Method for constructing a collaborative device group to ensure device reliability and reduce communication losses in the Internet of Things

Abstract

The present invention relates to a method for constructing a collaborative device group to ensure device reliability and reduce communication losses in the Internet of Things, belonging to the field of the Internet of Things. The method includes: S1: constructing a trust evaluation scheme for devices in the social Internet of Things: first, constructing identity-based trust and cognitive-based trust; then, comprehensively combining cognitive-based trust and identity-based trust by using the method of fuzzy inference; and finally, defuzzifying according to the centroid formula to obtain the final trust value; S2: constructing a social collaboration group with minimized total cost for task delegation in the social Internet of Things scenario. The present invention can maximize the reliability of task delegation, and at the same time reduce the transmission costs in wireless and wired networks and the communication and computing costs in MEC servers and SIoT devices.

Description

Technical Field

[0001] The present invention belongs to the field of Internet of Things, and relates to a method for constructing a collaborative device group to ensure device reliability and reduce communication loss in the Internet of Things. Background Art

[0002] With the rapid development and large-scale deployment of intelligent devices, in order to promote the development of intelligent applications and infrastructure, the Internet of Things (IoT) has been widely applied worldwide. To improve the efficiency of information and service discovery, researchers have integrated the concept of social relationships into IoT, giving rise to a new paradigm called Social Internet of Things (SIoT). In SIoT, intelligent devices can autonomously establish social relationships for different forms of collaboration and achieve local distributed data processing according to rules set by the owner. At the same time, the concept of Mobile Edge Computing (MEC) is to offload computing and traffic to the edge network to relieve the network load. The MEC server can reduce the data transmitted to the remote server by preprocessing and aggregating local SIoT data, and the collaboration between SIoT devices can process user requests locally and further reduce communication and computing overhead in the MEC network. Therefore, the collaboration between SIoT and MEC is crucial for reducing network communication and computing overhead. In addition, integrating social attributes into IoT may bring trust issues. In social networks, there are risks due to the asymmetry of information and time between the principal and the trustee during transactions. When a Service Requester (SR) publishes a service or task, multiple Service Providers (SPs) may respond and execute the task, and the SR must distinguish which SP is most suitable for providing such services. Therefore, in order to improve service quality, an efficient trust evaluation model should be established in SIoT, aiming to achieve dynamic real-time evaluation of the service capabilities of SPs.

[0003] In recent years, there have been many works on MEC data aggregation and trust evaluation in SIoT. Zhao et al. proposed a new architecture to aggregate IoT data in wireless networks by assigning MEC servers as IoT gateways. The architecture consists of three modules: First, the IoT network is divided into multiple subnets by a traffic-aware discretization method, and the traffic diversity is carefully examined. Second, the link quality and link correlation are evaluated by the proposed utility metrics. Based on the utility metrics, a deployment algorithm is proposed to determine the location of MEC servers. Shan et al. considered the initial energy of sensors and constructed a shortest-path aggregation tree to maximize the network lifetime. To improve scalability, a distributed protocol is proposed for constructing an aggregation tree based on a semi-matching algorithm. Chen et al. explored the aggregation scheduling with minimum delay in energy harvesting sensor (IoT) networks. They identified a new type of collision, called energy collision, and constructed an energy-collision-aware aggregation tree based on the remaining battery of each IoT to minimize the aggregation delay. They observed that the charging time is the main factor related to the aggregation delay, rather than the collision during the transmission process. However, the above works on SIoT have not explored the trade-off between social relationships and physical space, nor have they solved the communication and computing trade-off problems in SIoT and MEC networks.

[0004] The work on SIoT trust models has also received extensive attention. Nitti et al. proposed subjective and objective models for credibility evaluation. In the subjective model, a node will evaluate another node by combining its own and the experiences from other neighbor nodes. In the objective model, a distributed hash table is used for trust-related information dissemination, ensuring the consistency of information. Chen et al. proposed an adaptive trust model by referring to the trust model in social networks and combining direct and indirect credibility evaluations. They proposed social similarity metrics, including friendship similarity, social contact similarity, and interest community similarity as elements of indirect trust to maximize the application performance. Adaptive filtering is used to control the weight parameters to prevent malicious attacks. Based on this work, they further proposed metrics for honesty, cooperation, and community interests as indicators of social trust, and used air pollution detection and enhanced map travel assistance as typical IoT scenarios to verify the effectiveness of these metrics. Lin and Dong et al. considered the mutual trust between objects. In their trust model, the principal and the trustee will conduct bilateral evaluations on each other. In addition, real-world social networks, including Facebook, Google+, and Twitter, are used to simulate SIoT.

[0005] Although the above work has studied the SIoT and MEC data aggregation from multiple perspectives, there is still a lack of in-depth research on the following aspects in the construction of social collaboration groups in the Internet of Things: First, the communication and computing trade-off problem. SIoT devices with good social centrality can collaborate with more SIoT devices and effectively reduce communication losses, but the central SIoT device requires more computing resources to process local data. Second, the trade-off between social relationships and spatial locations. Adjacent SIoT devices in space may not be directly connected in the social network, and the collaboration of a group of SIoT devices that are close in the social network but far apart in space requires additional multi-hop transmissions. Third, the forwarding and aggregation trade-off. Although the MEC server can aggregate the data streams of multiple SIoT devices to reduce the size of the total data stream, when the shortest path from the SIoT device to the remote server does not include the MEC server and the path through the MEC server is longer, additional bandwidth consumption will occur. Fourth, most existing methods quantify the degree of trust based on the fixed characteristics of Internet of Things objects (such as their location and ownership), but the relationships between objects are variable during service interactions. Secondly, for the functions of the SP and the attributes of specific tasks, the capabilities and interests of the SP in providing services tend to be different. Fifth, the security issues in the real SIoT environment caused by malicious behaviors remain unresolved.

[0006] Therefore, there is an urgent need for a method for constructing a social collaboration device group in the Internet of Things to solve the trade-off problems of various costs in the system when the social Internet of Things system constructs a collaboration device group in the face of task requests and to ensure the reliable delivery of tasks. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a method for constructing a collaboration device group in the Internet of Things to ensure device reliability and reduce communication losses, so as to maximize the reliability of task delegation while reducing the transmission costs in wireless and wired networks and the communication and computing costs in the MEC server and SIoT devices.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for constructing a collaboration device group in the Internet of Things to ensure device reliability and reduce communication losses, specifically including the following steps:

[0010] S1: Construct a trust evaluation scheme for devices in the social Internet of Things: First, construct identification-based trust and cognition-based trust; then, use the method of fuzzy inference to synthesize cognition-based trust and identification-based trust, divide identification-based trust and cognition-based trust into different levels, and set relevant rules for fuzzy inference according to the principles of sociology and psychology, and infer the comprehensive trust level according to the fuzzy logic rules; finally, defuzzify according to the centroid formula to obtain the final trust value;

[0011] S2: Construct a social collaboration group that minimizes the total cost for task delegation in the social Internet of Things scenario: First, find the path with the lowest total cost for each Internet of Things device. When facing multiple requests, each request can be decomposed into a set of multiple tasks. A collaboration group needs to be constructed for each request to ensure that each task in the request is completed by a device. Since the tasks in multiple requests may have intersections, these tasks can be reused multiple times as long as they are delivered once. Screen the multiple collaboration groups constructed above. Specifically, define the unit task cost as the ratio of the cost of the collaboration group of the device to complete the task and the types of tasks it can complete in the request. Continuously find the device or collaboration group with the smallest unit task cost and add it to the final solution set. After iterating multiple times to ensure that all tasks are covered by devices, eliminate redundant devices or collaboration groups. Thus, the construction of the collaboration group is completed.

[0012] Furthermore, in step S1, construct identity-based trust, which specifically includes the following steps:

[0013] S101: Calculate the degree of identity of the service requester SR;

[0014] The quantification of SR's degree of identity with SP can be measured based on the degree of task completion of SP in historical transaction records. The degree of task completion is mainly measured from whether the task is completed and the quality of the completed task for successfully delivered tasks. The quality of task completion has different specific requirements according to different task requests. Transaction information that is too old in historical transaction records is not very likely to reflect the current relevant information of the device's task completion. A decay function can be set to set the weight ratio of transaction records with different freshness levels;

[0015] S102: Calculate the temporal interest degree of the service provider SP;

[0016] Divide the system time into multiple equal-sized time slices, and respectively count the frequency ratios of a device processing a certain specific task in the long time range and the short time range to the frequency of all task types as the long and short interest degrees. Comprehensively obtain the temporal interest preference for a certain period as the temporal interest degree of the device SP;

[0017] S103: According to the identity degree of SR and the temporal interest degree of SP the identity-based trust evaluation value can be obtained;

[0018] For task τ, device v j the identity-based trust is:

[0019]

[0020] where μ 1 and μ 2is the weighting coefficient, and μ 1 + μ 2 = 1.

[0021] Further, in step S101, calculating the degree of recognition of SR specifically includes: setting as the attenuation function of successful delivery, as the attenuation function of delivery failure; in order to maximize the guarantee that the task can be successfully delivered, the impact of the historical record of task delivery failure is greater, that is, when it fails once, it needs to be offset by multiple consecutive successful deliveries. The attenuation function is expressed as follows:

[0022]

[0023]

[0024] k f ≥ k s , th f ≥ th s

[0025] where t 0 represents the current time, represents the time when device v j successfully completed task τ in the historical record, represents the time when device v j failed to complete task τ in the historical record, th s represents the threshold of the difference between the recorded time of successfully completing the task and the current time, th f represents the threshold of the difference between the recorded time of failing to complete the task and the current time, k s and k f are constants and have no practical meaning;

[0026] Set to represent the quality evaluation function of device v j completing task τ, which is expressed as follows:

[0027]

[0028] where represents the time of requesting task τ , represents the maximum time difference between the requested task and the completed task in the historical results;

[0029] Define the degree of task completion as:

[0030]

[0031] where R τ∈{0, 1} indicates whether the task is successfully delivered. A successful delivery is 1, otherwise it is 0;

[0032] Therefore, device v i Regarding device v j Degree of recognition is:

[0033]

[0034] where TS represents the set of task types.

[0035] Furthermore, in step S102, calculate the temporal interest degree of SP, specifically including: Define the long-term interest degree LI j,τ,l indicating the degree of interest of a certain type of IoT device in a task within a long time range. The long-term interest degree LI j,τ,l can be expressed as:

[0036]

[0037] where l represents the serial number of the time slice where it is located, indicating the total number of times IoT device v L processes tasks within the long time range ΔT j , and indicating the total number of times device v L processes task τ within the long time range ΔT j . T represents the time period;

[0038] Define the short-term interest degree SI j,τ,l describing the interest of the device in a certain type of computing task within a short time range. The short-term interest degree SI j,τ,l can be expressed as:

[0039]

[0040] where l represents the serial number of the time slice where it is located, indicating the total number of times the device processes tasks within the short time range ΔT S , and indicating the total number of times device v S processes task τ within the short time range ΔT j ;

[0041] Taking into account both the long-term interest accumulation and the short-term interest change, define the temporal preference ρ j,τ,l describing the temporal interest of IoT device v j in task τ within the time slice. ρ j,τ,l can be expressed as:

[0042] ρ j,τ,l , = γ l ·LIj,τ,l +γ s ·SI j,τ,l

[0043] Among them, γ l and γ s are influence coefficients and satisfy γ l +γ s = 1;

[0044] Construct the temporal preference matrix of IoT devices for the task type set TS on the time slice, denoted as:

[0045] P j,TS,l = {ρ j,1,l , ρ j,2,l , …, ρ j,N,l}

[0046] Among them, N represents the number of task types;

[0047] The temporal interest degree is the concentrated embodiment of the interest change of IoT devices in different time slices; Set the temporal perception weight ω to represent the influence of the temporal preference matrix P j,TS,l on the temporal interest degree, and construct the temporal perception weight matrix ω T in the time period T as:

[0048]

[0049] Among them, L represents the number of time slices in the time period T;

[0050] Integrate the temporal preference matrix P j,TS,l of IoT devices in each time slice l , and the temporal perception weight ω , the temporal interest degree of IoT devices for the task type set TS

[0051]

[0052] Record the temporal preference values of IoT devices for different types of tasks in the time period in matrix form.

[0053] Furthermore, in step S1, construct cognitive trust, which specifically includes the following steps:

[0054] S111: According to the attributes of the task, complete the statistics of the matching strength of all corresponding characteristic attributes of the device SP to obtain the quantization value of the device's ability to complete a certain task type;

[0055] S112: Calculate the recommendation coefficient, defined as the comprehensive value of the social relationship strength and the similarity of task delegation;

[0056] S113: Calculate the cognitive trust $K_{BT}$ of device $j$ for task $\tau$ j For

[0057]

[0058] where $RT$ represents the set of reliable recommendations, represents the recommendation coefficient, and $E$ represents the recommendation coefficient matrix.

[0059] Furthermore, step S111 specifically includes: The feature - attribute matching strength can be expressed as:

[0060]

[0061] where represents the total number of tasks completed by device $v$ j with function while completing tasks containing attribute , represents the total number of tasks completed by device $v$ j while completing tasks containing attribute , $TS$ represents the set of task types, represents the matching set between the features of device $v$ j and the attributes of task $\tau$;

[0062] Define the ability of device $v$ j to complete task $\tau$ as:

[0063]

[0064] where represents the feature set of device $v$ j , and $A$ τ is the attribute set of task $\tau$; That is, if is less than the threshold $th$ m , it means that function is not helpful for completing task attribute .

[0065] Furthermore, step S112 specifically includes: The social - relationship strength $Do_{SR}$ between the recommended device $i$ and $SR$ i,SR is expressed as follows:

[0066]

[0067] where $NT$ represents the number of transactions between the recommended device and $SR$, and $\alpha$ and $\beta$ represent the initial $Do_{SR}$ of two social relationships existing between the recommended device and $SR$, and $0\leq\beta\lt\alpha\lt1$. As the number of transactions increases, the social - relationship strength will also increase slowly;

[0068] Correlation Sim of timing preferences among devices i,SR is defined as follows:

[0069]

[0070]

[0071] Among them, represents the timing preference for device i to delegate task τ, represents the average value of the timing preferences of device i for all tasks, represents the timing preference for the requester SR to delegate task τ, represents the average value of the timing preferences of the requester SR for all tasks, IS i,SR represents setting Sim i,SR between 0 and 1, and also represents the correlation of timing preferences;

[0072] Define the recommendation coefficient as:

[0073]

[0074] Among them, η 1 、η 2 represent the weighting coefficients;

[0075] The recommended value r i,j is expressed as follows:

[0076]

[0077] Among them, ρ 1 、ρ 2 represent the weighting coefficients respectively;

[0078] The similarity RS between device suggestions is expressed as follows:

[0079]

[0080] Among them, represents the average value of the suggestions of the receiving device;

[0081] Define the reliable advice set RT and the recommendation coefficient set E as:

[0082]

[0083]

[0084] Among them, r t represents the advice value of the recommending device i; If that is, the advice value given by the recommender deviates from the reference advice value by more than r t, and believes that this suggestion is not advisable.

[0085] Furthermore, in step S2, for each request, the process of constructing a collaboration group is as follows: for the data returned by each device, calculate the costs of the paths with MEC server participation in aggregation and the paths without MEC server participation in aggregation, including transmission cost, computing cost, and forwarding cost; select the path with the minimum cost among the paths without aggregation and with aggregation as the path for the device to return data.

[0086] For each device in the Internet of Things, set the uncertainty of task completion as the vertex weight according to its trust level, set the communication cost between devices with social relationships as the edge weight, set multiple task groups according to the task set of the request, and each group only contains devices that can complete the corresponding tasks. Use the GST (Group Steiner Tree) method to minimize the total weight and construct a collaboration group.

[0087] For multiple task requests, multiple collaboration groups will be constructed accordingly. However, when the requested data is returned to the remote server, some of the task data can be reused, and correspondingly some devices and collaboration groups are redundant.

[0088] Furthermore, in step S2, the cost of the path without aggregation is defined as:

[0089]

[0090] where c n represents the cost of device n transmitting data, f n represents the data stream size returned by device n, represents the cost of forwarding a unit of data, represents the cost of the remote server processing a unit of data, represents the shortest path from the base station to the remote server;

[0091] The cost of the path with aggregation is defined as:

[0092]

[0093] where represents the cost of the MEC server processing a unit of data, represents the shortest path from the MEC server to the remote server, represents the shortest path from the base station to the MEC server, and β' represents the ratio of the MEC server aggregating data.

[0094] Further, in step S2, the specific process of GST is as follows: Definition: The smallest group is the group with the smallest number of devices in the task group, the unconnected group is the other device groups in the task group except the smallest group, an initially empty connected device group, and an empty subtree;

[0095] S1: Find the minimum weight paths from all devices to the unconnected device groups;

[0096] S2: Initialize a device in the smallest group as a connected device in the connected device group (i.e., the initialized connected device group has only one device), and add the minimum weight paths from this device to each group in the unconnected device group set to a priority queue with the weight as the priority;

[0097] S3: Continuously pop the minimum weight path at the top of the priority queue, add the popped minimum weight path to the empty tree, add all the devices in the minimum weight path to the connected device group, and remove the device group corresponding to the minimum weight path from the unconnected device group;

[0098] S4: Record the minimum weight path in the priority queue. For the devices newly added to the connected queue, if the minimum weight path from them to the unconnected device group is less than the recorded path, update the recorded path to the new minimum weight path;

[0099] S5: Repeat steps S3 - S4 until the unconnected device group is empty, and the obtained subtree is the feasible solution;

[0100] Iteratively execute steps S2 - S5 for all devices in the smallest group, and the subtree with the smallest total weight is the solution of the cooperation group.

[0101] The beneficial effects of the present invention are as follows: By constructing a trust evaluation scheme for devices in the social Internet of Things and selecting a social cooperation group with the minimum total cost for task delegation, the present invention can maximize the reliability of task delegation, while reducing the transmission costs in wireless and wired networks and the communication and computing costs in MEC servers and SIoT devices.

[0102] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent description, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be learned from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following description. Description of the Drawings

[0103] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail preferably with reference to the accompanying drawings, where:

[0104] Figure 1It is a structural framework diagram of a social Internet of Things system;

[0105] Figure 2 It is a flowchart of the trust evaluation of the present invention. Specific implementation manners

[0106] The following uses specific specific examples to illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0107] Please refer to Figures 1 to 2 , the present invention provides a method for constructing a collaborative device group to ensure device reliability and reduce communication losses in the Internet of Things, which mainly consists of two parts:

[0108] 1) A trust evaluation scheme suitable for devices in the social Internet of Things.

[0109] Construct a trust evaluation scheme, which specifically includes:

[0110] Based on the principles of sociology and psychology, trust can be divided into identification-based trust and cognition-based trust. Specifically, for a service requester SR and a service provider SP, when there is a transaction history record between SR and SP, SR can form an identification degree with respect to SP, and at the same time, SP can form its own interest degree for a certain type of task in the past period. When there is a lack of interaction history between SR and SP, SR needs to refer to the suggestions from other people and the capabilities of SP itself to evaluate the trust of SP. Therefore, the present invention evaluates the credibility of devices in the SIoT by referring to the trust composition in the social trust theory and combining IBT and KBT, that is, the four components of SR's identification degree, SP's interest degree, SP's capabilities, and the suggestions of other devices.

[0111] When interacting between different devices, relevant transaction information will be saved locally. The transaction information includes relevant device information, whether the task is successful, the delivery time, and other information related to the task quality. The quantification of SR's recognition degree of SP can be measured based on the degree of task completion by SP in historical transaction records. The degree of task completion is mainly measured from whether the task is completed and the quality of the completed task for successfully delivered tasks. The quality of task completion has different specific requirements according to different task requests. Transaction information that is too old in the historical transaction record is less likely to reflect the relevant information of the device's current task completion. The weight ratio of transaction records with different freshness can be set by setting an attenuation function.

[0112] Set as the attenuation function for successful delivery, and

[0113]

[0114]

[0115] k f ≥k s ,th f ≥th s

[0116] where, t 0 represents the current time, represents the time when device v j successfully completed task τ in the historical record, represents the time when device v j failed to complete task τ in the historical record, th s represents the threshold of the time difference between the record time of successfully completing the task and the current time, th f represents the threshold of the time difference between the record time of failing to complete the task and the current time, k s and k f are constants and have no practical meaning.

[0117] Set to represent the quality evaluation function of device v j for completing task τ

[0118]

[0119] where, represents the time of requesting task τ, represents the maximum time difference between the requested task and the completed task in the historical result.

[0120] Define the task completion degree as follows:

[0121]

[0122] Among them, R τ ∈ {0, 1} represents whether the task is successfully delivered. If it is successfully delivered, it is 1; otherwise, it is 0.

[0123] Therefore, the degree of recognition of device v i for device v j is as follows: as follows:

[0124]

[0125] During the operation of Internet of Things devices, due to different usage scenarios, device purposes, and usage habits, Internet of Things devices will have a relatively high processing frequency for specific types of tasks, and similar to human usage habits, generate the interest of Internet of Things devices in a certain type of task. The interest of Internet of Things devices will change dynamically according to their operating status. In order to more accurately perceive the willingness of devices to complete tasks, by analyzing historical data of task processing, establish the temporal interest degree of Internet of Things devices for task types, and perceive idle devices with similar temporal interest changes to provide services for Internet of Things devices. In a long time range, the interest of Internet of Things devices in different types of tasks will accumulate due to the number of times of their task processing, thus generating a relatively stable long-term interest. In a short time range, the device will exhibit an interest drift phenomenon, thus changing its interest in different types of tasks. Divide the system time into multiple equal-sized time slices, and respectively count the frequency of a device processing a certain specific task in the long time range and the short time range as a proportion of the frequency of all task types as the long-term and short-term interest degrees, and comprehensively obtain the temporal interest preference for a certain period as the temporal interest degree of device SP.

[0126] Define the long-term interest degree LI j,τ,l to represent the degree of interest of Internet of Things devices in a certain type of task in a long time range. The long-term interest degree LI j,τ,l can be expressed as:

[0127]

[0128] Among them, l represents the serial number of the time slice where it is located, represents the total number of tasks processed by Internet of Things device v L within the long time range ΔT j , represents the total number of tasks τ processed by device v L within the long time range ΔT j .

[0129] Define the short-term interest SI j,τ,l Describe the interest of the device in a certain type of computing task within a short time range. The short-term interest SI j,τ,l Can be expressed as:

[0130]

[0131] Where l represents the serial number of the time slice where it is located, Represents the total number of times the device processes tasks within the short time range ΔT S Represents the total number of times the device v S Processes task τ within the short time range ΔT j

[0132] Taking into account the long-term interest accumulation and short-term interest changes, define the temporal preference ρ j,τ,l Describe the temporal interest of the IoT device v j In the time slice for task τ, ρ j,τ,l Can be expressed as:

[0133] ρ j,τ,l , = γ l ·LI j,τ,l + γ s ·SI j,τ,l

[0134] Where γ l And γ s Are influence coefficients and satisfy γ l + γ s = 1.

[0135] Construct the temporal preference matrix P of the IoT device for the task type set TS on the time slice j,TS,l , Expressed as:

[0136] P j,TS,l = {ρ j,1,l , ρ j,2,l , …, ρ j,28,l}

[0137] The temporal interest degree is a concentrated reflection of the interest change of the IoT device in different time slices. Set the temporal perception weight ω to represent the influence of the temporal preference matrix P j,TS,l In different time slices on the temporal interest degree, construct the temporal perception weight matrix ω within the time period T T

[0138]

[0139] Integrate the temporal preference matrix P of the IoT device within each time slice j,TS,l , And the temporal perception weight ω l ​​, the temporal interest degree of the IoT device in the set of task types TS can be expressed as:

[0140]

[0141] Record the temporal preference values of the IoT device for different types of tasks in matrix form.

[0142] Based on the degree of recognition of SR and the temporal interest degree of SP, the recognition-based trust evaluation value can be obtained. For device v for task τ j the recognition-based trust is

[0143]

[0144] where, μ 1 and μ 2 are weighting coefficients, and μ 1 +μ 2 = 1.

[0145] The quantification of the device's capabilities can be used to evaluate whether the device has sufficient communication, computing, and storage resources and capabilities to complete the task. As mentioned before, the business scenarios and task types in SIoT have a wide range of diversity. Therefore, directly evaluating the task performance before SP does not necessarily evaluate its comprehensive ability to complete the task. In addition, even for the same type of task, different requirements may be needed in different environments. Therefore, it becomes particularly important to evaluate whether the function of SP is suitable for the specific attributes of the task and to determine the degree to which this function determines the success or failure of the task. Based on historical interactions, establish the matching relationship between different device characteristics and different task attributes. Define the frequency of occurrence of a certain function of the device in successfully delivered tasks with specific attributes as the feature-attribute matching strength, so as to describe the importance of each feature to the attribute. The greater the feature-attribute matching strength, the more important the function is to the specific attribute. According to the attributes of the task, after statistically completing all the corresponding feature-attribute matching strengths of the device SP, the quantification value of the device's ability to complete a certain type of task is obtained.

[0146] The feature-attribute matching strength can be expressed as:

[0147]

[0148] where, represents the total number of times that device v j completes the task containing attribute in the case of having function , represents the total number of times that device v j completes the task containing attribute .

[0149] Define device v j The ability to complete task τ For

[0150]

[0151] Among them, That is, if Less than the threshold th m , it means that this function is not helpful for completing the task attribute Not helpful

[0152] Trust calculation should consider suggestions based on direct interaction from third parties, especially in the case where there is no historical interaction between SR and SP. To overcome the subjectivity of trust calculation, even with direct interaction, the trust mechanism should also consider the suggestions of others. Generally speaking, the recommended trust of a specific device can be calculated based on the recommendation information from other devices with social relationships. However, not all suggestions are trustworthy. Some malicious devices may give unreliable suggestion values and need to be further weighed

[0153] There may be social relationships between IoT devices. The interaction between devices with social relationships is more frequent than that of other devices, and the sense of identity generated is also higher. Therefore, those devices with social relationships with SR have greater influence, and SR is more likely to seek suggestions from these devices. To perceive the impact of the social relationship between devices on the recommendation value, the strength of the social relationship (Deep of Social Relationship, DoSR) can be used to measure the strength of the relationship between objects. At the same time, those devices that are more similar to SR may have closer task requirements to SR, and the similarity between them can be evaluated based on the temporal interest degree of the delegated task between SR and the recommender

[0154] First, it should be judged whether the time when the recommended device gives suggestions exceeds the time limit. If it exceeds the time limit, it is outdated data. In addition to SR and SP, those devices that have had interaction history with SP can all be used as recommended devices. The strength of the social relationship between the recommended device and SR can be obtained based on the social relationship between devices, and the similarity between SR and the recommended device regarding task delegation can be obtained based on the temporal preference of the delegated task between device SR and the recommended device. The comprehensive value of the strength of the social relationship and the similarity of task delegation is defined as the recommendation coefficient. The device with a higher recommendation coefficient is set as a trusted device. The suggestions of other recommended devices are compared with the suggestions of reliable devices. Those with a large gap are considered malicious information and the malicious information is removed. At the same time, based on the recommendation coefficient, all recommendation values that do not contain malicious information are statistically calculated as the value of the cognitive trust of device SP

[0155] The strength of the social relationship between the recommended device i and SR is expressed as follows: DoSR i,SR

[0156]

[0157] Among them, NT represents the number of transactions between the recommended device and the SR, and α and β represent the initial DoSR of two social relationships existing between the recommended device and the SR, and 0 ≤ β < α < 1. As the number of transactions increases, the strength of the social relationship will also increase slowly.

[0158] The correlation of the timing preference between devices is defined as follows:

[0159]

[0160]

[0161] Define the recommendation coefficient as:

[0162] η 1 + η 2 = 1

[0163] The recommended value is expressed as follows

[0164]

[0165] The similarity RS between device suggestions can be expressed as follows:

[0166]

[0167] Define the reliable advice set RT and the recommendation coefficient set E:

[0168]

[0169]

[0170] If that is, the recommended value given by the recommender deviates from the reference recommended value by more than r t , it is considered that this advice is not advisable.

[0171] Therefore, the cognitive trust of device j for task τ is

[0172]

[0173] Since trust is a concept with ambiguity and uncertainty, a fuzzy inference method is adopted to comprehensively recognize cognitive trust and identification trust. Cognitive trust and identification trust are divided into four levels: low, medium, high, and very high. According to the principles of sociology and psychology, relevant rules for fuzzy inference are set. Generally speaking, identification trust comes from the direct interaction between both parties, and its influence in comprehensive trust is greater. The comprehensive trust level is deduced according to the fuzzy logic rules, which is also divided into four levels: low, medium, high, and very high. Finally, the final trust value is obtained by defuzzification according to the centroid formula.

[0174] 2) A social collaboration group selection method suitable for minimizing the total cost for task delegation in the social Internet of Things scenario.

[0175] Construct a social collaboration group, which specifically includes:

[0176] The total cost from the delegated task to the successful delivery of the task in the SIoT includes the transmission cost of the Internet of Things devices, the computing cost of the MEC server and the remote server, and the forwarding cost of the switch. The selection strategy of the present invention minimizes the total cost by establishing an effective social collaboration group and selecting a suitable MEC server to aggregate data. First, find the path with the lowest total cost for each Internet of Things device, and then find a group of Internet of Things devices connected in the social layer to construct a collaboration group for the delegated task. For each request, the process of constructing the collaboration group is as follows:

[0177] For the data returned by each device, calculate the costs of the path with MEC server participation in aggregation and the path without MEC server participation in aggregation, including transmission cost, computing cost, and forwarding cost. For the path without aggregation, only find the shortest path between the base station corresponding to the device and the remote server. For the path with aggregation, since the transmission and forwarding costs are greatly related to the size of the data stream, the data stream becomes smaller after aggregation, so the transmission and forwarding costs after aggregation are reduced. Select the path with the lowest cost among the paths without aggregation and with aggregation as the path for the device to return data.

[0178] The cost of the path without aggregation is defined as

[0179]

[0180] The cost of the path with aggregation is defined as

[0181]

[0182] For each device in the Internet of Things, set the uncertainty of task completion as the vertex weight according to its trust level, set the communication cost between devices with social relationships as the edge weight, set multiple task groups according to the set of requested tasks, and each group only contains devices that can complete the corresponding tasks. Use the GST method to minimize the total weight and construct a collaboration group.

[0183] The specific process of GST is as follows:

[0184] Definition: The smallest group is the group with the smallest number of devices in the task group, the unconnected group is the group of other devices in the task group except the smallest group, an initially empty connected device group, and an empty subtree.

[0185] S1: First, find the minimum-weight paths from all devices to the unconnected device groups.

[0186] S2: Initialize a device in the smallest group as the connected device group (i.e., the initialized connected device group has only one device), and add the minimum-weight paths from this device to each group in the unconnected device group set to a priority queue with the weight as the priority.

[0187] S3: Continuously pop the minimum-weight path at the top of the priority queue, add the popped minimum-weight path to the empty tree, add all the devices in the minimum-weight path to the connected device group, and remove the device group corresponding to the minimum-weight path from the unconnected device group.

[0188] S4: Record the minimum minimum-weight path in the priority queue. For the devices newly added to the connected queue, if the minimum-weight path from them to the unconnected device group is less than the recorded path, update the recorded path to the new minimum minimum-weight path.

[0189] S5: Repeat steps S3 - S4 until the unconnected device group is empty. The obtained subtree is the feasible solution.

[0190] Iteratively execute steps S2 - S5 for all devices in the smallest group. The subtree with the minimum total weight is the solution of the collaboration group.

[0191] For multiple task requests, multiple collaboration groups will be constructed accordingly. However, when the requested data is returned to the remote server, some of the task data can be reused, and some of the devices and collaboration groups are redundant. Therefore, define the unit task cost as the ratio of the cost of the collaboration group of the device to complete the task and the types of tasks it can complete in the request. Continuously find the device or collaboration group with the minimum unit task cost and add it to the final solution set. After iterating multiple times to ensure that all tasks are covered by the devices, remove the redundant devices or collaboration groups. Thus, the construction of the collaboration group is completed.

[0192] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for constructing a collaborative device group to ensure device reliability and reduce communication losses in the Internet of Things, characterized in that, the method specifically includes the following steps: S1: Construct a trust evaluation scheme for devices in the social Internet of Things: First, construct identification-based trust and cognition-based trust; then, use the method of fuzzy inference to synthesize cognition-based trust and identification-based trust, divide identification-based trust and cognition-based trust into different levels, and set relevant rules for fuzzy inference according to the principles of sociology and psychology, and infer the comprehensive trust level according to the fuzzy logic rules; finally, defuzzify according to the centroid formula to obtain the final trust value; S2: Construct a social collaboration group that minimizes the total cost for task delegation in the social Internet of Things scenario: First, find the path with the lowest total cost for each Internet of Things device. When facing multiple requests, each request can be decomposed into a set of multiple tasks. A collaboration group needs to be constructed for each request to ensure that each task in the request is completed by a device; since the tasks in multiple requests may have intersections, these tasks can be reused multiple times as long as they are delivered once. Screen the multiple collaboration groups constructed above. Specifically, define the unit task cost as the ratio of the cost of the collaboration group of the device to complete the task and the types of tasks it can complete in the request. Continuously find the device or collaboration group with the smallest unit task cost to join the final solution set. After iterating multiple times to ensure that all tasks are covered by devices, eliminate redundant devices or collaboration groups. Thus, the construction of the collaboration group is completed; For each request, the process of constructing a collaboration group is as follows: For the data returned by each device, calculate the costs of the path with MEC server participation in aggregation and the path without MEC server participation in aggregation, including transmission cost, computing cost, and forwarding cost; select the path with the minimum cost among the paths without aggregation and with aggregation as the path for the device to return data; the cost of the path without aggregation is defined as: Among them, c n represents the cost of device n transmitting data, f n represents the data stream size returned by device n, represents the cost of forwarding a unit of data, represents the cost of the remote server processing a unit of data, represents the shortest path from the base station to the remote server; Cost of Aggregation Path Is defined as: Among them, represents the cost of the MEC server for processing unit data, represents the shortest path from the MEC server to the remote server, represents the shortest path from the base station to the MEC server, and β' represents the ratio of the MEC server for aggregating data; For each device in the Internet of Things, set the uncertainty of its task completion as the vertex weight according to its trust level, set the communication cost between devices with social relationships as the edge weight, set multiple task groups according to the task set of the request, and each group only contains devices that can complete the corresponding task. Use the GST method to minimize the total weight and construct a collaboration group; where GST represents Group Steiner Tree; For multiple task requests, multiple collaboration groups will be constructed accordingly. However, when the request data is returned to the remote server, some corresponding devices and collaboration groups are redundant.

2. The method for constructing a collaborative device group according to claim 1, characterized in that, in step S1, constructing identification-based trust specifically includes the following steps: S101: Calculate the identification degree of the service requester SR; The quantification of the identification degree of SR for SP is measured based on the degree of task completion of SP in the historical transaction record; the degree of task completion is measured from whether the task is completed and the quality of the completed task for the successfully delivered task; the quality of the completed task has different specific requirements according to different task requests; in the same historical transaction record, set a decay function to set the weight ratio of transaction records with different freshness; S102: Calculate the temporal interest degree of the service provider SP; Divide the system time into multiple equal-sized time slices, and respectively count the frequency ratio of a device processing a certain specific task in the long-term range and the short-term range to the frequency of all task types as the long-term and short-term interest degrees, and comprehensively obtain the temporal interest preference for a certain period as the temporal interest degree of the device SP; S103: According to the degree of recognition of SR and the temporal interest degree of SP obtain the recognition-based trust evaluation value; For task τ, device v j The identification-based trust is as follows: where, μ 1 and μ 2 are weighting factors, and μ 1 + μ 2 = 1.

3. The method for constructing a collaborative device group according to claim 2, characterized in that, In step S101, calculate the degree of recognition of SR, which specifically includes: setting as the attenuation function for successful delivery, as the attenuation function for delivery failure; the attenuation functions are expressed as follows: k f ≥ k s ,th f ≥ th s Among them, t 0 represents the current time, represents the time when device v j successfully completed task τ in the historical record, represents the time when device v j failed to complete task τ in the historical record, th s represents the threshold of the difference between the recorded time of successfully completing the task and the current time, th f represents the threshold of the difference between the recorded time of failing to complete the task and the current time, k s and k f is a constant and has no practical significance; Settings Representative device v j The quality evaluation function for completing task τ is expressed as follows: Among them, represents the time of requesting task τ, represents the maximum time difference between the requested task and the completed task in the historical results; Define the task completion degree as follows: Among them, R τ ∈ {0, 1} indicates whether the task is successfully delivered. A successful delivery is 1, otherwise it is 0; Therefore, device v i Regarding device v j Degree of recognition is as follows: where TS represents the set of task types.

4. The method for constructing a collaborative device group according to claim 3, characterized in that, In step S102, calculate the temporal interest degree of SP, specifically including: defining the long-term interest degree LI j,τ,l represents the interest degree of a certain type of IoT device in a task within a long time range, and the long-term interest degree LI j,τ,l is expressed as: Among them, l represents the serial number of the time slice where it is located. represents the total number of times the Internet of Things device v L processes tasks within the long time range ΔT j , and represents the total number of times the device v L processes the task τ within the long time range ΔT j . T represents the time period. Define the short-term interest SI j,τ,l Describe the device's interest in a certain type of computing task within a short time range. The short-term interest SI j,τ,l Is expressed as: Among them, represents the total number of times the device processes tasks within the short time range ΔT S ; represents the total number of times device v S processes task τ within the short time range ΔT j . Define the temporal preference ρ by comprehensively considering long-term interest accumulation and short-term interest changes j,τ,l Describe the Internet of Things device v j The temporal interest in task τ within a time slice, ρ j,τ,l Denoted as: ρ j,τ,l = γ l ·LI j,τ,l + γ s ·SI j,τ,l Among them, γ l and γ s are influence coefficients and satisfy γ l +γ s = 1; Construct the temporal preference matrix P of the IoT device for the task type set TS on the time slice j,TS,l , expressed as: P j,TS,l = {ρ j,1,l , ρ j,2,l ,..., ρ j,N,l} where N represents the number of task types; The temporal interest degree is a concentrated reflection of the interest changes of IoT devices in different time slices; set the temporal perception weight ω to represent the temporal preference matrix P in different time slices j,TS,l , and construct the temporal perception weight matrix ω within the time period T for its impact on the temporal interest degree T as follows: where L represents the number of time slices within the time period T; Integrate the temporal preference matrix P of IoT devices within each time slice j,TS,l , with the temporal perception weight ω l , the temporal interest degree of the IoT device in the task type set TS is expressed as: Record the timing preference values of IoT devices for different types of tasks during a period in matrix form.

5. The method for constructing a collaborative device group according to claim 1, characterized in that, in step S1, constructing a cognitive trust specifically includes the following steps: S111: According to the attributes of the task, the matching strength of all corresponding characteristic attributes of the device SP is statistically completed to obtain the quantization value of the device's ability to complete a certain task type; S112: Calculate the recommendation coefficient, defined as the comprehensive value of the social relationship strength and the similarity of task delegation; S113: Calculate the cognitive trust KBT of device j for task τ j For Among them, RT represents the set of reliable suggestions, ∈ i represents the recommendation coefficient, and E represents the recommendation coefficient matrix.

6. The method for constructing a collaborative device group according to claim 5, characterized in that, step S111 specifically includes: The matching strength of characteristic attributes is expressed as: Among them, represents device v j completes the total number of tasks including attributes in the case of having the function ; represents device v j completes the total number of tasks including attributes ; TS represents the set of task types, represents the matching set between the characteristics of device v j and the attributes of task τ; Define device v j The ability to complete task τ is as follows: Among them, represents the feature set of device v j and A τ is the attribute set of task τ; That is, if is less than the threshold th m , it means that the function is not helpful for completing the task attribute .

7. The method for constructing a collaborative device group according to claim 6, characterized in that, Step S112 specifically includes: recommending the social relationship strength DoSR between device i and SR i,SR It is expressed as follows: where NT represents the number of transactions between the recommended device and SR, α and β represent the initial DoSR of two social relationships existing between the recommended device and SR, and 0 ≤ β < α < 1. As the number of transactions increases, the social relationship strength will also increase slowly; Correlation Sim of timing preferences between devices i,SR is defined as follows: Among them, represents the timing preference for device i to delegate task τ, represents the average value of the timing preferences for all tasks of device i, represents the timing preference for requester SR to delegate task τ, represents the average value of the timing preferences for all tasks of requester SR, IS i,SR represents setting Sim i,SR between 0 and 1, and also represents the correlation of timing preferences; Define the recommendation coefficient ∈ i as: ∈ i = η 1 DoSR i,SR + η 2 IS i,SR ,η 1 + η 2 = 1 Among them, η 1 and η 2 represent weighting coefficients; Recommended value r i,j It is expressed as follows: Among them, ρ 1 and ρ 2 represent weighted coefficients respectively; The similarity RS between device suggestions is expressed as follows: Among them, represents the average value recommended by the receiving device; Define the reliable suggestion set RT and the recommendation coefficient set E as: Among them, r t represents the recommended value for the recommended device i; if that is, the recommended value given by the recommender deviates from the reference recommended value by more than r t , it is considered that this recommendation is not advisable.

8. The method for constructing a collaborative device group according to claim 1, characterized in that, in step S2, the specific process of GST is as follows: Define: The smallest group is the group with the smallest number of devices in the task group, the unconnected group is the other device groups in the task group except the smallest group, an initially empty connected device group, and an empty subtree; S1: Find the minimum weight path from all devices to the unconnected device groups; S2: Initialize a device in the smallest group as the connected device group, and add the minimum weight path from this device to each group in the unconnected device group set to a priority queue with the weight as the priority; S3: Continuously pop the minimum weight path at the top of the priority queue, add the popped minimum weight path to the empty tree, add all the devices in the minimum weight path to the connected device group, and remove the device group corresponding to the minimum weight path from the unconnected device group; S4: Record the minimum minimum weight path in the priority queue. For the devices newly added to the connected queue, if the minimum weight path from them to the unconnected device group is less than the recorded path, update the recorded path to the new minimum minimum weight path; S5: Repeat steps S3 - S4 until the unconnected device group is empty. The obtained subtree is the feasible solution; Iteratively execute steps S2 - S5 for all devices in the smallest group. The subtree with the smallest total weight is the solution of the collaborative group.

Images