An unmanned aerial vehicle data collection optimization method based on matrix completion and trust evaluation

Abstract

This invention proposes an optimized method for drone data acquisition based on matrix completion and trust assessment, applicable to drone forensics in IoT network defense. First, in matrix completion, the sampling point locations and data packet acquisition times are constructed into a matrix, and matrix completion technology is used to recover all information and select sampling points. Second, in the drone flight trajectory, the selected sampling points and the Elite Ant Trail Optimization (EATO) algorithm are combined to optimize the drone's flight trajectory. Third, in the trust evolution mechanism, the comprehensive trust level of sensor nodes is obtained through dual evaluation by neighboring nodes and the drone. The proposed method can solve the security risks and resource consumption problems of data acquisition from IoT smart devices, effectively identify malicious nodes, optimize the accuracy of trust assessment, improve network security, and reduce drone energy consumption.

Description

Technical Field

[0001] This invention belongs to the fields of IoT network security, intelligent drone digital forensics, wireless sensor networks and trust management. It relates to an optimization method for drone data acquisition based on matrix completion and trust assessment, which is applicable to the assessment of node trust in IoT sensor networks, the identification of malicious nodes, security assurance during intelligent drone data acquisition, and the optimization of data acquisition paths. Background Technology

[0002] The Internet of Things (IoT), by connecting a large number of sensors, actuators, controllers, and smart terminals to a network, enables extensive sensing, status monitoring, and data interaction of the physical world, and has been widely applied in smart cities, intelligent transportation, environmental monitoring, industrial manufacturing, healthcare, and agricultural production. With the continuous expansion of IoT device scale and the increasingly open deployment environment, IoT systems face increasingly prominent security challenges in data collection, transmission, storage, and processing. Due to the characteristics of numerous devices—resource-constrained, long-term unattended operation, dispersed distribution, and strong heterogeneity—attackers can easily disrupt network operations through node capture, identity forgery, injection of false data, data tampering, and denial-of-service attacks, leading to the leakage of sensitive information, distortion of monitoring results, and errors in system decision-making.

[0003] Traditional cybersecurity forensics methods primarily rely on fixed infrastructure, centralized log analysis, and manual intervention, and are typically applicable to relatively stable traditional network environments. However, given the massive number, wide distribution, and rapid topological changes of IoT nodes, traditional forensic methods suffer from significant shortcomings in real-time performance, mobility, and coverage. Drones, with their advantages of flexible deployment, high mobility, rapid access to target areas, and ease of performing mobile sensing tasks, can serve as mobile forensic vehicles to complete data collection, log gathering, status awareness, and evidence transmission tasks in complex environments. Therefore, utilizing intelligent drones for data collection and digital forensics in IoT scenarios is gradually becoming an important research direction in cybersecurity forensics.

[0004] In typical IoT-enabled smart drone forensics scenarios, sensor nodes are usually organized into multiple clusters based on region or function. Each cluster has a cluster head node responsible for collecting information from member nodes and interacting with external data acquisition terminals. The smart drone flies to each cluster head node or data acquisition station according to a predetermined task, collecting network operation status, business data, log information, and suspicious evidence, and then transmits the collected results back to the decision center for comprehensive analysis. While this model improves the flexibility and coverage of data collection, it also introduces new security risks. If the cluster head node or member nodes have been controlled by an attacker, the drone may collect tampered, forged, or incomplete data, affecting the authenticity and validity of subsequent forensic results. Furthermore, drones are typically limited by flight energy, range, and payload capacity; if the selection of collection points is unreasonable or the flight path planning is poor, it can easily lead to energy waste and decreased collection efficiency. Therefore, how to identify trustworthy nodes, filter malicious nodes, and optimize data collection strategies during drone forensics has become an urgent problem to be solved.

[0005] To address the challenges of IoT data analysis and behavior recognition, existing research has proposed various methods based on machine learning and deep learning. For example, some studies utilize unsupervised learning methods to extract behavioral patterns and risk features from vehicle-to-everything (V2X) or sensor network data, while others use recurrent neural networks to predict network security posture or jointly predict future security events and their timing. These methods can extract time-series information, behavioral patterns, and potential risks from large amounts of data, playing a positive role in security monitoring and situational awareness. However, most of these methods focus on data pattern extraction, trend prediction, and event recognition, making it difficult to assess the trustworthiness of the data source nodes themselves. When malicious nodes inject false data into the network, while the relevant models can analyze the data, they cannot guarantee the authenticity and reliability of the data from the source.

[0006] In the field of intrusion detection, researchers have proposed numerous detection models for the Internet of Things (IoT) and the Industrial IoT. For example, Attique et al. proposed an interpretable and data-efficient intrusion detection method for the Industrial IoT. This method employs a structure combining a bidirectional long short-term memory network (LSTM) with an adaptive attention mechanism to improve the model's ability to extract key attack features and temporal dependencies under limited training data conditions. Furthermore, a Shapley additive interpretation mechanism is introduced to enhance the interpretability and credibility of the detection results. This method achieved high detection accuracy on the CICIDS2017 and X-IIoTID datasets. Other studies have proposed intrusion detection methods that integrate convolutional neural networks, autoencoders, federated learning, and edge computing to improve the identification of network attacks, botnet attacks, and abnormal traffic. While these methods have shown some effectiveness in network traffic analysis and attack detection, their research focus is primarily on network intrusion detection and attack classification, typically targeting centralized or semi-centralized traffic analysis scenarios. They do not address trust modeling, malicious node filtering, and trust-based data collection decision-making issues for IoT nodes in intelligent drone forensics scenarios.

[0007] In the field of trust management, researchers have proposed various models and methods for assessing the trustworthiness of IoT nodes. For example, some studies have systematically defined the concepts of trust, trust transmission, trust updates, and contextual factors in IoT from a theoretical perspective, providing a basic framework for trust management. Other studies have categorized and summarized trust management schemes in social IoT, pointing out that existing methods are insufficient in considering aspects such as node reliability, data consistency, and energy consumption. Some studies have constructed dynamic decentralized trust management models under edge computing architectures, combining direct monitoring and indirect evaluation to achieve dynamic calculation of inter-device interaction trust. Still other studies have proposed dynamic evaluation models that comprehensively consider communication trust, data trust, and energy trust for wireless sensor networks, and trust-enhanced forwarding schemes based on contact probability, service level, and social attributes for mobile IoT. While these methods have all played a positive role in node trust assessment in different scenarios, most focus on static networks, edge collaboration, or data forwarding environments, lacking adaptation designs for intelligent drone mobile data collection scenarios, and failing to effectively solve the problem of independent verification of data source trustworthiness during drone forensics.

[0008] Existing technologies have the following shortcomings when applied to IoT-enabled smart drone forensics scenarios: First, most existing node trust assessment methods rely on nodes' own historical interaction records or neighbor recommendations, lacking independent assessment capabilities from the mobile data collection terminal, making them susceptible to malicious node collusion, false recommendations, and forged data. Second, while existing intrusion detection methods can identify attack behaviors in network traffic, they are difficult to directly apply to node trust screening during drone forensics, failing to guide drones to prioritize data collection from highly trustworthy nodes. Third, existing trust management methods and drone data collection strategies are usually independent of each other, failing to coordinate node trust with collection point selection, data redundancy control, and flight path planning, resulting in insufficient collection efficiency and security. Fourth, in multi-cluster, multi-node, and dynamically changing IoT environments, if the collection plan cannot be adjusted in real time based on node trust status, drones may collect a large amount of low-value or even false data, wasting flight energy and communication resources, and affecting the decision-making center's analysis results of the forensic information.

[0009] Therefore, there is an urgent need to propose a trust assessment method for IoT-based intelligent drone forensics, in order to dynamically and accurately assess the credibility of network nodes, identify and isolate malicious nodes, ensure the authenticity and integrity of drone-collected data, and improve data collection efficiency and mission execution security on this basis. Summary of the Invention

[0010] The purpose of this invention is to propose an optimized method for UAV data acquisition based on matrix completion and trust assessment, addressing issues in existing technologies such as uneven energy consumption of cluster head nodes, difficulty in identifying malicious nodes, unoptimized UAV flight trajectories, and data acquisition redundancy. By introducing matrix completion technology, improving the ant colony algorithm, and implementing a dual trust assessment mechanism, this invention can effectively detect malicious nodes, optimize the selection of sampling points, and dynamically optimize UAV flight trajectories while ensuring data acquisition quality, thereby improving the energy efficiency, security, and data acquisition efficiency of wireless sensor networks.

[0011] To achieve the above objectives, this invention provides an optimization method for UAV data acquisition based on matrix completion and trust assessment, the method comprising:

[0012] Using the trust level of sensor nodes as input, nodes are filtered based on the trust threshold to obtain a set of trustworthy nodes;

[0013] Nodes in the set of trusted nodes are used as candidate sampling points, and sampling points are selected based on matrix completion techniques to obtain a set of data collection stations.

[0014] Using a set of data collection stations as input, the improved ant colony algorithm is used for path optimization to obtain the optimized movement trajectory of the drone.

[0015] The node trust level is updated based on the number of successful and failed direct interactions between the node and its neighboring nodes within a preset time period, and the node trust level is obtained by evaluating the neighboring nodes.

[0016] The data packets collected by the drone are used as input, and the node trust level is updated according to the trust evaluation mechanism to obtain the updated node trust level.

[0017] The overall trust value of a node is obtained by fusing the trust values ​​of neighboring nodes with those of the drone.

[0018] In some embodiments, the trust level of the sensor nodes is used as input, and nodes are filtered according to a trust threshold to obtain a set of trustworthy nodes:

[0019] The evaluation trust value of each sensor node is compared with the preset trust threshold;

[0020] Nodes whose assessed trust value is greater than or equal to a preset trust threshold are identified as trusted nodes and a set of trusted nodes is formed.

[0021] Nodes whose assessed trust value is less than the preset trust threshold are identified as malicious nodes and excluded from the selection of sampling points.

[0022] In some embodiments, the nodes in the trusted node set are selected as candidate sampling points, and the sampling points are selected based on matrix completion technology to obtain the data collection station set as follows:

[0023] The sensor nodes within the monitoring area are constructed into an information matrix, and the number of sampling points to be selected in each column is determined based on matrix completion technology;

[0024] For the first column, sampling points are selected based on the uniform distribution probability;

[0025] For subsequent columns, the sampling probability is calculated based on the number of historically selected sampling points in each row, and sampling points are selected sequentially from high to low according to the probability until the required number of sampling points for each column is reached, thus obtaining the data collection station set.

[0026] In some embodiments, the data collection station set is used as input, and path optimization is performed based on the improved ant colony algorithm to obtain the optimized movement trajectory of the UAV:

[0027] The nodes in the data collection station set are treated as nodes in the graph, and the edges between the nodes are assigned initial pheromones.

[0028] For each iteration, the drone starts from the starting point, calculates the path selection probability based on the pheromones on the edges, and selects the next unvisited node based on the probability.

[0029] After the drone completes its access to all data collection stations, it records the path length of this iteration and updates the pheromones on each edge according to the path length. The shorter the path length, the greater the pheromone increment.

[0030] Repeat the iteration until the preset number of iterations is reached, and use the optimal path as the optimized movement trajectory of the drone.

[0031] In some embodiments, the number of successful and failed direct interactions between a node and its neighboring nodes within a preset time period is used as input, and the node's trust level is updated according to a trust evaluation mechanism to obtain the node's trust level evaluated by its neighboring nodes.

[0032] Record the number of successful interactions and the number of failed interactions between a node and its neighboring nodes within a preset time period;

[0033] Based on the number of successful interactions and the number of failed interactions in the past, combined with the number of interactions within a preset time period, update the cumulative number of successful interactions and the cumulative number of failed interactions;

[0034] Based on the updated cumulative number of successful interactions and cumulative number of failed interactions, calculate the node's trust value for its neighboring nodes;

[0035] The neighbor evaluation trust values ​​of a node are merged to obtain the neighbor evaluation trust degree of that node.

[0036] In some embodiments, using data packets collected by the drone as input to update node trust levels according to a trust assessment mechanism includes having the drone assess node trust levels:

[0037] The drone collects data packets from the data collection station; each data packet contains the source node identifier and data content.

[0038] The data packet is used as input to compare the node data obtained by the data collection station with the expected data;

[0039] If the error between the node data and the expected data exceeds the preset error range, the trust level of that node will be reduced.

[0040] If the error between the node data and the expected data is within the preset error range, the trust level of that node will be maintained or increased.

[0041] In some embodiments, the trust values ​​of neighboring nodes and the trust values ​​of the drone are fused to obtain the overall trust value of the node:

[0042] The overall trust value needs to take into account both neighbor node trust based on the number of interactions and drone assessment trust based on the information error rate.

[0043] As can be seen from the above technical solution, the purpose of this invention is to provide a UAV data acquisition optimization method based on matrix completion and trust assessment, including: taking the trust level of sensor nodes as input, filtering nodes according to a trust threshold to obtain a set of trustworthy nodes; taking nodes in the set of trustworthy nodes as candidate sampling points, selecting sampling points according to matrix completion technology to obtain a set of data collection stations; taking the set of data collection stations as input, optimizing the path according to an improved ant colony algorithm to obtain the optimized movement trajectory of the UAV; taking the number of successful and failed direct interactions between a node and its neighboring nodes within a preset time period as input, updating the node trust level according to a trust assessment mechanism to obtain the node trust level evaluated by neighboring nodes; taking the data packets collected by the UAV as input, updating the node trust level according to the trust assessment mechanism to obtain the updated node trust level.

[0044] This invention optimizes the selection of data collection points based on matrix completion technology, dynamically optimizes the flight trajectory of UAVs based on an improved ant colony algorithm, and comprehensively evaluates the trust level of nodes by combining a dual trust mechanism of neighbor node evaluation and UAV evaluation, so as to achieve efficient, safe and low-energy UAV data collection.

[0045] Compared with existing technologies, the proposed UAV data acquisition optimization method based on matrix completion and trust assessment has the following advantages and effects:

[0046] Existing data acquisition point selection methods suffer from severe data redundancy, requiring cluster head nodes to collect data packets from all sensor nodes, resulting in large data acquisition volumes and high energy consumption. This invention, however, introduces matrix completion technology, enabling the recovery of information for the entire monitoring area from data packets collected from only a small number of sampling points. This significantly reduces the total data acquisition volume, lowers node energy consumption, and extends network lifetime.

[0047] In existing technologies, cluster head nodes suffer from uneven energy consumption. A single cluster head node needs to handle a large number of data packets for receiving and forwarding, leading to excessive energy consumption and premature failure. This invention addresses this by dynamically selecting data collection stations and rotating them periodically, ensuring that each node carries roughly the same number of data packets, thus achieving balanced energy consumption distribution and improving the overall network lifetime.

[0048] Existing technologies suffer from insufficient malicious node detection capabilities. Malicious nodes with legitimate identities may discard data packets or send spoofed data packets, and traditional encryption and authentication methods cannot detect such internal attacks. This invention addresses this by introducing a dual trust assessment mechanism, where neighboring nodes and the drone separately assess the trustworthiness of nodes. This comprehensive assessment effectively identifies malicious nodes and excludes them from the sampling point selection, thereby improving network security.

[0049] Existing technologies often suffer from unoptimized drone flight trajectories, resulting in long flight distances and high energy consumption due to drones flying along fixed paths. This invention addresses this by using an improved ant colony algorithm to dynamically optimize the drone's trajectory, traversing all data collection stations along the shortest path. This significantly reduces the drone's flight distance, lowers energy consumption, and improves data acquisition efficiency.

[0050] Existing technologies often suffer from a lack of comprehensive trust assessment due to their reliance on a single dimension. Most methods depend solely on the average of interaction results or recommendations, neglecting the overall rigor of trust evaluation. This invention addresses this issue by employing a dual trust mechanism that combines neighbor node evaluation with drone evaluation. This approach comprehensively assesses node trust from multiple dimensions, thereby improving the accuracy and reliability of trust assessment.

[0051] Existing technologies suffer from insufficient dynamic adaptability, failing to respond promptly to changes in network topology and node behavior. This invention, however, improves system adaptability by periodically updating node trust levels and dynamically adjusting data collection station selection, enabling trust assessment and data acquisition strategies to adjust dynamically with network conditions.

[0052] In summary, this invention outperforms existing technologies in terms of data acquisition efficiency, energy consumption balance, network security, path optimization, and trust assessment accuracy, providing an efficient, safe, and low-energy-consumption solution for UAV data acquisition systems. Attached Figure Description

[0053] Figure 1 This is a system flowchart of the method of the present invention.

[0054] Figure 2 This is a network model diagram of the method of the present invention.

[0055] Figure 3 This is a matrix diagram illustrating the selection of data packets in the method of the present invention.

[0056] Figure 4 This is a schematic diagram illustrating the selection of sampling points in the method of the present invention.

[0057] Figure 5 This is a schematic diagram illustrating the addition of moving sampling points in the method of the present invention.

[0058] Figure 6 This is a data packet structure diagram of the method of the present invention.

[0059] Figure 7 This is a comparison chart showing the changes in the number of malicious nodes identified by the method of this invention.

[0060] Figure 8 This is a comparison chart of the changes in the average movement trajectory distance of the UAV using the method of this invention.

[0061] Figure 9This is a comparison chart of the maximum energy consumption of nodes in the method of this invention.

[0062] Figure 10 This is a comparison chart of network lifecycle changes in the method of this invention. Detailed Implementation

[0063] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0064] Figure 1 This is a system flowchart of the method of the present invention. The present invention mainly includes three modules: a data collection module, a comprehensive trust calculation module, and a trajectory optimization module.

[0065] Step (1) uses the trust level of the sensor node as input and filters the nodes according to the trust threshold to obtain a set of trustworthy nodes.

[0066] Step (1.1) compares the evaluation trust value of each sensor node with the preset trust threshold.

[0067] Step (1.2): Nodes whose evaluation trust value is greater than or equal to the preset trust threshold are determined as trusted nodes and form a set of trusted nodes;

[0068] Step (1.3) identifies nodes whose assessed trust value is less than the preset trust threshold as malicious nodes and excludes them from the selection of sampling points.

[0069] Step (1) Filter trusted nodes by the trust level of sensor nodes, so that the data center can collect data packets from “good” nodes as much as possible, thereby enhancing network security.

[0070] Step (2) selects the nodes in the set of trusted nodes as candidate sampling points and selects the sampling points according to the matrix completion technique to obtain the set of data collection stations.

[0071] Step (2.1) involves constructing an information matrix from the sensor nodes within the monitoring area, and determining the number of sampling points to be selected in each column using matrix completion techniques. The information matrix is ​​as follows:

[0072] (1)

[0073] In the formula, Indicates the location of the sampling point. It is the time for collecting data packets. Indicates IUVA's location In time Information collected at that time Indicates IUVA's location In time No information was collected at the time.

[0074] Matrix for restoring information for:

[0075] (2)

[0076] In the formula, Indicates the location In time Recovery information at that time.

[0077] Based on matrix completion technology, the total number of selected sampling points is:

[0078] (3)

[0079] In the formula, It is a constant.

[0080] The number of sampling points selected in each column is:

[0081] (4)

[0082] In the formula, Indicates the time when data packets were collected.

[0083] Step (2.2): For the first column, select sampling points based on a uniform probability distribution. The probability of being selected as a sampling point is:

[0084] ( (5)

[0085] In the formula, probability This refers to the probability that a grid cell (row i, column j) is selected as a sampling point. Indicates the location of the sampling point.

[0086] Step (2.3): For subsequent columns, calculate the sampling probability based on the number of historically selected sampling points in each row, and select sampling points in descending order of the probability until the required number of sampling points for each column is reached, thus obtaining the data collection station set.

[0087] For the List , The calculation is as follows:

[0088] (6)

[0089] In the formula, the number of sampling points selected from column 1 to column (j-1) in row i is: For column j, the number of selected sampling points from column 1 to (j-1) in different rows are respectively }, in the collection

[0090] combine In the middle, the maximum value is The minimum value is .

[0091] If all sampling points If all values ​​are the same, the system will randomly select from this column. One sampling point.

[0092] Step (2) selects sampling points based on matrix completion technology, so that the system collects data packets only at a few locations, reducing data redundancy, reducing estimation error, and ensuring higher estimation accuracy.

[0093] Step (3) takes the data collection station set as input and performs path optimization based on the improved ant colony algorithm to obtain the optimized movement trajectory of the UAV.

[0094] Step (3.1) takes the nodes in the data collection station set as nodes in the graph and assigns initial pheromones to the edges between the nodes.

[0095] Step (3.2): For each iteration, the UAV starts from the starting point, calculates the path selection probability based on the pheromones on the edges, and selects the next unvisited node based on the probability. The path selection probabilities are as follows:

[0096] (7)

[0097] In the formula, the path selection probability This refers to the node Select node The probability of path selection. It is the edge The pheromone on the surface, its initial value is ; From node To the node Path heuristic factors; It is a drone from the node The set of nodes for the next step after starting.

[0098] Step (3.3): After the UAV completes its access to all data collection stations, record the path length for this iteration and update the pheromones on each edge according to the path length. The shorter the path length, the greater the pheromone increment. The formula for pheromone update is as follows:

[0099] (8)

[0100] In the formula, It is the pheromone evaporation coefficient, and .

[0101] These are pheromones released by the drone, and their values ​​are shown below:

[0102] (9)

[0103] The values ​​are as follows:

[0104] (10)

[0105] In the formula, This is the optimal routing path in this iteration. Optimal routing path The edges in will be added The pheromones. Among them It is a parameter. It is a routing path The length.

[0106] Step (3.4) is repeated iteratively until the preset number of iterations is reached, and the optimal path is used as the optimized movement trajectory of the UAV.

[0107] Step (3) uses the Elite Ant Trajectory Optimization Algorithm to accelerate the convergence of the shortest path, effectively shorten the movement trajectory length of the UAV, and significantly reduce the energy consumption of the UAV.

[0108] Step (4) takes the number of successful and failed direct interactions between a node and its neighboring nodes within a preset time period as input, updates the node's trust level according to the trust evaluation mechanism, and obtains the node trust level evaluated by the neighboring nodes.

[0109] Step (4.1) Record the number of successful interactions and the number of failed interactions between the node and its neighboring nodes within a preset time period.

[0110] Step (4.2): Based on the historical number of successful interactions and the historical number of failed interactions, combined with the number of interactions within the preset time period, update the cumulative number of successful interactions and the cumulative number of failed interactions. The number of successful or failed interactions for a node is shown below:

[0111] (11)

[0112] In the formula, and Indicates that the node comes from to The number of successful or failed interactions during the period; time , ; and Indicates the node within a time period The number of successful or failed interactions within the scope; It is a coefficient for the number of update interactions.

[0113] The values ​​are shown below:

[0114] (12)

[0115] In the formula, Indicates from arrive The difference between the cumulative historical successful interaction rate and the most recent successful interaction rate, and .

[0116] if or This indicates that the node The behavior is consistent and reliable, and the number of recent successful interactions is reliable. Therefore, .

[0117] like , then it represents a node Their behaviors differ significantly. In the first scenario... , i.e., node The historical success rate of interactions is lower than the recent success rate. This effectively prevents malicious nodes from rapidly increasing the success rate of interactions through deception or masquerading, thereby enhancing trust. The second scenario... ,node When the historical success rate of a node is higher than its recent success rate, the probability of the node engaging in malicious behavior increases accordingly. The system can quickly identify malicious nodes.

[0118] Step (4.3): Based on the updated cumulative successful interactions and cumulative failed interactions, calculate the node's trust value for its neighboring nodes. exist Time to node The trust level assessment is as follows:

[0119] (13)

[0120] Step (4.4) involves fusing the trust values ​​of all neighboring nodes for that node to obtain the neighboring trust score of that node. Neighboring nodes exist Time to node The trust level assessment is as follows:

[0121] (14)

[0122] Step (4) evaluates trust calculation based on the number of successful and failed direct interactions between adjacent nodes, effectively preventing malicious nodes from rapidly increasing the success rate of interactions and increasing node trust through deception or disguise. At the same time, the system can quickly identify malicious nodes.

[0123] Step (5): Using the data packets collected by the UAV as input, the node trust is calculated based on the difference between the values ​​perceived by the node and the data packets collected by the UAV from the data collection station, thus obtaining the UAV's assessed trust value. The node trust formula is as follows:

[0124] (15)

[0125] Step (5) assesses trust by evaluating the differences between the data packets received by the UAV before and after, avoids the corruption of trust results by false and forged data packets, achieves accurate calculation of UAV trust, and improves the reliability of node trust.

[0126] Step (6) merges the trust values ​​assessed by neighboring nodes with the trust values ​​assessed by the drone to obtain the overall trust value of the node. (Comprehensive Trust Level) As shown below.

[0127] (16)

[0128] Step (6) assesses the trust value of nodes by evaluating the trust value of neighboring nodes and the trust value of drones, thereby achieving a comprehensive assessment of node trust, improving the accuracy and robustness of trust assessment, and enhancing the security and stability of the drone network.

[0129] Figure 2This is a network model diagram of the method of this invention. It includes a cluster of sensor nodes and an Intelligent Unmanned Aerial Vehicle (IUVA). Each sensor node contains a starting point and a target point, as well as several other nodes divided into clusters. Each cluster has a cluster head node (CH). Sensor nodes in the network acquire sensor information from the surrounding environment and transmit data packets to their respective CHs. The CHs then send the sensor information to the IUVA. The IUVA is used to collect sensor information from the CHs and is considered to have sufficient energy and data storage space, eliminating the need for landing and recharging. It performs data collection and forwarding tasks according to a specified path.

[0130] Figure 3 This is a matrix diagram illustrating the data packet selection method of the present invention. The present invention collects data packets from a set of trusted sensor nodes. Through matrix completion technology, the locations of sampling points in the monitoring area and the time of data packet collection can be constructed into a matrix, recording the information collected by the nodes. Using matrix completion technology, only a small number of nodes need to send data packets to the CH (Chief Sensor), allowing some data packets within the monitoring area to be collected. Subsequently, matrix completion technology can be used to recover all the information within the entire monitoring area. The total number of collected data packets is significantly reduced, resulting in a substantial decrease in energy consumption.

[0131] Figure 4 This is a schematic diagram illustrating the sampling point selection method of the present invention. The system collects data packets only at a few locations. The grid in each column of the matrix is ​​divided into white and green portions. Each column has a different priority when selecting sampling points. To ensure accuracy in estimation, the sampling points are selected column-wise. The probability of a sampling point being selected is adjusted based on the number of sampling points selected in the current row. That is, the more sampling points selected in a row, the lower the probability of a sampling point being selected in that column, and vice versa.

[0132] Figure 5 This is a schematic diagram illustrating the addition of moving sampling points in the method of this invention. The optimized drone trajectory in this invention will pass through other sampling points. To determine the location of malicious nodes, the drone collects data packets sent from these sampling points along its path. The more data packets the drone collects, the more reliable its trust assessment of the nodes becomes.

[0133] Figure 6 This is a data packet structure diagram of the method of the present invention. Each data packet consists of the following parts: a data field, which displays the content of the data packet; a source identifier, which displays the identifier of the source node that generated the data packet; a destination identifier, which displays the identifier of the destination node that received the data packet; and a final identifier, which displays the identifier of the final node that sent the data packet.

[0134] Figure 7This is a comparison chart showing the change in the number of malicious nodes identified by the method of this invention. The TEMC scheme identifies more malicious nodes than the SRMC scheme. In the SRMC scheme, nodes evaluate the trust level of neighboring nodes based on the number of successful and failed interactions, while drones do not evaluate the trust level of neighboring nodes. Malicious nodes may make false trust assessments of neighboring nodes to disrupt the system, leading to incorrect decisions by the data center. However, this drone system also evaluates the trust level of nodes in their movement trajectories in the TEMC scheme. JAVs are considered trusted nodes in the system. The trust level assessments of nodes are mostly reliable. Therefore, the number of malicious nodes identified in the TEMC scheme is greater than that in the SRMC scheme. This demonstrates the effectiveness of the scheme.

[0135] Figure 8 This is a comparison chart of the average movement trajectory distance changes of the UAVs using the method of this invention. Under the TEMC scheme, the movement distance of the UAV varies across different rounds, while under the SRMC scheme, the movement distance of the UAV is greater than the other two schemes. The movement trajectory of the UAV does not change with rounds under the SRMC and TAG schemes. This is because the location of the sampling points in the TEMC scheme varies with rounds, resulting in differences in the movement distance of the UAV across different rounds. In the TAG scheme, each cluster selects a node as the CH (Chain Leader), and then optimizes the movement trajectory to consider all CH cases. However, the SRMC scheme does not optimize the movement trajectory of the UAV. Since neither the CH nor the movement trajectory of the UAV changes in the SRMC scheme, the value of the UAV movement distance remains unchanged in both schemes.

[0136] Figure 9 This is a comparison chart of the maximum node energy consumption of the method of this invention. Compared with other schemes, the maximum node energy consumption is lowest in the TEMC scheme, while it is highest in the TAG scheme. This is because the TEMC scheme uses matrix completion technology to select sampling points; the selected sampling points can transmit data packets, while other nodes do not need to send data packets to the CH (CH). The TEMC scheme then uses an optimized ant colony algorithm to optimize the movement trajectory of the selected sampling points. In the SRMC scheme, each cluster selects one node as the CH; the SRMC scheme also uses matrix completion technology to select sampling points, but the sampling points in each cluster only transmit data packets to their CH. The SRMC scheme does not use an optimized ant colony algorithm. However, in the TAG scheme, each cluster selects one node as the CH, and other nodes in the cluster send data packets to the CH. The TAG scheme then uses an optimized ant colony algorithm to optimize the movement trajectory of the UAV, while also considering the existence of the CH. Therefore, the number of data packets transmitted by nodes in the TEMC and SRMC schemes is less, while the number of data packets transmitted by nodes in the TAG scheme is more.

[0137] Figure 10 This is a comparison chart of network lifecycle changes in the method of this invention. Compared to other schemes, the TEMC scheme has the longest network lifecycle, while the TAG scheme has the shortest. The network lifecycle in all three schemes decreases with increasing round number. The network lifecycle is determined by the maximum energy consumption of network nodes. The TAG scheme has the highest energy consumption per node, while the TEMC scheme has the lowest, demonstrating the effectiveness of the TEMC scheme.

[0138] In summary, the TEMC scheme for the Internet of Things (IoT) described in this invention ensures network security while maintaining network lifespan. In the TEMC scheme, the locations of sampling points and the time of data packet collection together form a matrix. The locations of the sampling points constitute the rows of the matrix, while the time of data packet collection at each sampling point constitutes the columns. Sampling points are selected using matrix completion technology, selecting only a small portion at a time, and then the drone's movement trajectory within that time period is optimized using an optimized ant colony algorithm. Therefore, because the number of data packets collected in the TEMC scheme is relatively small, it effectively reduces node energy consumption and improves network lifespan. Simultaneously, to ensure network security, the TEMC scheme proposes a security mechanism to identify malicious nodes. In this mechanism, the trustworthiness of a node is evaluated not only by neighboring nodes but also by the drone itself. This mechanism effectively prevents malicious nodes from damaging the network and also significantly improves the accuracy of node trustworthiness identification.

[0139] Furthermore, it should be noted that in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element. Since this invention is a UAV data acquisition optimization method based on matrix completion and trust assessment, which can stack other factors affecting trust values, to avoid confusion, all references to this invention refer to this UAV data acquisition optimization method based on matrix completion and trust assessment.

[0140] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0141] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing UAV data acquisition based on matrix completion and trust assessment, characterized in that, The method includes: Using the trust level of sensor nodes as input, nodes are filtered based on the trust threshold to obtain a set of trustworthy nodes; Nodes in the set of trusted nodes are used as candidate sampling points, and sampling points are selected based on matrix completion techniques to obtain a set of data collection stations. Using a set of data collection stations as input, the improved ant colony algorithm is used for path optimization to obtain the optimized movement trajectory of the drone. The node trust level is updated based on the number of successful and failed direct interactions between the node and its neighboring nodes within a preset time period, and the node trust level is obtained by evaluating the neighboring nodes. Using the data packets collected by the drone as input, the node trust level is updated according to the trust assessment mechanism to obtain the updated node trust level. The overall trust value of a node is obtained by fusing the trust values ​​of neighboring nodes with those of the drone.

2. The UAV data acquisition optimization method based on matrix completion and trust assessment according to claim 1, characterized in that, The process involves using the trust level of sensor nodes as input and filtering nodes based on a trust threshold to obtain a set of trustworthy nodes. The evaluation trust value of each sensor node is compared with the preset trust threshold; Nodes whose assessed trust value is greater than or equal to the preset trust threshold are identified as trusted nodes and a set of trusted nodes is formed. Nodes whose assessed trust value is less than the preset trust threshold are identified as malicious nodes and excluded from the selection of sampling points.

3. The UAV data acquisition optimization method based on matrix completion and trust assessment according to claim 1, characterized in that, The process involves selecting candidate sampling points from the set of trusted nodes using matrix completion techniques, resulting in the following data collection station set: The sensor nodes within the monitoring area are constructed into an information matrix, and the number of sampling points to be selected in each column is determined based on matrix completion technology; For the first column, sampling points are selected based on the uniform distribution probability; For subsequent columns, the sampling probability is calculated based on the number of historically selected sampling points in each row, and sampling points are selected sequentially from high to low according to the probability until the required number of sampling points for each column is reached, thus obtaining the data collection station set.

4. The UAV data acquisition optimization method based on matrix completion and trust assessment according to claim 1, characterized in that, The process involves using a set of data collection stations as input and optimizing the path using an improved ant colony algorithm to obtain the optimized movement trajectory of the UAV: The nodes in the data collection station set are treated as nodes in the graph, and the edges between the nodes are assigned initial pheromones. For each iteration, the drone starts from the starting point, calculates the path selection probability based on the pheromones on the edges, and selects the next unvisited node based on the probability. After the drone completes its access to all data collection stations, it records the path length of this iteration and updates the pheromones on each edge according to the path length. The shorter the path length, the greater the pheromone increment. Repeat the iteration until the preset number of iterations is reached, and use the optimal path as the optimized movement trajectory of the drone.

5. The UAV data acquisition optimization method based on matrix completion and trust assessment according to claim 1, characterized in that, The node trust score is updated based on the number of successful and failed direct interactions between the node and its neighboring nodes within a preset time period, using the number of such interactions as input and a trust evaluation mechanism. The node trust score evaluated by the neighboring nodes is then obtained as follows: Record the number of successful interactions and the number of failed interactions between a node and its neighboring nodes within a preset time period; Based on the number of successful interactions and the number of failed interactions in the past, and combined with the number of interactions within the preset time period, update the cumulative number of successful interactions and the cumulative number of failed interactions; Based on the updated cumulative number of successful interactions and cumulative number of failed interactions, calculate the node's trust value for its neighboring nodes; The neighbor evaluation trust values ​​of a node are merged to obtain the neighbor evaluation trust degree of that node.

6. The UAV data acquisition optimization method based on matrix completion and trust assessment according to claim 1, characterized in that, The process involves using data packets collected by the drone from the data collection station as input, updating the node trust level based on the trust assessment mechanism, and obtaining the node trust level assessed by the drone as follows: The data packets collected by the drone are input, and the node trust is calculated based on the difference between the value perceived by the node and the data packets collected by the drone from the data collection station, thus obtaining the drone's assessment trust value.

7. The UAV data acquisition optimization method based on matrix completion and trust assessment according to claim 1, characterized in that, The process of fusing the trust values ​​of neighboring nodes with the trust values ​​of the drone yields the overall trust value of the node: The trust values ​​of neighboring nodes and the drone are used as inputs, and the overall trust is calculated using a weighted fusion method.

Images