Edge computing-oriented dynamic trust evaluation method and system

Abstract

The invention discloses an edge computing-oriented dynamic trust evaluation method. The method comprises the following steps of: selecting an edge node having historical interaction with a requesting edge node from a belonging edge node cluster as a to-be-evaluated edge node; evaluating a direct trust value between the requesting edge node and each to-be-evaluated edge node and a recommended trust value of each to-be-evaluated edge node; performing dynamic weighted aggregation on the direct trust value between the request edge node and the to-be-evaluated edge nodes and the recommended trust value of the to-be-evaluated edge nodes to obtain comprehensive trust values of the to-be-evaluated edge nodes and the request edge node, and selecting the to-be-evaluated edge node with the maximum comprehensive trust value as an interactive edge node; the trust degrees of the requesting edge node and the interaction edge node are updated based on the interaction result of the two. According to the dynamic trust data updating method based on the incentive mechanism, the trust data updating amplitude of the method is adaptively adjusted according to a trust calculation result so as to ensure that the updated trust data is closer to a true value.

Description

Technical Field

[0001] The present invention belongs to the field of privacy protection technology. More specifically, the present invention relates to a dynamic trust assessment method and system for edge computing. Background Art

[0002] With the advent of the big data era and the development of the Internet of Things and 5G networks, the popularity of terminal devices such as smartphones and IoT devices has generated massive amounts of data. Systems such as smart factories, smart homes, and smart transportation based on cloud computing technology are very sensitive to network delays, and traditional cloud computing centers have difficulty making decisions quickly from the perspective of computing power, bandwidth, and storage. In order to alleviate the load of cloud computing centers and make full use of edge node resources, the concept of edge computing is proposed to assign computing, storage, and other tasks to nodes close to objects or data sources to assist cloud computing centers in making quick decisions.

[0003] With the development of edge computing technology, the edge computing environment faces severe security issues. A large amount of user privacy data is stored on edge nodes, but edge computing lacks a complete protection mechanism. The frequent interaction of edge nodes increases the risk of privacy leakage. In addition, how to select interactive edge nodes to ensure higher interaction quality and true computing results is also a challenge now. Trust is a concentrated embodiment of security, privacy, and availability. Therefore, highly trusted nodes often also have high security, privacy, and availability. Trust evaluation can effectively motivate nodes to participate in interaction and establish trust relationships between edge nodes. It is one of the important solutions to alleviate the above security issues.

[0004] At present, there are few studies on trust evaluation methods in edge environments. Some scholars have proposed to directly migrate the trust evaluation methods in traditional cloud computing, P2P, and wireless sensor networks to edge computing environments. Faced with the contradiction between massive data scale and edge nodes with limited resources, traditional methods do not meet the lightweight characteristics required by edge computing environments and cannot be fully applied to edge computing environments. In addition, in the study of trust evaluation methods in edge environments, Wang et al. proposed a trust evaluation method based on spatiotemporal factors in edge environments, which takes into account the time factor of trust and the spatial geographical factors of edge nodes, and simultaneously evaluates the trust evaluation of both interacting nodes, thereby improving the reliability of trust evaluation. However, this method involves too few recommended nodes, resulting in overly subjective trust evaluation, and does not balance the influence of subjective and objective factors on the trust evaluation results, and performs poorly in resisting malicious node attacks. Summary of the invention

[0005] The present invention provides a dynamic trust evaluation method for edge computing, aiming to improve the above-mentioned problems.

[0006] The present invention is implemented as follows: a dynamic trust evaluation method for edge computing, the method comprising the following steps:

[0007] (1) Edge node ed r Initiate an interaction request at the edge node ed r Select the edge node ed from the edge node cluster r Other edge nodes with historical interactions are used as edge nodes to be evaluated. j ;

[0008] (2) Evaluation request edge node ed r With each edge node to be evaluated j The historical trust between the request edge nodes is evaluated using Newton's cooling law. r With each edge node to be evaluated j The direct trust value between the edge nodes to be evaluated based on the truth value discovery algorithm j Recommended trust value;

[0009] (3) Request edge node ed r and the edge node to be evaluated ed j The direct trust value between the edge nodes and the edge nodes to be evaluated j The recommended trust values ​​of each edge node to be evaluated are dynamically weighted and aggregated to obtain ed j With the request edge node ed r The comprehensive trust value of the edge node with the largest comprehensive trust value is selected as the requesting edge node ed i The interactive edge nodes t ;

[0010] (4) Based on the request edge node ed r Interact with edge nodes t The interaction results will update the edge node ed r For edge nodes t degree of trust.

[0011] Furthermore, edge nodes r For edge nodes j The direct trust value DT edr,edj The calculation formula is as follows:

[0012]

[0013] in, Indicates edge node ed r For edge nodes j The historical trust, t old Indicates historical trust The update time, t now represents the current time, and k is the attenuation coefficient.

[0014] Furthermore, each edge node to be evaluated ed j The specific process of obtaining the recommended trust value is as follows:

[0015] In edge node cluster D k Find the edge node to be evaluated in ed j Edge nodes with historical interactions i , calculate the edge node ed i For edge nodes j Direct trust value Direct trust value As the corresponding edge node ed i The perception data is taken as the recommended trust value output by the previous iteration as the true value, and the edge node ed is calculated based on the distance between the perception data and the true value. i The weight in the current iteration Then update the edge node ed j Recommended trust value Start the next iteration until the edge node ed is evaluated j The true value converges, and the converged true value is used as the corresponding evaluation edge node ed j The recommended trust value.

[0016] Furthermore, edge nodes i The weight of The calculation formula is as follows:

[0017]

[0018] in, Indicates edge node ed i For edge nodes j Direct trust value The recommended trust value output from the previous iteration The distance between The formula is as follows:

[0019]

[0020] in, Indicates the edge node to be evaluated ed j At each edge node i The standard deviation of the direct trust values ​​between .

[0021] Furthermore, edge nodes j Recommended trust value The calculation formula is as follows:

[0022]

[0023] in, For edge nodes ed i For edge nodes j The direct trust value of For edge nodes ed i The weight of .

[0024] Furthermore, edge nodes r For edge nodes j The comprehensive trust value The calculation formula is as follows:

[0025]

[0026] Among them, α represents the evaluation weight of the comprehensive trust value, Indicates edge node ed i For edge nodes j The direct trust value of Indicates edge node ed j The recommended trust value.

[0027] Furthermore, the calculation formula of the weight α is as follows:

[0028]

[0029] in, Indicates that the edge node ed is obtained in the last iteration i The perceptual data weight, Representation weight The standard deviation of is used to normalize the weight α using the maximum normalization method.

[0030] Further, in the request edge node ed r For interactive edge nodes t Historical trust When The calculation formula is as follows:

[0031]

[0032] in, Request edge node ed r Interact with edge nodes t Updated trust, Request edge node ed r Interact with edge nodes t The historical trust Request edge node ed r Interact with edge nodes t The comprehensive trust value of , θ is the incentive factor.

[0033] Furthermore, at the edge node ed r For interactive edge nodes t If this interaction is the first interaction between the two parties, if the interaction is successful, the edge node ed will be requested. r For interactive edge nodes t The comprehensive trust value of the requesting edge node ed r For interactive edge nodes t The degree of trust, that is If this interaction fails, the edge node ed will be requested r For interactive edge nodes t The comprehensive trust value of the requesting edge node ed r For interactive edge nodes t The historical trust of edge nodes ed r For interactive edge nodes t Trust

[0034] The present invention is implemented as follows: a dynamic trust evaluation system for edge computing, the system comprising:

[0035] Edge nodes,Edge nodes send interaction requests based on demand;

[0036] Multiple edge servers communicate with edge nodes in the corresponding edge node cluster. The edge servers receive interaction requests from edge nodes in the edge node cluster, find the interacting edge nodes of the requesting edge nodes among the edge nodes that have historical interactions with the requesting edge nodes based on the above-mentioned dynamic trust evaluation method for edge computing, and update the trust between the requesting edge nodes and the interacting edge nodes after the interaction is completed, and upload them to the cloud computing layer;

[0037] The cloud computing layer is used to store or update data uploaded by the edge server.

[0038] The present invention introduces Newton's cooling law to deal with the time decay problem of trust data, thereby improving the accuracy of direct trust calculation of candidate nodes. On this basis, the candidate node and all its historical interaction nodes are used as recommended nodes. Through a truth value discovery algorithm, based on the direct trust of all recommended nodes in the candidate node, the true value is iteratively calculated as the recommended trust, thereby reducing the influence of subjective factors on trust evaluation as much as possible. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1A flow chart of a dynamic trust assessment method for edge computing provided by an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of the structure of a dynamic trust assessment system for edge computing provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The specific implementation modes of the present invention are further explained in detail below by describing the embodiments with reference to the accompanying drawings, so as to help those skilled in the art to have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention.

[0042] Direct Trust Value Request edge node ed r Edge nodes to be evaluated j The quantification of the ability to complete collaborative tasks is based on the historical interaction records between the two, and the value range is [0,1].

[0043] Recommended Trust Value Edge nodes to be evaluated j Edge nodes with which it has historical interactions in the edge cluster to which it belongs, edge nodes ed j The quantification of the ability to complete collaborative tasks is based on the other nodes in the edge cluster to the node ed j The recommended weight is obtained, and the value range is [0,1].

[0044] Historical Trust Request edge node ed r Interact with edge nodes t The trust evaluation result at time t will be dynamically updated based on the interactive feedback. r With edge nodes ed t The trust value between them ranges from [0,1].

[0045] Figure 1 A flow chart of a dynamic trust assessment method for edge computing provided by an embodiment of the present invention, the method specifically comprises the following steps:

[0046] (1) Request edge node ed r Initiate an interaction request at the edge node ed r Select the edge node ed from the edge node cluster r Other edge nodes with historical interactions are used as edge nodes to be evaluated. j , put the edge node set D′ to be evaluated k ;

[0047] D k For edge servers kManaged edge node cluster, edge node cluster D k There are m edge nodes in it, edge node ed r ,ed j is the edge node cluster D k The edge device within the r ,ed j ∈D k , the edge node cluster D k Inner and request edge node ed r Edge nodes with historical interactions j Put the evaluation edge node set D′ k ,ed j ∈D′ k , j=1,2,...,n.

[0048] (2) Evaluation request edge node ed r With each edge node to be evaluated j The historical trust between the request edge nodes is evaluated using Newton's cooling law. r With each edge node to be evaluated j The direct trust value between them is then used to evaluate the edge node set D′ based on the truth value discovery algorithm. k Each edge node to be evaluated in ed j Recommended trust value;

[0049] In the embodiment of the present invention, Newton's law of cooling means that when the temperature of an object is higher than the ambient temperature, its own temperature gradually cools due to heat conduction. It defines the law of temperature change over time. Trust has dynamic properties and gradually approaches unfamiliarity, that is, a state of no trust, over time. The dynamic decay of trust has a similar process to cooling, which is specifically implemented as follows:

[0050]

[0051] in, represents the edge node ed at time t r For edge nodes j The historical trust of E represents the approaching value of environmental trust, and k is the attenuation coefficient. When t is the current time t now When , historical trust is equal to direct trust, that is,

[0052] By deducing from the above formula (1), we can know that the edge node ed r For edge nodes j Direct trust value The calculation formula is as follows:

[0053]

[0054] in, Indicates edge node ed r For edge nodes j The historical trust, t old Indicates historical trust The update time, t now Indicates the current time.

[0055] The above direct trust value The calculation process is as follows:

[0056]

[0057] Integrate both sides of equation (3):

[0058]

[0059] According to formula (4), formula (5) can be obtained:

[0060]

[0061] Where B is the factor for solving the differential equation. Transforming formula (5) yields formula (6):

[0062]

[0063] Further transformation yields formula (7):

[0064]

[0065] According to the initial conditions, represents the historical trust at time t0, and is substituted into the above formula (7) to obtain:

[0066]

[0067] According to formula (8), formula (9) can be obtained:

[0068]

[0069] When t is the current time t now When , the historical trust is direct trust, that is, Therefore, by substituting the above formulas (8) and (9) into formula (7), we can obtain:

[0070]

[0071] As time goes by, trust eventually approaches 0, that is, E = 0, which leads to formula (2).

[0072] The weights of direct trust values ​​are aggregated through the truth value discovery algorithm, and the recommended trust values ​​are iterated multiple times to obtain more accurate recommended trust data.

[0073] Indicates edge node ed r For edge nodes j The direct trust value of the data is used as the perception data in the truth discovery algorithm to evaluate the true value. Indicates edge node ed j The recommended trust value is the true value calculated by the truth value discovery algorithm as the recommended trust value. The truth value discovery algorithm is divided into two stages: weight update and truth value evaluation. The specific process is as follows:

[0074] In the weight update phase, in the edge node cluster D k Find the edge node to be evaluated ed j Edge nodes with historical interactions i , i=1,2,...,s, edge node ed i The weight of The calculation formula is as follows:

[0075]

[0076] in, Represents the perception data of each node The recommended trust value output from the previous iteration Since the trust data is continuous data, the distance function The formula is as follows:

[0077]

[0078] in, Indicates the edge node to be evaluated ed j At each edge node i The standard deviation of the direct trust values ​​between .

[0079] In the truth evaluation phase, the edge nodes ed are evaluated j Recommended trust value (true value) The calculation formula is as follows:

[0080]

[0081] Based on formula (13), the current round evaluation edge node ed is calculated j Recommended trust value Start the next iteration until the edge node ed is evaluated j The true value converges, and the converged true value is used as the corresponding evaluation edge node edj The recommended trust value.

[0082] By introducing Newton's cooling law to deal with the time decay problem of trust data, the accuracy of direct trust calculation of the candidate node is improved. On this basis, the candidate node and all its historical interaction nodes are used as recommended nodes. Through the truth value discovery algorithm, based on the direct trust of all recommended nodes in the candidate node, the true value is iteratively calculated as the recommended trust, which minimizes the impact of subjective factors on trust evaluation.

[0083] (3) Request edge node ed r and the edge node set D′ to be evaluated k Edge nodes to be evaluated j The direct trust value between them and the set of edge nodes to be evaluated D′ k Edge nodes to be evaluated j The recommendation trust values ​​are dynamically weighted and aggregated to obtain the edge node set D′ to be evaluated k The edge nodes to be evaluated in each edge ed j With the request edge node ed r The comprehensive trust value of the edge node with the largest comprehensive trust value is selected as the requesting edge node ed i The interactive edge nodes t ;

[0084] Direct trust value through aggregation and recommended trust value Calculate the comprehensive trust value Comprehensive trust is the final result of trust evaluation. The calculation formula of comprehensive trust value is as follows:

[0085]

[0086] Among them, α represents the evaluation weight of the comprehensive trust value, and the value of weight α changes dynamically. Indicates edge node ed i For edge nodes j The direct trust value of Indicates edge node ed j The recommended trust value, Indicates edge node ed r For edge nodes j The comprehensive trust value.

[0087] In the embodiment of the present invention, the method for determining the value of the weight α is specifically as follows:

[0088] According to the weight of the true value in the last iteration result of the recommendation trust evaluation process, the importance of the recommendation trust value in the comprehensive trust evaluation is evaluated. iPerception data weight Standard Deviation Evaluate the weight α of the recommended trust value in the comprehensive trust value. The calculation formula of weight α is as follows:

[0089]

[0090] in, Indicates that the edge node ed is obtained in the last iteration i The perceptual data weight, Representation weight The standard deviation of , and the maximum normalization method is used to normalize the weight α.

[0091] When the recommendation trust evaluation is performed through the truth value discovery algorithm, in the last iteration, after the weight update phase, the perception data weight of each recommendation node is obtained. This weight reflects the importance of each recommendation node's direct trust in the candidate node for the final true value calculation. If the discrete degree of the weight set is large, it reflects that the recommendation trust of the candidate node is more controversial. Conversely, the reliability of the recommendation trust of the candidate node is higher. Based on this, the weight of the recommendation trust value in the comprehensive trust value is evaluated through the standard deviation of the perception data weight of each recommendation node, which effectively balances the influence of subjective and objective factors on the trust evaluation results, and further improves the accuracy of the trust evaluation.

[0092] (4) Based on the request edge node ed r Interact with edge nodes t The interaction results will update the edge node ed r For edge nodes t degree of trust.

[0093] In the request edge node ed r Interact with edge nodes t After the interaction is completed, the edge node feedbacks the interaction results and dynamically updates the edge node based on the incentive mechanism. r For edge nodes t Due to different trust evaluation methods, trust updates are divided into request edge nodes ed r For interactive edge nodes t There are two situations: with historical trust and without historical trust. The update process of the new task degree is explained for the above two situations respectively, as follows:

[0094] (1) Request edge node ed r For interactive edge nodes t Has historical credibility When The calculation formula is as follows:

[0095]

[0096] in, Request edge node ed r Interact with edge nodes t Updated trust, Request edge node ed r Interact with edge nodes t Historical trust, Request edge node ed r Interact with edge nodes t The comprehensive trust value of , θ is the incentive factor.

[0097] (2) At the edge node ed r For interactive edge nodes t When this interaction is the first interaction between the two parties, the edge node ed r For interactive edge nodes t There is no historical trust. If the interaction is successful, the edge node ed will be requested. r For interactive edge nodes t The comprehensive trust value of the requesting edge node ed r For interactive edge nodes t The degree of trust, that is If this interaction fails, the edge node ed will be requested r For interactive edge nodes t The comprehensive trust value of the requesting edge node ed r For interactive edge nodes t The historical trust of the edge node ed is updated based on formula (16) and formula (17) r For interactive edge nodes t Trust

[0098] According to the interaction results, the trust is updated based on the incentive mechanism, and the updated trust participates in the next round of trust evaluation. The comprehensive trust obtained by the direct trust evaluation based on Newton's cooling law and the recommended trust based on the truth discovery algorithm is closer to the true reflection of trust. Therefore, the difference between the comprehensive trust and the historical trust is used as the trust update amplitude factor, which makes the trust update closer to the true value, greatly enhances the sustainability of the trust evaluation method, and greatly improves the ability to resist malicious node attacks.

[0099] Figure 2 A structural diagram of a dynamic trust assessment system for edge computing provided in an embodiment of the present invention. For ease of explanation, only the part related to the embodiment of the present invention is shown. The system includes:

[0100] Edge nodes are divided into different clusters according to their geographical location and functions. Each cluster is managed by an edge server. Edge nodes send interaction requests based on demand and request other edge nodes to interact collaboratively. Edge nodes are smartphones, cameras, vehicles, sensors, or industrial robots, etc.

[0101] Multiple edge servers communicate with edge nodes in the corresponding edge node cluster. The edge server receives interaction requests from edge nodes in the edge node cluster, finds the interacting edge nodes of the requesting edge node among the edge nodes that have historical interactions with the requesting edge node based on the above-mentioned dynamic trust evaluation method for edge computing, and updates the trust between the requesting edge node and the interacting edge node after the interaction is completed, and uploads it to the cloud computing layer.

[0102] The cloud computing layer is used to store or update data uploaded by the edge server.

[0103] In an edge computing environment, edge servers usually have a small amount of resources. Relying on the characteristics of the edge computing environment to manage edge nodes in clusters, the dynamic trust assessment method for edge computing has lower time complexity and space complexity, which meets the lightweight characteristics required by the edge computing environment.

[0104] The present invention has been described exemplarily. Obviously, the specific implementation of the present invention is not limited to the above-mentioned method. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention, or the concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A dynamic trust evaluation method for edge computing, characterized in that: The method comprises the following steps: (1) Edge node ed r Initiate an interaction request at the edge node ed r Select the edge node ed from the edge node cluster to which it belongs r Edge nodes with historical interactions are used as edge nodes to be evaluated. j ; (2) Evaluation request edge node ed r With each edge node to be evaluated ed j The historical trust between the request edge nodes is evaluated using Newton's cooling law. r With each edge node to be evaluated ed j The direct trust value between the edge nodes to be evaluated based on the truth value discovery algorithm j Recommended trust value; (3) Request edge node ed r and the edge node to be evaluated ed j The direct trust value between the edge nodes and the edge nodes to be evaluated j The recommended trust values ​​of each edge node to be evaluated are dynamically weighted and aggregated to obtain ed j With the request edge node ed r The comprehensive trust value of the edge node with the largest comprehensive trust value is selected as the interactive edge node. t ; (4) Based on the request edge node ed r Interact with edge nodes t The interaction results will update the edge node ed r For edge nodes t degree of trust.

2. The dynamic trust assessment method for edge computing as claimed in claim 1, characterized in that: Edge Node r For edge nodes j Direct trust value The calculation formula is as follows: in, Indicates edge node ed r For edge nodes j The historical trust, t old Indicates historical trust The update time, t now represents the current time, and k is the attenuation coefficient.

3. The dynamic trust assessment method for edge computing as claimed in claim 1, characterized in that: Each edge node to be evaluated j The specific process of obtaining the recommended trust value is as follows: In edge node cluster D k Find the edge node to be evaluated in ed j Edge nodes with historical interactions i , calculate the edge node ed i For edge nodes j Direct trust value Direct trust value As the corresponding edge node ed i The perception data of the previous iteration is used as the recommended trust value, and the edge node ed is calculated based on the distance between the perception data and the true value. i The weight in the current iteration Then update the edge node ed j Recommended trust value Start the next iteration until the edge node ed is evaluated j The true value converges, and the converged true value is used as the corresponding evaluation edge node ed j The recommended trust value.

4. The dynamic trust assessment method for edge computing as described in claim 3 is characterized in that: Edge Node i The weight of The calculation formula is as follows: in, Indicates edge node ed i For edge nodes j Direct trust value The recommended trust value output from the previous iteration The distance between The formula is as follows: in, Indicates the edge node to be evaluated ed j At each edge node i The standard deviation of the direct trust values ​​between .

5. The dynamic trust assessment method for edge computing as described in claim 3 is characterized in that: Edge Node j Recommended trust value The calculation formula is as follows: in, For edge nodes ed i For edge nodes j The direct trust value of For edge nodes ed i The weight of .

6. The dynamic trust assessment method for edge computing according to claim 1, characterized in that: Edge Node r For edge nodes j The comprehensive trust value The calculation formula is as follows: Among them, α represents the evaluation weight of the comprehensive trust value, Indicates edge node ed i For edge nodes j The direct trust value of Indicates edge node ed j The recommended trust value.

7. The dynamic trust assessment method for edge computing as described in claim 6 is characterized in that: The calculation formula of weight α is as follows: in, Indicates that the edge node ed is obtained in the last iteration i The perception data weight, W represents the weight The standard deviation of is used to normalize the weight α using the maximum normalization method.

8. The dynamic trust assessment method for edge computing according to claim 1, characterized in that: In the request edge node ed r For interactive edge nodes t Historical trust When The calculation formula is as follows: in, Request edge node ed r Interact with edge nodes t Updated trust, Request edge node ed r Interact with edge nodes t The historical trust Request edge node ed r Interact with edge nodes t The comprehensive trust value of , θ is the incentive factor.

9. The dynamic trust assessment method for edge computing as described in claim 8 is characterized in that: At the edge node r For interactive edge nodes t If this interaction is the first interaction between the two parties, if the interaction is successful, the edge node ed will be requested. r For interactive edge nodes t The comprehensive trust value of the requesting edge node ed r For interactive edge nodes t The degree of trust, that is If this interaction fails, the edge node ed will be requested r For interactive edge nodes t The comprehensive trust value of the requesting edge node ed r For interactive edge nodes t The historical trust of edge nodes ed r For interactive edge nodes t Trust 10. A dynamic trust assessment system for edge computing, characterized in that: The system comprises: Edge nodes,Edge nodes send interaction requests based on demand; Multiple edge servers communicate with edge nodes in the corresponding edge node cluster, and the edge servers receive interaction requests from edge nodes in the edge node cluster, and find the interacting edge nodes of the requesting edge nodes among the edge nodes that have historical interactions with the requesting edge nodes based on the dynamic trust evaluation method for edge computing described in any one of claims 1 to 9, and update the trust between the requesting edge nodes and the interacting edge nodes after the interaction is completed, and upload them to the cloud computing layer; The cloud computing layer is used to store or update data uploaded by the edge server.

Images