal cluster;

[0036] If not, delete the historical runtime data contained in the normal cluster.

[0037] This application first uses a clustering algorithm to cluster the historical operation data of each underwater IoT facility to obtain normal clusters and abnormal clusters. Through clustering, this application can group historical operation data with similar characteristics into one category and distinguish the normal operation status and abnormal operation status of underwater IoT facilities.

[0038] Subsequently, this application designates the centroid of the normal cluster as the first data and the centroid of the abnormal cluster as the second data. The centroid, as a representative point of the data within a cluster, extracts the characteristics of normal and abnormal operating states. Centroid extraction simplifies and improves the efficiency of subsequent Euclidean distance calculations while preserving key feature information of the data within the cluster. This helps to quickly determine the similarity between real-time operating data and normal or abnormal operating states, avoiding blind searching and comparison in large amounts of data and improving the efficiency and accuracy of data processing. When the Euclidean distance between the real-time operating data and the first data is less than the Euclidean distance between the real-time operating data and the second data, it indicates that the real-time operating data is more similar to historical operating data within the normal cluster. Deleting historical operating data contained in the abnormal cluster reduces the computational load in the similarity judgment step, further improving processing efficiency. Conversely, this application removes historical operating data contained in the normal cluster to further improve processing efficiency. Through the above data filtering operations, this application removes historical operating data that does not conform to the characteristics of the current real-time operating data, improving computational efficiency and reducing the consumption of computing resources.

[0039] Optionally, the calculation model for the underwater environmental risk level is as follows:

[0040] ;

[0041] Where R is the numerical value of the underwater environmental risk level; Weights for stress data; Pressure data for underwater IoT facilities; The pressure withstand threshold for underwater IoT facilities; Weights for temperature data; Temperature data for underwater IoT facilities; The optimal operating temperature for underwater IoT facilities; The maximum temperature that underwater IoT facilities can withstand; The minimum temperature that underwater IoT facilities can withstand; Weights for depth data; Depth data for underwater IoT facilities; This refers to the maximum depth that underwater IoT facilities can withstand.

[0042] Optionally, the method further includes:

[0043] The correlation between temperature and depth data is calculated using the Pearson correlation coefficient algorithm. Based on this correlation, an intersection term is introduced into the underwater environmental risk level calculation model to obtain an extended underwater environmental risk level calculation model. The underwater environmental risk level of the i-th underwater IoT facility is then recalculated using the extended underwater environmental risk level calculation model and recorded as the new underwater environmental risk level. Finally, the risk assessment steps are performed.

[0044] This application employs the Pearson correlation coefficient algorithm to calculate the correlation between temperature and depth data, enabling precise numerical quantification of the strength and direction of the relationship between these two variables. The Pearson correlation coefficient ranges from -1 to 1, and its specific value provides a clear understanding of whether the temperature and depth data of the underwater IoT facility's environment are positively, negatively, or uncorrelated, as well as the degree of correlation.

[0045] Subsequently, based on the calculated correlation, this application introduces interaction terms into the calculation model for underwater environmental risk levels, resulting in an extended calculation model. The introduction of interaction terms allows the calculation model to consider the interaction between temperature and depth variables, rather than simply considering the impact of each variable on the risk level individually. The extended calculation model can more comprehensively capture this complex variable relationship, enhancing its expressive power and enabling it to more accurately describe the degree of underwater environmental risk.

[0046] Subsequently, this application recalculated the underwater environmental risk level of the i-th underwater IoT facility using an extended underwater environmental risk level calculation model, denoted as the new underwater environmental risk level. The new risk level more accurately reflects the current environmental risk status of the underwater IoT facility. Performing a risk assessment after obtaining the new underwater environmental risk level provides a clear decision-making basis for the operation and management of the underwater IoT facility. The risk assessment results can help managers understand the level of risk faced by the facility and take corresponding measures based on the risk level.

[0047] Optionally, the method further includes:

[0048] A sensitivity matrix is ​​constructed based on the extended underwater environmental risk level calculation model and environmental data. The sensitivity value of the i-th underwater IoT facility is calculated based on the sensitivity matrix. The element in the i-th row and j-th column of the sensitivity matrix represents the partial derivative of the extended underwater environmental risk level calculation model of the i-th underwater IoT facility with respect to the j-th environmental data.

[0049] The false alarm rate is statistically analyzed. Based on the sensitivity value and the false alarm rate, a new warning threshold is calculated using the Youden index method. The risk assessment steps are then re-executed based on the new warning threshold.

[0050] This application constructs a sensitivity matrix based on an extended underwater environmental risk level calculation model and environmental data. The element in the i-th row and j-th column of the sensitivity matrix represents the partial derivative of the extended underwater environmental risk level calculation model for the i-th underwater IoT facility with respect to the j-th environmental data. The introduction of partial derivatives accurately quantifies the sensitivity of the calculation model to changes in each environmental data point. The sensitivity matrix visually displays the sensitivity of different underwater IoT facilities to various environmental data points. By viewing the sensitivity matrix, it is possible to quickly compare the sensitivity of different underwater IoT facilities to the same environmental data, as well as the differences in sensitivity of the same underwater IoT facility to different environmental data points, helping researchers and managers better understand the complex relationship between underwater IoT facilities and the environment.

[0051] Subsequently, this application calculates the sensitivity value of the i-th underwater IoT facility based on the sensitivity matrix and calculates the false alarm rate. Then, based on the sensitivity value and false alarm rate, a new warning threshold is calculated using the Youden index method. This comprehensively considers the sensitivity and false alarm situation of the underwater IoT facility, thus optimizing the warning threshold. Traditional warning thresholds do not fully consider the sensitivity of underwater IoT facilities to environmental changes, leading to a high number of false alarms in practical applications. By introducing a sensitivity value, the new warning threshold can more accurately reflect the risk level of underwater IoT facilities under different environmental conditions, thereby reducing the false alarm rate and improving the reliability and effectiveness of warning information.

[0052] Subsequently, this application re-executes the risk assessment step based on the new warning threshold, which enables the risk assessment results to be more accurate. The new warning threshold can better distinguish between normal and risky situations, reduce over-intervention or ignoring of true risks due to false alarms, and reduce unnecessary waste of resources.

[0053] This application enables dynamic adjustment of warning thresholds based on actual environmental data and false alarm occurrences, enhancing its adaptability to different environments and operating conditions. Simultaneously, by introducing mathematical models and calculation methods, it achieves quantitative analysis and optimization of the sensitivity and false alarm rate of underwater IoT facilities, improving the level of intelligence and thus enabling more intelligent responses to complex underwater environmental risks.

[0054] Secondly, this application provides a data security protection system for underwater IoT facilities, which adopts the following technical solution:

[0055] Underwater IoT facility data security protection system. Underwater IoT includes multiple underwater IoT facilities, including:

[0056] The data acquisition module is used to deploy corresponding edge nodes on underwater IoT facilities, collect environmental data of the location of the i-th underwater IoT facility using the i-th edge node, calculate the underwater environmental risk level of the i-th underwater IoT facility based on the environmental data, and obtain the underwater environmental risk level of all underwater IoT facilities. The environmental data includes pressure data, temperature data and depth data.

[0057] The risk assessment module is used to determine whether the risk level of the k-th underwater environment exceeds a preset warning threshold.

[0058] If the risk level of the kth underwater environment is greater than the preset warning threshold, the operation data and target data stored in the kth underwater IoT facility are obtained, the operation data and target data are encrypted, and the encrypted operation data and target data are transmitted to the server.

[0059] If the risk level of the kth underwater environment is not greater than the preset warning threshold, then the risk level of the (k+1)th underwater environment is updated to the risk level of the kth underwater environment, and the risk assessment module is triggered until all underwater environmental risk levels have been traversed.

[0060] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when processed by a processor, is used to implement the method described in the first aspect.

[0061] Fourthly, this application provides a server on which the system described in the second aspect is loaded, the server comprising: a processor, and a memory communicatively connected to the processor;

[0062] The memory is provided with the computer-readable storage medium described in the third aspect, and the computer-readable storage medium stores a computer program.

[0063] When the processor processes a computer program stored on the computer-readable storage medium, it implements the method described in the first aspect.

[0064] In summary, this application includes at least one of the following beneficial technical effects:

[0065] 1. This application can respond to changes in the underwater environment in real time, promptly detect underwater IoT facilities with risks, and encrypt the transmission of operational and target data in the event of risks, thereby reducing the risk of theft or tampering of operational and target data during transmission and improving data security.

[0066] 2. This application, while protecting the data privacy of each underwater IoT facility, implements the training and updating of distributed models (i.e., local models) through a federated learning framework, further improving the security of data stored internally by underwater IoT facilities. Attached Figure Description

[0067] Figure 1 This is a flowchart of Embodiment 1 of this application;

[0068] Figure 2 This is a flowchart of Embodiment 2 of this application. Detailed Implementation

[0069] The following combination Figures 1 to 2 This application will be described in further detail.

[0070] Example 1: This example discloses a method for data security protection of underwater IoT facilities. The underwater IoT includes multiple underwater IoT facilities, as described above. Figure 1 The method includes: S11 data acquisition and S12 risk assessment. First, corresponding edge nodes are deployed on the underwater IoT facility. Environmental data from each edge node is collected, and the underwater environmental risk level of all underwater IoT facilities is calculated based on this data. Then, it is determined whether the k-th underwater environmental risk level is greater than a preset warning threshold. If it is, the operational data stored in the k-th underwater IoT facility is acquired, encrypted, and then transmitted to the server. If it is not greater, the (k+1)-th underwater environmental risk level is updated to the k-th underwater environmental risk level, and the risk assessment steps continue until all underwater environmental risk levels have been traversed. The execution process of this embodiment is as follows:

[0071] S11 Data Acquisition: Since the underwater IoT consists of multiple underwater IoT facilities, this embodiment deploys at least one edge node on each underwater IoT facility. These edge nodes are capable of sensing and collecting data about the environment surrounding the underwater IoT facilities in real time.

[0072] Specifically, edge nodes deployed on the i-th underwater IoT facility collect environmental data about its location. This environmental data includes pressure, temperature, and depth data. Pressure data reflects the water pressure experienced by the underwater IoT facility; changes in water pressure indicate instability in the marine environment or potential structural problems. Temperature data reflects the temperature of the underwater environment; abnormal temperature fluctuations can affect the normal operation of the electronic components inside the underwater IoT facility. Depth data specifies the depth of the underwater IoT facility.

[0073] Based on the environmental data collected from the i-th underwater IoT facility, the underwater environmental risk level of the i-th underwater IoT facility is calculated using an underwater environmental risk level calculation model, as follows:

[0074] ;

[0075] Where R is the numerical value of the underwater environmental risk level; Weights for stress data; Pressure data for underwater IoT facilities, in Pa; The pressure withstand threshold for underwater IoT facilities, in Pa; Weights for temperature data; Temperature data for underwater IoT facilities, in °C; The optimal operating temperature for underwater IoT facilities, in °C; The maximum temperature that underwater IoT facilities can withstand, in °C; The minimum temperature that underwater IoT facilities can withstand, in °C; Weights for depth data; Depth data for underwater IoT facilities, in meters (m). The maximum depth that an underwater IoT facility can withstand, measured in meters (m).

[0076] In other embodiments, the degree of risk of the underwater environment can also be calculated using the following model:

[0077] ;

[0078] ;

[0079] ;

[0080] ;

[0081] Where R is the numerical value of the underwater environmental risk level; This represents the stress risk value. Temperature risk value; Value for deep risk; Pressure data for underwater IoT facilities, in Pa; The pressure withstand threshold for underwater IoT facilities, in Pa; Temperature data for underwater IoT facilities, in °C; Depth data for underwater IoT facilities, in meters (m). The maximum depth that an underwater IoT facility can withstand, measured in meters (m).

[0082] In other embodiments, a calculation model for the degree of underwater environmental risk can be set according to requirements, which will not be elaborated in this embodiment.

[0083] Subsequently, this embodiment sequentially collects environmental data and calculates the risk level of all underwater IoT facilities, ultimately obtaining the underwater environmental risk level of all underwater IoT facilities.

[0084] S12 Risk Assessment determines whether the risk level of the k-th underwater environment exceeds the preset warning threshold.

[0085] If so, meaning the risk level of the k-th underwater environment exceeds the preset warning threshold, it indicates that the underwater IoT facility is in a high-risk environment and may be damaged due to harsh conditions. Subsequently, the operational data and target data stored within the k-th underwater IoT facility are acquired. The operational data contains crucial information such as the facility's operating status and parameters, which is essential for subsequent analysis, troubleshooting, and determining whether the underwater IoT facility is operating normally. The target data includes data related to the data acquisition or exploration tasks performed by the underwater IoT facility.

[0086] To prevent the loss of operational and target data due to damage to underwater IoT facilities and to improve data security and integrity, this embodiment encrypts the acquired operational and target data. The encryption process employs encryption algorithms (such as symmetric or asymmetric encryption algorithms) to transform the original data into a difficult-to-decipher ciphertext form. Only the recipient (i.e., the server) with the corresponding decryption key can recover the original data. After encryption, the encrypted operational and target data is transmitted to the server. The server, as the central data storage and management unit, can further process and analyze the encrypted data, for example, storing operational and target data at risk of loss.

[0087] If the judgment result is negative, meaning the risk level of the k-th underwater environment is not greater than the preset warning threshold, it indicates that the current environmental risk of the underwater IoT facility is low, and there is no need to encrypt and transmit the internally stored operational data for the time being. In this case, this embodiment will update the risk level of the (k+1)-th underwater environment to the risk level of the k-th underwater environment, meaning that the focus of the assessment will shift to the next underwater IoT facility, and the risk assessment in S12 will continue. Through this iterative process, the underwater environmental risk level of each underwater IoT facility will be judged sequentially until all underwater environmental risk levels have been traversed.

[0088] By adopting the above scheme, this embodiment transmits the operational data and target data stored in the underwater IoT facility in a relatively harsh environment to the server. For underwater IoT facilities in a safe environment, the collected operational data and target data are still stored inside the facility. This application improves the security of the operational data and target data stored in the underwater IoT facility by reducing the number of transmissions of the operational data and target data stored in the underwater IoT facility to the server.

[0089] In principle, this embodiment utilizes edge nodes to collect environmental data of the underwater IoT facility's location when it starts operating, and continuously calculates and judges the level of underwater environmental risk within the underwater IoT facility. When the level of underwater environmental risk exceeds a preset warning threshold, the operating data and target data stored within the underwater IoT facility can be encrypted and transmitted to the server in a timely manner, reducing the risk of data loss.

[0090] Example 2: Refer to Figure 2 The difference between this embodiment and Embodiment 1 is that the method further includes:

[0091] S21 collects real-time operational data and records the operational data stored in each underwater IoT facility as historical operational data. Historical operational data includes records of various operational parameters of the underwater IoT facility under different environmental conditions and operating states, such as historical operational status parameters and historical control commands, reflecting the past operational characteristics and performance of the underwater IoT facility.

[0092] Real-time operational data is collected, providing an immediate reflection of the underwater IoT facility's current operational status and offering high timeliness. Various sensors and data acquisition devices deployed on the underwater IoT facility continuously acquire environmental awareness information and operational parameters. For example, within the underwater IoT facility, parameters such as motor speed and power consumption are obtained.

[0093] S22 modeling and training, based on a federated learning framework, constructs local models for each underwater IoT facility and a global model for the server. The local models are designed specifically for the operational data and characteristics of each underwater IoT facility, enabling them to better capture its operational patterns and features. Lightweight neural network models, such as 1D-CNN or lightweight fully connected networks, can be used, as long as the local model, after training, can determine whether the real-time operational data is in a normal or abnormal state. The global model, on the other hand, is a synthesis and optimization of the various local models, aiming to improve the overall performance and generalization ability of the model. For example, the FedAvg algorithm is used to aggregate the model parameters of each underwater IoT facility to form a global model.

[0094] Federated learning is a distributed machine learning framework that allows multiple stakeholders (i.e., various underwater IoT facilities) to collaboratively train a global model without sharing the original data.

[0095] During model training, labeled historical operational data is used to train the local model. Labeling allows the local model to explicitly learn the operational states or categories corresponding to different data, such as normal operation and abnormal operation. The local model is repeatedly trained using a large amount of labeled historical operational data, continuously adjusting its parameters to gradually enable it to accurately identify and predict the operational status of underwater IoT facilities. After multiple rounds of training, the trained local model is obtained.

[0096] S23 Data Filtering employs clustering algorithms (such as K-Means clustering, spectral clustering, etc.) to cluster the historical operational data of each underwater IoT facility. Clustering algorithms group data according to similarity, grouping data with similar characteristics into one category. Through clustering, this embodiment divides the historical operational data into normal clusters and abnormal clusters. The normal cluster contains data of the underwater IoT facility under normal operating conditions, reflecting the facility's normal operating mode and characteristics; the abnormal cluster contains data of the underwater IoT facility under faulty, abnormal, or other abnormal operating conditions.

[0097] The centroid of the normal cluster is designated as the first data point, and the centroid of the abnormal cluster is designated as the second data point. The centroid is the center point of all data points in a cluster and represents the average characteristic of the data in that cluster.

[0098] Calculate the Euclidean distance between the real-time running data and the first data point, and denote it as the third data point; calculate the Euclidean distance between the real-time running data and the second data point, and denote it as the fourth data point. Euclidean distance is a commonly used distance metric that reflects the degree of similarity between two data points; the smaller the distance, the higher the degree of similarity.

[0099] Next, it is determined whether the third data point is less than the fourth data point. If the third data point is not less than the fourth data point, it indicates that the real-time running data is closer to the centroid of the abnormal cluster, meaning the real-time running data is more likely to be in an abnormal state. In this case, historical running data contained within the normal cluster is deleted. If the third data point is less than the fourth data point, it indicates that the real-time running data is closer to the centroid of the normal cluster, meaning the real-time running data is more likely to be in a normal running state. In this case, historical running data contained within the abnormal cluster is deleted. Otherwise, historical running data contained within the normal cluster is deleted.

[0100] S24 Similarity Assessment: Calculate the similarity between the real-time operational data of the i-th underwater IoT facility and each remaining historical operational data point of that facility (i.e., the remaining historical operational data after data filtering in S23). Similarity can be calculated using various methods, such as cosine similarity or Pearson correlation coefficient. By employing the above methods, this step can measure the degree of similarity between real-time operational data and historical operational data.

[0101] Determine whether the similarity is greater than a preset similarity threshold.

[0102] If the similarity score is greater than a preset similarity threshold, it indicates that the real-time operational data has a high degree of similarity to the remaining historical operational data, belonging to the same operational state or category as these historical data. The real-time operational data is then input into the trained local model, which classifies the data to obtain the output of the trained local model. The output indicates whether the underwater IoT facility is in a normal or abnormal operational state.

[0103] If the similarity is not greater than the preset similarity threshold, it means that the similarity between the real-time running data and the historical running data is low. In this case, execute S25 to update the parameters to further optimize the local model.

[0104] S25 updates parameters by obtaining the parameters of the trained local model, denoted as local parameters. These local parameters contain the knowledge and features learned by the local model during training, reflecting the operational patterns and characteristics of the underwater IoT facility. The local parameters are then encrypted. Encryption can employ symmetric or asymmetric encryption algorithms to convert the local parameters into ciphertext, reducing the risk of them being stolen or tampered with during transmission.

[0105] Each underwater IoT facility uses its own stored operational data to train a local model and obtain local parameters. The local model is trained using gradient descent, with the goal of minimizing the local loss function. Underwater IoT facilities rely on local parameters. Encryption is performed (e.g., using homomorphic encryption or differential privacy techniques) to obtain the encrypted local parameters. This reduces the risk of data privacy leaks during the upload process.

[0106] Next, the encrypted local parameters Uploaded to the server. The server, as the central hub for local parameter aggregation and management, aggregates the encrypted local parameters uploaded by various underwater IoT facilities. Aggregation processing can employ methods such as weighted averaging and federated averaging to combine and optimize the parameters of multiple local models, generating global parameters for the global model.

[0107] ;

[0108] ;

[0109] in, is a global parameter; N is the total number of underwater IoT facilities; Let m be the amount of local parameter data for the i-th underwater IoT facility; m is the total amount of local parameter data for all underwater IoT facilities.

[0110] Global parameters represent the overall knowledge and features learned from data from various underwater IoT facilities, which can improve the performance and generalization ability of the global model.

[0111] The server then encrypts the global parameters to obtain the encrypted global parameters. To enhance the security of global parameters during transmission, encrypted global parameters are sent to each underwater IoT facility. Upon receiving the encrypted global parameters, each underwater IoT facility decrypts them to recover the original global parameters.

[0112] Finally, each underwater IoT facility updates the local parameters in its trained local model based on the decrypted global parameters, obtaining the updated local model. The update methods include:

[0113] Direct replacement: Replace local parameters Directly replace with global parameters .

[0114] Weighted fusion: combining local parameters With global parameters Perform weighted fusion.

[0115] ;

[0116] These are the parameters of the updated local model corresponding to the i-th underwater IoT facility; For the local parameters corresponding to the i-th underwater IoT facility The weight.

[0117] Gradient descent: This method uses global parameters... As initial values, continue training the local model using historical running data:

[0118] ;

[0119] in, These are the parameters of the updated local model corresponding to the i-th underwater IoT facility; Let be the learning rate of the local model corresponding to the i-th underwater IoT facility; Let be the gradient of the local model loss function corresponding to the i-th underwater IoT facility.

[0120] Through the aforementioned parameter update mechanism, each local model can continuously absorb data and knowledge from other underwater IoT facilities, thereby achieving continuous optimization and performance improvement of the local model to better adapt to changes in the facility's operating environment and emerging data characteristics.

[0121] S26 classification involves inputting real-time running data into the updated local model to obtain classification results.

[0122] By adopting the above scheme, this embodiment constructs a local model capable of classifying real-time operational data in each underwater IoT facility through a federated learning framework, reducing the data transmission between the underwater IoT facility and the server, and reducing the risk of real-time operational data being tampered with during transmission.

[0123] Example 3: This example differs from Example 1 in that the method further includes:

[0124] The correlation between temperature and depth data was calculated using the Pearson correlation coefficient algorithm. The Pearson correlation coefficient measures the degree of linear correlation between two variables, with values ​​ranging from -1 to 1. The closer the absolute value is to 1, the stronger the correlation.

[0125] Based on the calculated correlation, a cross term is introduced into the calculation model for underwater environmental risk level. The introduction of the cross term more accurately reflects the impact of the interaction between temperature and depth on risk, thus obtaining an extended calculation model for underwater environmental risk level. The underwater environmental risk level of the i-th underwater IoT facility is recalculated using the extended model, denoted as the new underwater environmental risk level, and this is used as the basis for performing the S12 risk assessment.

[0126] Taking the calculation model in Example 1 as an example, the extended calculation model for the underwater environment risk level in this example is as follows:

[0127] ;

[0128] in, This represents the numerical value for the expanded underwater environmental risk level. The correlation between temperature and depth data.

[0129] Since there is a relationship between pressure data and depth data in water that can be derived through a calculation model, this embodiment does not calculate the correlation between pressure data and temperature data. In other embodiments, depth data can be converted into pressure data to extend the calculation model for the degree of underwater environmental risk. Alternatively, the cross terms of temperature data and pressure data, and temperature data and depth data can be added to the calculation model for the degree of underwater environmental risk.

[0130] In other embodiments, the calculation method for the extended underwater environmental risk level is similar to that in this embodiment, and will not be repeated here.

[0131] A sensitivity matrix is ​​constructed based on the extended underwater environmental risk level calculation model and environmental data (pressure, temperature, depth). The sensitivity matrix is ​​a two-dimensional matrix, where the element in the i-th row and j-th column represents the partial derivative of the extended underwater environmental risk level calculation model of the i-th underwater IoT facility with respect to the j-th environmental data.

[0132] Partial derivatives reflect the rate of change of underwater environmental risk when a certain environmental data undergoes a small change. For example, for the i-th underwater IoT facility, its partial derivative with respect to pressure data represents the change in the facility's underwater environmental risk level when the pressure data changes by one unit. By calculating the value of each element in the sensitivity matrix, a comprehensive understanding of the impact of various environmental data on the risk levels of different facilities can be obtained.

[0133] The sensitivity value of the i-th underwater IoT facility is calculated based on a sensitivity matrix. The sensitivity value comprehensively considers the facility's sensitivity to all environmental data and is a quantitative indicator of the facility's overall sensitivity; the higher the sensitivity value, the more sensitive the underwater IoT facility is to environmental data. It helps assess the sensitivity of underwater IoT facilities to risk changes under different environmental conditions.

[0134] The calculation model for the sensitivity value is as follows:

[0135] ;

[0136] In other embodiments, the sensitivity value can also be calculated using the L1 norm, with the following calculation model:

[0137] ;

[0138] The false alarm rate (FAR) is the probability of incorrectly classifying a situation as risky during risk assessment. Based on the sensitivity value and the FAR, a new warning threshold is calculated using the Youden index method. The calculation model for the new warning threshold is as follows:

[0139] ;

[0140] in, A represents the new warning threshold; B represents the sensitivity value; and C represents the false alarm rate.

[0141] The new warning thresholds comprehensively consider the facility's sensitivity to environmental changes and the likelihood of misjudgments during the assessment process. A high sensitivity value indicates a strong response to environmental changes, requiring a lower warning threshold to detect potential risks more promptly. Conversely, a high false alarm rate indicates numerous errors in the assessment process, necessitating an appropriate increase in the warning threshold to reduce unnecessary warnings. Re-performing the S12 risk assessment based on the new warning thresholds improves the accuracy and reliability of the risk assessment, more effectively ensuring the safe operation of underwater IoT facilities.

[0142] Example 4: This example discloses a data security protection system for underwater IoT facilities. The underwater IoT includes multiple underwater IoT facilities, and the system includes:

[0143] The data acquisition module is used to deploy at least one edge node (such as a sensor node or embedded device) in each underwater IoT facility. It is responsible for the acquisition and preliminary processing of local environmental data, calculating the underwater environmental risk level of the i-th underwater IoT facility based on the environmental data, obtaining the underwater environmental risk level of all underwater IoT facilities, and transmitting the underwater environmental risk level of all underwater IoT facilities to the risk assessment module. The environmental data includes pressure data, temperature data, and depth data.

[0144] The risk assessment module is used to sequentially determine whether the risk level of each underwater environment exceeds a preset warning threshold.

[0145] If the risk level of an underwater environment exceeds a preset warning threshold, the operation data and target data stored in the underwater IoT facility corresponding to the risk level of the underwater environment exceeding the preset warning threshold are obtained, the operation data and target data are encrypted, and the encrypted operation data and target data are transmitted to the server.

[0146] If the current underwater environmental risk level is not greater than the preset warning threshold, the next underwater environmental risk level will be used as the current underwater environmental risk level, and the risk assessment module will be triggered until all underwater environmental risk levels have been traversed.

[0147] Example 5: This example provides a computer-readable storage medium storing a computer program thereon, characterized in that the computer program, when processed by a processor, is used to implement the methods described in Examples 1-3.

[0148] Example 6: This example provides a server, characterized in that the server is loaded with the system described in Example 4, the server comprising: a processor, and a memory communicatively connected to the processor;

[0149] The memory is provided with the computer-readable storage medium described in Embodiment 5, and the computer-readable storage medium stores a computer program.

[0150] When the processor processes the computer program stored on the computer-readable storage medium, it implements the methods described in Examples 1-3.

[0151] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for data security protection of underwater IoT facilities, wherein the underwater IoT includes multiple underwater IoT facilities, characterized in that, The method includes: Data Acquisition: Deploy corresponding edge nodes on underwater IoT facilities, use edge nodes to collect environmental data of the location of the i-th underwater IoT facility, calculate the underwater environmental risk level of the i-th underwater IoT facility based on the environmental data, and obtain the underwater environmental risk level of all underwater IoT facilities. The environmental data includes pressure data, temperature data and depth data. Risk assessment: Determine whether the risk level of the k-th underwater environment exceeds the preset warning threshold. If so, then obtain the operation data and target data stored in the kth underwater IoT facility, encrypt the operation data and target data, and transmit the encrypted operation data and target data to the server; If not, update the (k+1)th underwater environmental risk level to the kth underwater environmental risk level, and perform the risk assessment steps until all underwater environmental risk levels have been traversed. The method further includes: The correlation between temperature and depth data is calculated using the Pearson correlation coefficient algorithm. Based on this correlation, an intersection term is introduced into the calculation model of underwater environmental risk level to obtain an extended calculation model of underwater environmental risk level. A sensitivity matrix is ​​constructed based on the extended underwater environmental risk level calculation model and environmental data. The sensitivity value of the i-th underwater IoT facility is calculated based on the sensitivity matrix. The element in the i-th row and j-th column of the sensitivity matrix represents the partial derivative of the extended underwater environmental risk level calculation model of the i-th underwater IoT facility with respect to the j-th environmental data. The false alarm rate is statistically analyzed. Based on the sensitivity value and the false alarm rate, a new warning threshold is calculated using the Youden index method. The risk assessment steps are then re-executed based on the new warning threshold.

2. The method for data security protection of underwater IoT facilities according to claim 1, characterized in that, The method further includes: Collect real-time operational data: Record the operational data stored in each underwater IoT facility as the historical operational data of each underwater IoT facility, and collect the real-time operational data of each underwater IoT facility. Modeling and Training: Based on the federated learning framework, local models of various underwater IoT facilities and global models of servers are constructed. The local models are trained using labeled historical operating data to obtain the trained local models. Update parameters: Obtain the parameters of the trained local model, denoted as local parameters, encrypt the local parameters, and upload the encrypted local parameters to the server. The server aggregates the encrypted local parameters to generate global parameters for the global model. The server encrypts the global parameters and distributes them to each underwater IoT facility. Each underwater IoT facility decrypts the encrypted global parameters and updates the local parameters in the trained local model according to the decrypted global parameters to obtain the updated local model. Classification: Input the real-time running data into the updated local model to obtain the classification results.

3. The method for data security protection of underwater IoT facilities according to claim 2, characterized in that, The method further includes: Similarity assessment: Calculate the similarity between the real-time operational data of the i-th underwater IoT facility and each historical operational data point of the i-th underwater IoT facility, and determine whether the similarity exceeds a preset similarity threshold. If so, the real-time running data will be input into the trained local model to obtain the output result of the trained local model; If not, then proceed with the parameter update step.

4. The method for data security protection of underwater Internet of Things facilities according to claim 3, characterized in that, Before performing the similarity determination step, the following steps are also included: Clustering algorithms are used to cluster the historical operational data of each underwater IoT facility to obtain normal clusters and abnormal clusters. The centroid of the normal cluster is recorded as the first data and the centroid of the abnormal cluster is recorded as the second data. The Euclidean distance between the real-time operational data and the first data is calculated and recorded as the third data. The Euclidean distance between the real-time operational data and the second data is calculated and recorded as the fourth data. Determine if the third data is less than the fourth data. If so, delete the historical runtime data contained within the abnormal cluster; If not, delete the historical runtime data contained in the normal cluster.

5. The method for data security protection of underwater Internet of Things (IoT) facilities according to any one of claims 1-4, characterized in that, The calculation model for the underwater environment risk level is as follows: ； Where R is the numerical value of the underwater environmental risk level; Weights for stress data; Pressure data for underwater IoT facilities; The pressure withstand threshold for underwater IoT facilities; Weights for temperature data; Temperature data for underwater IoT facilities; The optimal operating temperature for underwater IoT facilities; The maximum temperature that underwater IoT facilities can withstand; The minimum temperature that underwater IoT facilities can withstand; Weights for depth data; Depth data for underwater IoT facilities; This refers to the maximum depth that underwater IoT facilities can withstand.

6. The method for data security protection of underwater Internet of Things facilities according to claim 5, characterized in that, The method further includes: The underwater environmental risk level of the i-th underwater IoT facility is recalculated using the extended underwater environmental risk level calculation model, and recorded as the new underwater environmental risk level. The risk assessment steps are then performed.

7. A data security protection system for underwater IoT facilities, wherein the underwater IoT comprises multiple underwater IoT facilities, characterized in that, The system is applicable to the method as described in any one of claims 1-6, the system comprising: The data acquisition module is used to deploy corresponding edge nodes on underwater IoT facilities, collect environmental data of the location of the i-th underwater IoT facility using the i-th edge node, calculate the underwater environmental risk level of the i-th underwater IoT facility based on the environmental data, and obtain the underwater environmental risk level of all underwater IoT facilities. The environmental data includes pressure data, temperature data and depth data. The risk assessment module is used to determine whether the risk level of the k-th underwater environment exceeds a preset warning threshold. If the risk level of the kth underwater environment is greater than the preset warning threshold, the operation data and target data stored in the kth underwater IoT facility are obtained, the operation data and target data are encrypted, and the encrypted operation data and target data are transmitted to the server. If the risk level of the kth underwater environment is not greater than the preset warning threshold, then the risk level of the (k+1)th underwater environment is updated to the risk level of the kth underwater environment, and the risk assessment module is triggered until all underwater environmental risk levels have been traversed.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is processed by a processor, it is used to implement the method as described in any one of claims 1-6.

9. A server, characterized in that, The server is loaded with the system as described in claim 7, the server comprising: a processor, and a memory communicatively connected to the processor; The memory is provided with a computer-readable storage medium as described in claim 8, and a computer program is stored on the computer-readable storage medium. When the processor processes a computer program stored on the computer-readable storage medium, it implements the method as described in any one of claims 1-6.

Images