Multi-source trust analysis method based on blockchain and IoT edge task collaboration

Abstract

The present application relates to a multi-source trust analysis method based on blockchain and IoT edge task collaboration. The method comprises: upon receiving a collaboration request from a device, obtaining the resource threshold required for the device's current task, obtaining the resource vector owned by each device in the device set corresponding to the current task, analyzing the capability matching degree between each device in the device set and the device's current task, analyzing the capability trust of each device in the device set at the current moment, obtaining the direct trust of each device, calculating the weight of its feedback trust and indirect feedback trust, and then combining the direct trust, capability trust, and feedback trust to analyze the global trust value of each device j in the device set D, updating the historical global trust value, and determining the device with the highest global trust value from the device set D to collaborate with device i on the current task, thereby improving data security.

Description

Technical Field

[0001] The present application relates to the field of Internet of Things technology, and in particular to a multi-source trust analysis method based on blockchain and Internet of Things edge task collaboration. Background Art

[0002] With the rapid development of IoT and 5G communication technologies, which now provide more reliable and powerful communication support, there is growing demand for more advanced mobile services, which means that applications will become more complex and time-sensitive. The rapid increase in the number of mobile user devices is significantly increasing the transmission pressure on communication networks and the computing pressure on cloud computing centers. To address this problem, edge computing has been proposed in recent years as a method for processing the large amounts of data generated by IoT edge devices. As computing resource providers, IoT edge devices can interact in real time and collaboratively process the generated data. This reduces communication latency and conserves computing resources, significantly alleviating the transmission pressure on communication networks and the computing pressure on cloud computing centers.

[0003] Due to the open, dynamic, and autonomous nature of edge devices, as well as the lack of robust basic protections, trust between devices is lacking. This has led to numerous attacks, such as message forgery, message tampering, and reply attacks, which can lead to device fraud, easily cause issues with task writing between edge devices, and expose large amounts of private data. Therefore, establishing trust in edge devices and effectively distinguishing malicious devices has become a major challenge in IoT edge task collaboration. Trust mechanisms are particularly effective in combating internal network attacks, enabling network operators to effectively reward honest devices and punish malicious ones. By evaluating the service quality of different devices, malicious devices can be filtered out and access control issues can be efficiently addressed, ensuring the reliability of edge computing networks in authorization, authentication, and service provision. However, due to the widespread use of trust mechanisms, various security attacks have been designed to deceive edge computing networks and mislead trust mechanisms.

[0004] In related technologies, for the trust issue of edge computing, trust assessment is based entirely on information such as historical interaction records of edge devices to establish trust. The accuracy of trust calculation is low, and the possibility of tampering by malicious devices is high, resulting in low data security. Summary of the Invention

[0005] Based on this, it is necessary to address the above technical problems and provide a multi-source trust analysis method based on the collaboration of blockchain and IoT edge tasks, which can improve the accuracy of trust calculation, reduce the possibility of tampering by malicious devices, and improve data security.

[0006] A multi-source trust analysis method based on blockchain and IoT edge task collaboration is applied to an edge proxy server. The method includes:

[0007] When receiving a collaboration request from device i, obtain the resource threshold T required for the current task of device i. ik ={δ i1 ,δ i2 ,……,δ in}, i∈D, D={1, 2, 3...m}, where D is the device set, m is the number of devices in the device set, δ in Indicates the size of the nth capability resource required by the current task of device i;

[0008] Get the resource vector N owned by each device j in the device set D corresponding to the current task jk ={ε j1 ,ε j2 ,……,ε jn}, j∈D, j≠i, D={1, 2, 3……m}, where ε jn Indicates the size of the nth resource owned by device j for the current task;

[0009] According to the resource threshold required by the current task of device i and the resource vector of each device j in the device set D corresponding to the current task, the matching degree of the current task capability of each device j in the device set D and device i is analyzed.

[0010] According to the capability matching degree of each device j in the device set D at the current moment, the capability trust of device j in the device set D at the current moment is analyzed. Among them, H(P ij ) is the capability matching degree P of the current task between device j and device i ij Information entropy of

[0011] Obtain the direct trust between devices 1 to m in device set D and each device j in device set D;

[0012] Calculate the weight of the feedback trust based on the number of interactions between devices 1 to m in device set D and each device j in device set D in, is the number of interactions between device i and device j in the current time window t;

[0013] Analyze the indirect feedback trust of each device j in the device set D Among them, ω zj represents the weight of device z’s trust in device j’s feedback, is the direct trust between devices z and j in the current time window t;

[0014] Combine the direct trust, capability trust and feedback trust to analyze the global trust value of each device j in the device set D

[0015] The historical global trust value is updated according to the global trust value of each device j in the device set D, and the device with the highest global trust value is determined from the device set D to collaborate with the device i on the current task.

[0016] The multi-source trust analysis method based on blockchain and IoT edge task collaboration obtains the resource threshold required for the current task of device i when receiving a collaboration request from device i, and obtains the resource vector of each device j in the device set D corresponding to the current task. Then, according to the resource threshold required for the current task of device i and the resource vector of each device j in the device set D corresponding to the current task, the capability matching degree of each device j in the device set D with the current task of device i is analyzed, and according to the capability matching degree of each device j in the device set D at the current moment, the capability trust of each device j in the device set D at the current moment is analyzed, and the capability matching degree of each device j in the device set D at the current moment is obtained. The direct trust between device m and each device j in device set D is calculated, and the weight of its feedback trust is calculated according to the number of interactions between device 1 to device m in device set D and each device j in device set D, and the indirect feedback trust of each device j in device set D is analyzed. Then, the global trust value of each device j in device set D is analyzed by combining the direct trust, the capability trust and the feedback trust, thereby improving the accuracy of trust calculation and reducing the possibility of tampering by malicious devices. The historical global trust value is updated according to the global trust value of each device j in device set D, and the device with the highest global trust value is determined from device set D to collaborate with device i on the current task, thereby improving data security. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of the overall architecture of a multi-source trust analysis method based on blockchain and IoT edge task collaboration in one embodiment;

[0018] Figure 2 Schematic diagram of a trusted computing architecture based on multi-source feedback and blockchain in one embodiment;

[0019] Figure 3 A schematic diagram of the blockchain system model structure in one embodiment;

[0020] Figure 4 1. A flowchart of a multi-source trust analysis method based on blockchain and IoT edge task collaboration in one embodiment;

[0021] Figure 5 FIG1 is a schematic diagram showing a comparison result of global aggregation time in a network environment with 20% malicious devices and 20% cooperative devices in one embodiment;

[0022] Figure 6 FIG. 1 is a schematic diagram showing a comparison result of global aggregation time in a network environment with 40% malicious devices and 40% cooperative devices in one embodiment;

[0023] Figure 7 FIG1 is a schematic diagram showing a comparison result of task failure rates in a network environment with 10% malicious devices and 10% cooperative devices in one embodiment;

[0024] Figure 8 FIG1 is a schematic diagram showing a comparison result of task failure rates in a network environment with 20% malicious devices and 20% cooperative devices in one embodiment;

[0025] Figure 9 2 is a schematic diagram showing comparison results of task failure rates in a network environment with 40% malicious devices and 40% cooperative devices in one embodiment. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0027] In related technologies, for the trust issue of edge computing, trust assessment is based entirely on information such as historical interaction records of edge devices to establish trust. The accuracy of trust calculation is low, and the possibility of tampering by malicious devices is high, resulting in low data security.

[0028] This application provides a multi-source trust analysis method based on blockchain and IoT edge task collaboration. It uses edge proxy servers to replace cloud computing centers. Trust evaluation calculations are mainly performed by edge proxy servers, which solves the problem of excessive resource consumption of edge devices. Based on multi-source feedback calculations, this application also takes into account the interaction between devices and edge proxy servers, making the evaluation of device trust more comprehensive and accurate. In response to the problems of heterogeneity and instability of device resources, capability trust evaluation is adopted to evaluate the device capabilities of collaborators before task collaboration. At the same time, a feedback information fusion algorithm based on objective information entropy theory is adopted to overcome the limitations of artificial or subjective weighted trust factors in traditional trust schemes.

[0029] To protect device privacy and interaction data, this patent also utilizes an IoT consortium chain with elastic block generation and dynamic device management to maintain records of edge collaboration interactions and update trust values ​​for edge devices. The tasks, results, and real-time trust values ​​of nodes in the collaboration process are recorded in a blockchain, publicly accessible and verifiable by other edge nodes. This prevents tampering by malicious devices and ensures data security. Leveraging smart contracts, the blockchain automatically penalizes misbehaving nodes and holds stakeholders accountable upon detection of any untrustworthy collaboration. The chain categorizes transactions into urgent and ordinary transactions based on their time sensitivity within the application. Task interaction records are ordinary transactions, while trust value updates are urgent transactions. This prioritizes trust updates, expedites device trust value updates, and improves the efficiency of edge task collaboration. Using an edge proxy server to maintain the chain conserves computing and storage resources on edge devices while protecting device privacy and data security.

[0030] The task collaboration computing environment of the multi-source trust analysis method based on blockchain and IoT edge task collaboration provided in this application is divided into three layers: network layer, proxy layer and device layer. The generation and collection of data originate from the device layer, and the processing of data and the calculation of the trust value of the device mainly rely on the proxy layer. The global trust value of the device is calculated based on four trust types: direct trust between edge devices (based on direct interaction records between edge devices), direct trust from edge proxy servers, feedback trust from other edge devices, and device capability trust. Among them, direct trust is a subjective evaluation of the service quality provided by edge devices, and feedback can provide an efficient and reliable method for establishing a trust relationship based on credibility between edge devices.

[0031] In IoT edge computing environments, a multi-source feedback trust aggregation model based on time decay and interaction frequency can effectively detect malicious nodes and further identify malicious nodes with dynamic identities. To address attack challenges such as badmouthing, goodmouthing, and collusion attacks in trust assessment mechanisms, this application designs a multi-source feedback trust aggregation model based on time decay and interaction frequency, introducing a trust assessment mechanism to measure the reliability of resource nodes. First, the resource provider calculates direct trust based on time decay. Then, the edge agent calculates feedback trust based on interaction frequency and sends it to the task requester. Finally, the task requester integrates its own calculated direct trust and aggregated feedback trust into the resource provider's global trust. This multi-source feedback trust aggregation model provides a reliable basis for resource allocation and scheduling. Furthermore, an IoT consortium chain with elastic block generation and dynamic device management is used to maintain interaction records of edge task collaboration and trust feedback from edge devices. This chain stores records that can be publicly accessed and verified by other devices, preventing tampering by malicious devices and ensuring data security and privacy.

[0032] The computing resource providers for edge task collaboration are dynamic, complex, and unreliable, making their collaborative services difficult to trust. In this environment, the interaction between the task requester and the resource provider usually cannot guarantee the service behavior of the resource node. Insufficient information about the resource node will make it impossible to determine which resources can be matched, thereby reducing system efficiency and making the system more unreliable. Therefore, effectively evaluating the behavior of resource nodes through the edge proxy server is conducive to allocating node resources to tasks. The trust value will also be dynamically updated with the behavior of the resource node to ensure the timeliness of the trust. Behavior-based trust is used to evaluate the credibility of the node, which is generally determined by the historical behavior records and current behavior characteristics of the evaluated node. A large number of research results show that an effective trust mechanism can ensure the reliable development of the network and efficient collaboration of tasks. The definitions involved in this application are as follows.

[0033] Definition 1. Direct trust of device i or edge proxy server k in another device j (including: D_D direct trust and B_D direct trust): D_D direct trust is based on the historical interaction records between device j and device i, and B_D direct trust is based on the historical interaction records between device j and edge proxy server k. They are evaluations of a device's capabilities in collaborative tasks.

[0034] Definition 2. Indirect feedback trust of edge proxy server k on device j (i.e., indirect feedback trust of device i on device j, B_D indirect feedback trust): Edge proxy server k receives the trust value uploaded from the edge device, sets a weight for each related device that directly interacts with j based on the interaction frequency, and performs a weighted calculation on all related devices that directly interact with j to obtain a comprehensive evaluation.

[0035] Definition 3. Capability Trust: Capability trust indicators include multiple aspects, such as available time, CPU computing power, energy maintenance capability, time cost of each business unit, storage space size, transmission bandwidth, etc.

[0036] Definition 4. Global trust of edge proxy server k on device j: It is a comprehensive evaluation of edge device j, which is obtained by aggregating capability trust, D_D direct trust, B_D direct trust and B_D indirect feedback trust.

[0037] Among them, the overall architecture diagram of the multi-source trust analysis method based on blockchain and IoT edge task collaboration is as follows: Figure 1 shown. Figure 1The main components of the IoT are task requesters, resource providers, and edge proxy servers. Both task requesters and resource providers are edge devices. In the architecture of an edge task collaborative computing network based on the Internet of Things and multi-source feedback, edge proxy servers undertake some data analysis, processing, and management tasks, bringing various potential advantages. For example, edge proxy servers provide a wide range of services to a large number of heterogeneous edge devices, quickly processing data from IoT edge devices, thereby reducing network latency, quickly responding to task requesters' task requests, and addressing the task requirements and QoS requirements of large-scale IoT systems. Edge proxy servers primarily undertake three tasks: computing feedback trust, aggregating global trust, maintaining a consortium chain to ensure the security of interactive data, and allocating appropriate resource providers to jointly assist in completing IoT system tasks.

[0038] The reliable trust computing architecture based on multi-source feedback and blockchain is divided into two layers: the device layer and the agent layer. Their composition and functions are as follows.

[0039] 1) The device layer consists of a large number of heterogeneous devices. During task collaboration, collaborating devices transmit data using technologies such as Wi-Fi and communicate with edge proxy servers via wireless communication base stations or gateways. Before completing a collaborative task, edge devices send request messages to the edge proxy server to ensure the reliability of the interacting edge devices.

[0040] 2) Due to the open IoT network, there may be some malicious edge devices that generate malicious feedback, making the trust calculation of the proxy layer unreliable. Therefore, this application expands the traditional feedback mechanism. The edge proxy server is used to monitor the interactive behavior of edge devices, summarize the feedback information of edge devices, and feed it back to the edge devices. In addition, unlike the traditional feedback mechanism, this patent uses the edge proxy server to perform trust aggregation calculation based on multi-source feedback information, which not only reduces transmission costs and latency, but also improves computing efficiency and reliability.

[0041] Among them, the diagram of the trusted computing architecture based on multi-source feedback and blockchain is as follows: Figure 2 As shown in Figure 1, the device layer generates direct trust, which is fed back to the agent layer. The agent layer generates recommended trust for the device layer through computational aggregation. For example, B1 is responsible for monitoring the service behavior of edge devices. Edge devices periodically feed back information about S i B1 calculates the feedback trust and global trust between different edge devices. If device S i Hope to get device S j The global trust value of S i Send a request to B1, and B1 feeds the result back to S i .

[0042] Among them, the schematic diagram of the blockchain system model structure is as follows Figure 3 As shown, the IoT consortium chain used in this application consists of two types of nodes: 1) edge proxy server node (SN) and 2) member node (MN). MN refers to the edge device registered in the SN and the blockchain network. SN in this application is the edge proxy server. Each device is marked with a universal identifier in the hardware. It is managed by the SN. Each SN is responsible for generating system parameters and managing member nodes. There is a local storage on the SN for storing the raw data generated by the member node devices. In addition, SN is also a complete node of the blockchain because they save records of the entire block of the blockchain. Each SN can be selected as a miner or validator in different rounds. Both SN and MN can send and receive transactions in the blockchain, which are defined as MN transactions and SN transactions respectively.

[0043] This application adopts an IoT application-oriented consortium chain. For example, some applications may generate a large number of new transactions in a short period of time (for example, the interactive collaboration records of edge devices in this patent). These are ordinary transactions because they are not critical transactions that need to be processed immediately. All transactions are sent to the edge proxy server node (SN), where v transactions are integrated into the SN transaction for efficient verification and storage. The SN then sends its transactions to the verifier. The miners package these transactions into a common block. In this patent solution, urgent access to data may be required (for example, the trust value of the edge device, which needs to be updated in a timely manner), and such transactions are urgent transactions. In this case, the data requesting device MN sends an urgent MN transaction directly to the verifier. Within a predefined interval, the miners integrate the urgent transactions into the urgent block.

[0044] In the IoT environment, edge devices can dynamically join and leave, posing risks to the efficiency and stability of task execution. Therefore, to address the issue of node trust evaluation, this application introduces a trust mechanism into the IoT task collaboration model. Trust values ​​are generated based on the historical behavior of edge devices to measure their trustworthiness and are regularly updated as edge devices perform their actions. A multi-source feedback trust aggregation mechanism based on time decay is established, and a trust model is designed to evaluate the trust values ​​of service nodes, detect nodes with anomalous behavior, and enhance the reliability of the task collaboration system. The edge computing trust system should dynamically evaluate the trust of service resources to ensure that resource nodes that meet user requirements participate in completing the requested service tasks. This model involves two entities: a device set D = {1,…,i,…,j,…,m}, with m devices, and a set of edge proxy servers B = {1,…,k,…,s}, with s edge proxy servers.

[0045] In one embodiment, Figure 4As shown in FIG, a multi-source trust analysis method based on blockchain and IoT edge task collaboration is provided. The method is applied to an edge proxy server as an example, and includes the following steps:

[0046] Step 1: When receiving a collaboration request from device i, obtain the resource threshold T required for the current task of device i. ik ={δ i1 ,δ i2 ,……,δ in}, i∈D, D={1, 2, 3...m}, where D is the device set, m is the number of devices in the device set, δ in Indicates the size of the nth capability resource required by the current task of device i.

[0047] Among them, the resource threshold T required by the current task ik It includes service duration, CPU computing power, energy endurance, time cost of each service unit, storage space size, and transmission bandwidth. n represents the number of evaluation indicators.

[0048] Step 2: Get the resource vector N of each device j in the device set D corresponding to the current task jk ={ε j1 ,ε j2 ,……,ε jn}, j∈D, j≠i, D={1, 2, 3……m}, where ε jn Indicates the size of the nth resource owned by device j for the current task.

[0049] Among them, device j corresponds to the resource vector N owned by the current task jk Including service duration, CPU computing power, energy endurance, time cost of each service unit, storage space size, and transmission bandwidth.

[0050] Step 3: Analyze the matching degree between the current task capability of each device j in device set D and device i based on the resource threshold required by the current task of device i and the resource vector of each device j in device set D corresponding to the current task.

[0051] Among them, P ij The matching degree of the current task capabilities of device j and device i is the minimum ratio of the resources available to the current task to the resources required for the task. Considering that the matching accuracy of different elements in the same task may vary, the minimum value is selected as the final value. In other words, the node resource shortage is used to represent the overall accuracy.

[0052] Step 4: Analyze the capability trust of each device j in device set D at the current moment based on the capability matching degree of each device j in device set D at the current moment. Among them, H(P ij ) is the capability matching degree P of the current task between device j and device i ij Information entropy.

[0053] Among them, the capability trust C of device j at the current moment ij The capability trust evaluation of device i on device j is based on the capability trust design. This effectively reduces the chances of low-capability devices participating in collaborative tasks. Device capability trust is determined by the requirements of the corresponding service task. Capability trust evaluation metrics include service duration, CPU computing power, energy endurance, time cost per service unit, storage space, and transmission bandwidth.

[0054] It should be understood that determining the capability trust of device j at the current moment from the perspective of information entropy and using information entropy to improve the accuracy of trust assessment results can effectively quantify the risks brought by low-capability devices.

[0055] Determining the capability trust of device j at the current moment from the perspective of information entropy can effectively describe the capability trust of IoT edge devices. The capability matching degree P of the current task of device j and device i is ij The calculation formula of information entropy is:

[0056] H(P ij )=-P ij log2P ij -(1-P ij )log2(1-P ij )

[0057] Among them, H(P ij ) is the capability matching degree P of the current task between device j and device i ij Information entropy.

[0058] Trust value is not only related to node capabilities but also to the success rate and duration of the services provided by the evaluated node in the past. Furthermore, the degree of trust also indicates the evaluated node's ability to provide reliable resources and stable interactions within the edge network. This application uses trust as a key parameter to measure the dynamic performance of node resources.

[0059] Step 5: Obtain the direct trust between device 1 to device m in device set D and each device j in device set D.

[0060] Direct trust is achieved by understanding the nature of the devices or interactions involved in task collaboration. Over a period of time, a record is generated based on the ratio of the number of successful task executions to the number of task requests. Let the service quality record between i and j based on the current time window t be as follows:

[0061]

[0062] in, is the service quality record between i and j in the current time window t, The ratio of successfully executed tasks to the total number of task requests within the current time window t. As time passes, the longer the time interval, the smaller the impact of the trust value, showing a downward trend. Therefore, its time decay function is as follows:

[0063]

[0064] Where r is a time slot in the current time window t, Indicates the time decay coefficient for calculating direct trust.

[0065] The direct trust of device i on device j in time slot r based on the time decay function D_D is expressed as:

[0066]

[0067] in, represents the service quality record of device i and device j at time slot r, represents its time attenuation coefficient, It represents the direct trust of device i on device j in time slot r.

[0068] Step 6: Calculate the weight of the feedback trust based on the number of interactions between device 1 to device m in device set D and each device j in device set D. in, is the number of interactions between device i and device j in the current time window t.

[0069] Indirect trust feedback means that proxy server k regularly broadcasts request messages within the regional cluster it manages. Simultaneously, all devices send trust values ​​back to the edge proxy server in the cluster. Edge proxy server k then stores these trust values ​​in a matrix.

[0070] Assume that there are m devices D = {1,2,…,j,…,m} in the edge computing cluster. The edge proxy server will periodically broadcast request messages in the device cluster. In response, all devices in the cluster will forward their direct trust values ​​to other devices, and then the edge proxy server will save these trust values ​​in the matrix In the example, the matrix is ​​as follows:

[0071]

[0072] in, is the direct trust between devices i and j in the current time window t. When i = j, this value needs to be discarded when feeding back to the proxy aggregation.

[0073] The direct trust collected by the edge proxy server is aggregated into an indirect trust value. In addition, the number of interactions also affects the trust value of the resource node. The more interactions there are, the higher the frequency between the two nodes, and the greater the impact of the feedback trust. For example, the weight of the feedback trust of device i to device j is recorded as ω ij :

[0074]

[0075] in, is the number of interactions between device i and device j in the current time window t.

[0076] Step 7: Analyze the indirect feedback trust of each device j in the device set D Among them, ω zj represents the weight of device z’s trust in device j’s feedback, is the direct trust between devices z and j in the current time window t.

[0077] Among them, indirect feedback trust is the weighted sum of the indirect feedback from m devices to device j.

[0078] Step 8: Combine the direct trust, capability trust and feedback trust to analyze the global trust value of each device j in the device set D.

[0079] Step 9: Update the historical global trust value according to the global trust value of each device j in the device set D, and determine the device with the highest global trust value from the device set D to collaborate with the device i on the current task.

[0080] The multi-source trust analysis method based on blockchain and IoT edge task collaboration obtains the resource threshold required for the current task of device i when receiving a collaboration request from device i, and obtains the resource vector of each device j in the device set D corresponding to the current task. Then, according to the resource threshold required for the current task of device i and the resource vector of each device j in the device set D corresponding to the current task, the capability matching degree of each device j in the device set D with the current task of device i is analyzed, and according to the capability matching degree of each device j in the device set D at the current moment, the capability trust of each device j in the device set D at the current moment is analyzed, and the capability trust of each device j in the device set D is obtained. The direct trust between device 1 to device m and each device j in device set D is calculated, and the weight of their feedback trust is calculated according to the number of interactions between device 1 to device m in device set D and each device j in device set D, and the indirect feedback trust of each device j in device set D is analyzed. Then, the global trust value of each device j in device set D is analyzed by combining the direct trust, the capability trust and the feedback trust, thereby improving the accuracy of trust calculation and reducing the possibility of tampering by malicious devices. According to the global trust value of each device j in device set D, the device with the highest global trust value is determined from device set D to collaborate with device i on the current task, thereby improving data security.

[0081] In one embodiment, the global trust value of the device j is for:

[0082]

[0083] in, is the global trust value of device j, α is the weight of device i’s direct trust in device j in the current time window t, β is the weight of edge proxy server k’s direct trust in device j in the current time window t, and γ is the weight of device i’s indirect feedback trust in device j in the current time window t. is the direct trust of device i on device j in the current time window t, is the direct trust of edge proxy server k on device j in the current time window t, is the indirect feedback trust of device i to device j in the current time window t.

[0084] Among them, the direct trust of edge proxy server k on device j in time slot r based on the time decay function B_D is expressed as:

[0085]

[0086] in, represents the service quality record of edge proxy server k and device j at time slot r, represents its time decay coefficient.

[0087] Here, α + β + γ = 1. If the interaction between devices is more successful, the two will trust each other more, and therefore the weight of direct trust (i.e., α and β) is higher.

[0088] In one embodiment, the weight of the direct trust of device i on device j in the current time window t is determined as follows:

[0089] According to the request frequency of device i to device j in the current time window t, the weight of direct trust of device i to device j in the current time window t is determined. The weight of direct trust of device i to device j in the current time window t is:

[0090] α=1-v n

[0091] Where v is a hyperparameter ranging from 0 to 1, and n is the number of interactions between devices i and j in the current time window t.

[0092] It should be understood that existing solutions often manually assign weights, resulting in a lack of flexibility in the weighting of trust factors. Based on the low latency and flexibility of edge computing, this application uses an edge proxy server to monitor requests from different devices within a certain timeframe and sets the weight of the trust factor based on the frequency of requests from device i to device j within the current time window.

[0093] In one embodiment, determining, based on the global trust values ​​of each device j in the device set D, a device with the highest global trust value from the device set D to collaborate with the device i on the current task includes:

[0094] According to the global trust value of each device j in the device set D, the global trust value update formula is used to update the historical global trust value to determine the latest global trust value of each device j in the device set D;

[0095] According to the latest global trust value of each device j in the device set D, a device with the highest global trust value is determined from the device set D to collaborate with the device i on the current task.

[0096] In one embodiment, the global trust value update formula is:

[0097]

[0098] in, is the latest global trust value, σ is the trust reward factor, and θ is the update adjustment factor.

[0099] In one embodiment, the update adjustment factor θ is:

[0100]

[0101] It should be understood that if the global trust value of the device is higher than its historical global trust value, the historical global trust value of the device should be increased appropriately; otherwise, its historical global trust value should be decreased. This trust update mechanism can effectively prevent the attack of good words.

[0102] To further demonstrate the advantages and beneficial effects of the multi-source trust analysis method based on blockchain and IoT edge task collaboration in this application, the following analysis is conducted:

[0103] (1) Complexity analysis

[0104] a) Space complexity

[0105] Assume there are m edge nodes, each with c neighboring nodes. Neighboring nodes (i.e., devices) feed the direct trust value of the evaluated node back to the edge proxy server. An evaluated edge node (i.e., device) has c messages, and there are m evaluated edge nodes. Therefore, a total of mc messages are stored locally by the edge proxy server, and the global trust value of m edge nodes is stored, resulting in a space complexity of O(2mc + m).

[0106] b) Time complexity

[0107] For direct trust between devices i and j, there are q windows used to record the number of interactions between the nodes within time period t. In the trust calculation system proposed in this application, it is assumed that there are m edge nodes, each with c neighbor nodes, that is, the time complexity is O(mtc). After receiving direct trust feedback, the edge proxy server needs to calculate an indirect trust value for each evaluated edge node, that is, O(mc). The time complexity is O(mtc+mc+m).

[0108] (2) Security Analysis

[0109] In various practical scenarios of edge task collaboration, different types of edge devices exist. Malicious devices can provide invalid services and feedback during interactions, causing collaborative services to crash and user information to be compromised. Assume that there are two types of edge devices in the system: normal devices and malicious devices. Normal devices provide valid services and provide accurate feedback. Malicious devices provide invalid services and provide misleading feedback. The following malicious attack models are considered: bad word attacks, good word attacks, and collusion attacks.

[0110] This application's multi-source trust analysis method based on blockchain and IoT edge task collaboration was experimentally compared with three approaches: TOSMFTM (Task Offloading Strategy Based on Multi-feedback Trust Mechanism), CLMT (Credible and Lightweight Multidimensional Trust), and RTCM (a reliable trust computing mechanism based on the integration of multi-source feedback and fog computing). Table 1 shows the comparison of the different approaches.

[0111] Table 1 Comparison of different solutions

[0112]

[0113] In trust computation mechanisms based on multi-source feedback, the primary threat comes from malicious feedback from IoT edge devices. Several performance mechanisms were designed to enable a comprehensive comparison with other trust mechanisms. The performance was evaluated primarily from two perspectives: computational efficiency and reliability under varying percentages of malicious devices and varying percentages of collaborative devices.

[0114] The simulation parameters used in the experiment are shown in Table 3. The simulation computer is configured with a CPU of 4.1G, a memory of 16G, and a hard disk of 1T. A total of 1000 devices are deployed in the simulator, and a total of 30 agents are deployed in the network. The total time step of the simulation run is 200, and the time window of the trust calculation is 20. The percentage of malicious devices is set to 10%, 20%, and 40% [1][2][7]. When the proportion of collaborative devices is 10%, 20%, and 40%, it means that the edge collaborative system is in the idle, busy, and highly busy states respectively.

[0115] Table 3 Simulation parameters

[0116]

[0117] The experiment uses the global aggregation time (R) to evaluate the computational efficiency of the proposed trust mechanism. R is the total trust aggregation time and can be used to evaluate the computational efficiency of the entire network system. Most previous work has focused on system behavior when the system is stable. This application considers the aggregation time as an important indicator for measuring the system's stability, especially in dynamic and large-scale IoT computing environments. R is calculated as follows:

[0118] R=t1-t0

[0119] Wherein, t1 is the timestamp of global aggregation completion, and t0 is the timestamp when device i requests the edge proxy server k for the global trust value of device j.

[0120] The failure rate of computing tasks is recorded as F, which is used to reflect the reliability of the trusted computing system. The lower the F value, the higher the reliability of the trust mechanism. In this set of experiments, it is assumed that most edge proxy servers in the IoT edge computing are trustworthy collaborators. This IoT edge computing environment is very similar to the actual situation, and most brokers are honest and trustworthy. This set of experiments considers several types of edge computing environments: (1) an honest and idle IoT computing environment; (2) a busy and dishonest IoT computing environment; (3) a highly dishonest and highly busy IoT computing environment. The calculation of F is as follows:

[0121]

[0122] Among them, x f is the number of task failures, and x is the total number of tasks, and the task failure rate can be calculated.

[0123] Computational efficiency evaluation: The comparison results of global aggregation time in a network environment with 20% malicious devices and 20% collaborative devices are shown in the figure below. Figure 5 shown. Figure 5 Comparative results of GCT are presented for an IoT edge computing network with 1,000 devices. In this set of experiments, the percentage of malicious nodes was set to 20%, and the percentage of collaborating devices was set to 20%, representing a busy and dishonest system, with 20% of the devices being dishonest and 20% requesting collaboration with other devices. From a global aggregation time perspective, the proposed trust mechanism achieved significantly shorter aggregation times than TOSMF™, CLMT, and RTCM. This demonstrates that in a dishonest and busy Internet computing environment, the proposed trust mechanism achieves superior computational efficiency compared to TOSMF™, CLMT, and RTCM.

[0124] The comparison results of global aggregation time in a network environment with 40% malicious devices and 40% collaborative devices are shown in the figure below. Figure 6 As shown. Figure 6 In the data, the proportion of malicious devices is 40% and the proportion of cooperative devices is 40%, which means that the system is a highly busy and highly dishonest system, in which 40% of the devices are dishonest and 40% of the devices require cooperation with other devices. Figure 6As can be seen, the proposed trust mechanism outperforms CLMT, TOSMFTM, and RTCM from the perspective of global convergence time. In this highly dishonest and highly busy IoT computing environment, the proposed trust mechanism achieves the shortest global convergence time. This demonstrates that the proposed trust mechanism is computationally more efficient than CLMT, TOSMFTM, and RTCM in this highly dishonest and highly busy IoT computing environment.

[0125] Resistance to malicious node evaluation: Figures 7 to 9 The results show a comparison of task failure rates for different proportions of malicious and cooperative devices. In this set of experiments, we assume that the IoT computing environment is trustworthy and all edge proxy servers are honest. The percentage of malicious devices is set to 10%, 20%, and 40%, representing honest, dishonest, and highly dishonest network environments, respectively.

[0126] The comparison results of task failure rate in a network environment with 10% malicious devices and 10% cooperative devices are shown in the figure below. Figure 7 shown. Figure 7 The results show an honest idle network environment in which the proportion of malicious devices is only 10%. The task failure rates of all four schemes are very low. These results reflect that these four network environments have few malicious nodes and have a certain degree of reliability.

[0127] The comparison results of task failure rate in a network environment with 20% malicious devices and 20% cooperative devices are shown in the figure below. Figure 8 shown. Figure 8 The proportion of malicious devices is 20%, and the proportion of collaborative devices is 20%. The results show that when the proportion of malicious devices is set to 20%, the trust mechanism proposed in this application has a lower task failure rate than TOSMFTM, CLMT, and RTCM in a dishonest and busy IoT computing environment.

[0128] The comparison results of task failure rate in a network environment with 40% malicious devices and 40% cooperative devices are shown in the figure below. Figure 9 shown. Figure 9 The proportion of malicious devices is 40%, and the proportion of cooperative devices is 40%, which shows that the system is highly busy and highly dishonest. Figure 9 From the perspective of mission failure rate, the trust mechanism proposed in this application is better than TOSMFTM, CLMT and RTCM.

[0129] It should be understood that although Figure 4The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 4 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0130] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0131] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A multi-source trust analysis method based on blockchain and IoT edge task collaboration, applied to edge proxy servers, characterized by: The method comprises: When receiving a collaboration request from device i, obtain the resource threshold T required for the current task of device i. ik ={δ i1 , δ i2 ,......,δ in }, i∈D, D={1, 2, 3...m}, where D is the device set, m is the number of devices in the device set, δ in Indicates the size of the nth capability resource required by the current task of device i; Get the resource vector N owned by each device j in the device set D corresponding to the current task jk ={ε j1 , ε j2 ,......,ε jn }, j∈D, j≠i, D={1, 2, 3...m}, where ε jn Indicates the size of the nth resource owned by device j for the current task; According to the resource threshold required by the current task of device i and the resource vector of each device j in the device set D corresponding to the current task, the matching degree of the current task capability of each device j in the device set D and device i is analyzed. According to the capability matching degree of each device j in the device set D at the current moment, the capability trust of each device j in the device set D at the current moment is analyzed. Among them, H(P ij ) is the capability matching degree P of the current task between device j and device i ij Information entropy of Obtain the direct trust between devices 1 to m in device set D and each device j in device set D; Calculate the weight of the feedback trust based on the number of interactions between devices 1 to m in device set D and each device j in device set D in, is the number of interactions between device i and device j in the current time window t; Analyze the indirect feedback trust of each device j in the device set D Among them, ω zj represents the weight of device z’s trust in device j’s feedback, is the direct trust between devices z and j in the current time window t; Combine the direct trust, capability trust and feedback trust to analyze the global trust value of each device j in the device set D The historical global trust value is updated according to the global trust value of each device j in the device set D, and the device with the highest global trust value is determined from the device set D to collaborate with the device i on the current task.

2. The method according to claim 1, characterized in that The global trust value of the device j for: in, is the global trust value of device j, α is the weight of device i’s direct trust in device j in the current time window t, β is the weight of edge proxy server k’s direct trust in device j in the current time window t, and γ is the weight of device i’s indirect feedback trust in device j in the current time window t. is the direct trust of device i on device j in the current time window t, is the direct trust of edge proxy server k on device j in the current time window t, is the indirect feedback trust of device i to device j in the current time window t.

3. The method according to claim 2, characterized in that The weight of the direct trust of device i on device j in the current time window t is determined as follows: According to the request frequency of device i to device j in the current time window t, the weight of direct trust of device i to device j in the current time window t is determined. The weight of direct trust of device i to device j in the current time window t is: α＝1-v n Where v is a hyperparameter ranging from 0 to 1, and n is the number of interactions between devices i and j in the current time window t.

4. The method according to claim 1, wherein The updating of the historical global trust value according to the global trust value of each device j in the device set D, and determining the device with the highest global trust value from the device set D to collaborate with the device i on the current task, includes: According to the global trust value of each device j in the device set D, the global trust value update formula is used to update the historical global trust value to determine the latest global trust value of each device j in the device set D; According to the latest global trust value of each device j in the device set D, a device with the highest global trust value is determined from the device set D to collaborate with the device i on the current task.

5. The method according to claim 4, characterized in that The global trust value update formula is: in, is the latest global trust value, α is the trust reward factor, and θ is the update adjustment factor.

6. The method according to claim 5, characterized in that The update adjustment factor θ is:

7. The method according to claim 1, characterized in that The matching degree P of the current task capability of device j and device i ij The information entropy is: H(P ij )=-P ij log2P ij -(1-P ij )log2(1-P ij ) Among them, H(P ij ) is the capability matching degree P of the current task between device j and device i ij Information entropy.

Images