Method and device for MEC task offloading with joint resource allocation and privacy protection

Abstract

The present application discloses a method and device for MEC task offloading with joint resource allocation and privacy protection, and relates to the field of computer technology. The method includes: obtaining information of tasks to be processed; constructing a communication model, a computing model and a joint privacy entropy model, and establishing an optimization model for the association between task offloading ratio, network selection, resource allocation, task processing time and joint privacy entropy; solving the optimization model using a double-layer offloading algorithm, and obtaining corresponding strategies; selecting and allocating tasks to different offloading locations according to the strategies, and selecting corresponding network transmission data. The present application combines resource allocation and privacy protection in the task offloading process, while taking into account multi-network data transmission, which can further reduce the terminal task processing time and ensure the privacy security of task offloading. At the same time, the present application adopts a hierarchical mode to construct an offloading model and design an offloading algorithm, which reduces the complexity of the joint task offloading problem.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method and device for MEC task offloading with joint resource allocation and privacy protection. Background Art

[0002] MEC, the full name of which is Multi-Access Edge Computing, is a new computing method that can provide users with three-layer resources: cloud server, edge server, and terminal. Multi-access edge computing can increase the available resources of the terminal and improve the efficiency of task processing by offloading computing tasks from the terminal to the edge server and cloud server. However, users need to make accurate decisions when deciding whether to execute tasks locally on the terminal or offload them to the edge server or cloud server. Incorrect offloading decisions will not only fail to reduce the pressure on the terminal, but will also lead to more consumption due to additional data transmission.

[0003] In the prior art, there is a problem of coarse granularity in the task offloading process, which will leave the entire task to be executed locally on the terminal, or offload it to the edge server or cloud server at one time. At the same time, the prior art does not fully consider the resource allocation and privacy protection issues in the task offloading process, and cannot meet the efficiency of resource allocation in the task offloading process, and cannot guarantee the security of the offloading process. In addition, the data transmission process in the task offloading process of the prior art is single, which cannot meet the transmission between multiple servers.

[0004] Therefore, there is an urgent need to provide a new multi-access edge computing task offloading method to solve the above-mentioned problems existing in the prior art. Summary of the invention

[0005] The embodiments of the present application provide a method and device for MEC task offloading with joint resource allocation and privacy protection, thereby solving the problems of long task processing time and low privacy and security performance due to the limitations and defects of related technologies in the prior art.

[0006] In the first aspect, an embodiment of the present application provides a method for MEC task offloading with joint resource allocation and privacy protection, including: obtaining information about tasks to be processed, the information about tasks to be processed including the amount of input data of the tasks to be processed, the number of CPU cycles required to process the tasks to be processed, the level of leakage information of the tasks to be processed, the maximum tolerance time of the tasks to be processed, and the maximum energy consumption of the tasks to be processed; establishing an associated optimization model through a communication model, a computing model, and a joint privacy entropy model; solving the associated optimization model to obtain a task offloading strategy, a network selection strategy, and a resource allocation strategy; allocating the unloading locations of the tasks to be processed in proportion according to the task offloading strategy, selecting the corresponding network to transmit data based on the network selection strategy, and allocating the computing resources required for the tasks to be processed at the unloading locations through the resource allocation strategy.

[0007] In combination with the first aspect, in a possible implementation manner, the unloading location includes a terminal, an edge server, and a cloud server.

[0008] In combination with the first possible implementation manner of the first aspect, in a second possible implementation manner, the communication model supports the use of multiple wireless networks to simultaneously perform data transmission between the terminal and the edge server, and data transmission between the edge server and the cloud server.

[0009] In combination with the first aspect, in a third possible implementation method, the joint privacy entropy model includes task privacy entropy and data privacy entropy; the task privacy entropy is used to obtain the leakage probability and leakage level of the task to be processed at different unloading locations; the data privacy entropy is used to obtain the leakage probability and leakage level of the data of the task to be processed during the data transmission process.

[0010] In combination with the first aspect, in a fourth possible implementation method, the associated optimization model is established through the communication model, the computing model and the joint privacy entropy model, including: constructing the communication model, the computing model and the joint privacy entropy model; obtaining the network selection of the task to be processed based on the communication model, obtaining the task unloading ratio, resource allocation ratio, task processing time and minimized task processing time of the task to be processed at the unloading location through the computing model, and obtaining the joint privacy entropy and maximized joint privacy entropy of the task to be processed according to the joint privacy entropy model; integrating the task unloading ratio, the network selection, the resource allocation ratio and the task processing time, and establishing an associated optimization model of the task processing time and the joint privacy entropy; the associated optimization model includes an objective function and multiple constraints.

[0011] In combination with the fourth possible implementation manner of the first aspect, in a fifth possible implementation manner, the minimization of the task processing time and the maximization of the joint privacy entropy are used as the optimization objectives of the objective function, and the optimization objectives are normalized; the normalized optimization objectives are transformed using a linear weighted method to obtain a function value: Among them, U n represents the normalized linear weighted value of the optimization target, N represents the number of tasks to be processed, X represents the strategy after the task offloading strategy and the network selection strategy of the tasks to be processed are coupled, X=[α n ,β n,k ,γ n,k ] n×(2k+1) , where α n Indicates the proportion of tasks to be processed executed on the terminal, β n,k represents the proportion of the pending tasks executed on the edge server, γ n,k represents the proportion of the pending tasks executed in the cloud server, k represents the kth network, F is the resource allocation strategy for the pending tasks, in, represents the computing resources allocated when the pending task is processed at the terminal, represents the computing resources allocated when the pending task is processed on the edge server, represents the computing resources allocated when the pending task is processed on the cloud server, l represents that the pending task is executed on the terminal, es represents that the pending task is executed on the edge server, and c represents that the pending task is executed on the cloud server.

[0012] In combination with the first aspect, in a sixth possible implementation method, the solving the associated optimization model includes: solving the associated optimization model based on a two-layer unloading algorithm of an improved dung beetle optimizer; the two-layer unloading algorithm is divided into two layers, the upper layer of the two-layer unloading algorithm is based on the improved dung beetle optimizer to solve the optimal task offloading strategy and the optimal network selection strategy of the associated optimization model; the lower layer of the two-layer unloading algorithm is based on the improved dung beetle optimizer to solve the optimal resource allocation strategy of the associated optimization model.

[0013] In combination with the sixth possible implementation manner of the first aspect, in a seventh possible implementation manner, the double-layer unloading algorithm based on the improved dung beetle optimizer solves the associated optimization model, including: using Bernoulli chaos mapping to initialize the population of the double-layer unloading algorithm: Wherein, ψ is the control parameter, z1 is the mapping output, and z0 represents the initialization generation of the chaotic mapping; the dung beetle population is initialized, an initial position is set for each dung beetle, and it is expanded to the D-dimensional space. In the boundary range, the initialization of the dung beetle population after adding the chaotic mapping is expressed as: D0=Lb+z1(Ub-Lb); wherein, D0 represents the position of the dung beetle population initialization, Ub represents the upper bound of the problem to be solved, and Lb represents the lower bound of the problem to be solved; an adaptive parameter adjustment strategy is used to dynamically adjust the number of producers during the operation of the unloading algorithm; when the adaptive parameter adjustment strategy updates the number of producers, the producer ratio will be reduced, and the reduction expression of the producer ratio is: Among them, P percent represents the producer ratio, P u represents the upper bound of the producer ratio, P l Iter represents the lower bound of the producer ratio, iter represents the current number of iterations, and Iter max Indicates the maximum number of iterations.

[0014] In a second aspect, an embodiment of the present application provides an apparatus for a method of performing MEC task offloading with joint resource allocation and privacy protection, the apparatus comprising: a processor; a memory for storing processor executable instructions; when the processor executes the executable instructions, it implements the method described in the first aspect or any possible implementation method of the first aspect.

[0015] In a third aspect, an embodiment of the present application provides a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium includes a medium for storing a computer program or instructions, which, when executed, enables the method described in the first aspect or any possible implementation method of the first aspect to be implemented.

[0016] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0017] This application transforms the multi-access edge computing task offloading problem into an associated optimization model based on a communication model, a computing model, and a joint privacy entropy model, thereby solving the problem that the offloading decisions made during the multi-access edge computing task offloading process are insufficient to meet high task processing efficiency, high resource allocation efficiency, high local terminal processing pressure, and high energy consumption during additional data transmission from the perspective of resource allocation and privacy protection.

[0018] At the same time, the present application obtains the optimal solution for the task offloading strategy, network selection strategy and resource allocation strategy by solving the associated optimization model. According to the above-mentioned task offloading strategy, the tasks are allocated to the corresponding offloading locations in proportion. According to the network selection strategy, the corresponding network is selected to transmit data. Finally, the computing resources are allocated to the corresponding offloading location through the resource allocation strategy. The resource allocation problem and the privacy protection problem in the task offloading process are solved. The efficiency and security of resource allocation in the task offloading process are achieved. At the same time, the present application can realize multi-terminal transmission between edge servers and other terminals, reduce the occupancy rate of network transmission, and improve the transmission speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 A schematic diagram of an application scenario of the method for joint resource allocation and privacy-protected MEC task offloading provided in an embodiment of the present application;

[0021] Figure 2 A flowchart of a method for joint resource allocation and privacy protection for MEC task offloading provided in an embodiment of the present application;

[0022] Figure 3 A schematic diagram of a two-layer unloading algorithm flow based on an improved dung beetle optimizer provided in an embodiment of the present application;

[0023] Figure 4 A schematic diagram of target values ​​for different average leakage numbers (λ) and different numbers of tasks using the unloading method of the present application provided in an embodiment of the present application;

[0024] Figure 5 A schematic diagram of target values ​​for different numbers of networks (k) using the offloading method of the present application provided in an embodiment of the present application;

[0025] Figure 6 A schematic diagram of target values ​​under different weight factors (δ) and different numbers of tasks using the unloading method of the present application provided in an embodiment of the present application;

[0026] Figure 7 A schematic diagram of the performance comparison of the unloading method of the present application and other technologies at different iteration times provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] Figure 1 Schematic diagram of the application scenario of the method for joint resource allocation and privacy protection of MEC task offloading provided in the embodiment of the present application. As shown in the figure, the multi-access edge computing system includes a terminal layer, an edge layer, and a cloud layer. The terminal and the edge server can be connected through a variety of wireless links, wired links, etc.

[0029] Exemplarily, the terminal may be an industrial device such as an intelligent monitoring device or an intelligent robot, and the wireless network may be a wireless wide area network, a wireless local area network, a wireless metropolitan area network, a wireless personal area network, etc. The edge layer includes multiple edge servers. The cloud layer includes multiple cloud servers.

[0030] The following Figure 1 Taking the application scenario shown as an example, the MEC task offloading method for joint resource allocation and privacy protection disclosed in this application is described in detail.

[0031] Figure 2 is a flow chart of a method for MEC task offloading with joint resource allocation and privacy protection in an embodiment of the present application, such as Figure 2 As shown, the flow chart of the MEC task offloading method includes the following steps:

[0032] Step 101: Obtain information of the task to be processed. The information of the task to be processed includes the amount of input data of the task to be processed, the number of CPU cycles required to process the task to be processed, the level of leakage information of the task to be processed, the maximum tolerance time of the task to be processed, and the maximum energy consumption of the task to be processed.

[0033] In the embodiment of the present application, the information of the task to be processed is shown in the following formula.

[0034]

[0035] In the formula, S n Indicates pending tasks, D n is the amount of input data for the task to be processed, C n The number of CPU cycles required to process the pending tasks, y n is the level of leakage information of the task to be processed, is the maximum tolerable time to complete pending tasks, The maximum energy required to complete the task to be processed.

[0036] Step 102: Establish an association optimization model through the communication model, the calculation model and the joint privacy entropy model.

[0037] In an embodiment of the present application, the communication model can support the use of multiple wireless networks to simultaneously transmit data between various unloading locations, and the unloading locations include terminals, edge servers, and cloud servers.

[0038] Therefore, the communication model supports the use of multiple wireless networks for simultaneous data transmission between terminals and edge servers, and between edge servers and cloud servers.

[0039] Exemplarily, the construction process of the communication model is as follows:

[0040] In this application, terminal n can use multiple networks to transmit data at the same time, where n represents the number of the terminal, n = 1, 2, 3, ..., N. The signal-to-noise ratio SNR (English name: Signal to Noise Ratio) reaching the edge server through network k (k represents the number and / or number of assignable networks) is expressed as:

[0041]

[0042] Among them, SNR n,k represents the signal-to-noise ratio of terminal n to the edge server through network k, p n,k Select the transmission power of network k for the task to be processed, h n,k represents the channel gain of terminal n selecting network k, p n,q Select the transmission power of network q for the task to be processed, h n,q Indicates the channel gain of terminal n selecting network q. Δ 2 It is additive white Gaussian noise (AWGN, full name in English: Additive White Gaussian Noise).

[0043] When the task to be processed is offloaded to the edge server, its data uplink transmission rate is expressed as:

[0044]

[0045] in, represents the data uplink rate of terminal n through network k, B represents the channel bandwidth; k≤K, K represents the maximum communication resource of network k that can be allocated on the edge server; w n,k represents the network k allocation indicator matrix, when w n,k = 1, the kth network resource in the edge server is allocated to the user (i.e., terminal n), when w n,k When =0, it means no allocation.

[0046] When the pending tasks are offloaded to the cloud server, they are first transmitted to the edge server and then to the cloud server for offloading. The uplink rate of the pending tasks in the edge server to the cloud server is expressed as:

[0047]

[0048] Among them, R n represents the uplink rate of data transmitted from the edge server to the cloud server through the task to be processed, B C is the bandwidth from edge server to cloud server, P n,mec is the transmission power from the edge server to the cloud server for sending tasks to be processed, q n represents the channel gain from the edge server to the cloud server; σ 2 Represents the noise power between the edge server and the cloud server.

[0049] Exemplarily, the process of establishing the calculation model is as follows.

[0050] Use vector X = [α n ,β n,k ,γ n,k ] n×(2k+1) represents the strategy after coupling the task offloading strategy and network selection strategy of the task to be processed, where α n It indicates the proportion of pending tasks processed locally on the terminal after being offloaded, β n,k represents the proportion of pending tasks that are offloaded to edge servers through network k, γ n,k It indicates the proportion of pending tasks that are offloaded to the cloud server through network k.

[0051] Using Vectors represents the resource allocation strategy for tasks to be processed, where represents the computing resources allocated when the pending tasks are processed locally on terminal n. represents the computing resources allocated when the pending tasks are offloaded to the edge server through network k, Indicates the computing resources allocated when the pending tasks are offloaded to the cloud server for processing.

[0052] It should be noted that the processing time of the task to be processed at each offloading location (terminal, edge server and cloud server) needs to be calculated through the calculation model. Specifically:

[0053] The processing time of the pending task on terminal n is:

[0054]

[0055] in, Indicates the processing time of the pending task on terminal n.

[0056] The processing time of the pending task offloaded to the edge server through the network k is:

[0057]

[0058] in, It represents the processing time of the pending tasks offloaded to the edge server through network k.

[0059] The processing time of the pending tasks offloaded to the cloud server through the network k is:

[0060]

[0061] in, It represents the processing time of the pending tasks offloaded to the cloud server through network k.

[0062] Exemplarily, the above calculation model-based calculation of the processing time of the task to be processed at each unloading location can also obtain the minimized task processing time.

[0063] The embodiment of the present application also provides a joint privacy entropy model, which is as follows.

[0064] Different entropy values ​​are set for each task, and the privacy security of the task is measured by establishing a joint privacy entropy model. The joint privacy entropy model includes task privacy entropy and data privacy entropy. The construction of the joint privacy entropy model is shown in the following formula.

[0065] H(S n y n )=H(S n )+H(y n ).

[0066] Among them, H(S n y n ) represents the joint privacy entropy, H(S n ) represents data privacy entropy, H(y n ) represents the task privacy entropy.

[0067] Task privacy entropy is used to obtain the leakage probability and leakage level of the pending task at different offloading locations.

[0068] Data privacy entropy is used to obtain the data leakage probability and leakage level of the task to be processed during the data transmission process.

[0069] Based on the information entropy theory, considering that the terminal also has the possibility of leaking information, sufficient location randomness should be maintained to avoid any location privacy leakage. The degree of randomness of the location can be measured by the entropy value, using the discrete variable y n It represents the level of leakage of the tasks to be processed at different locations. The relationship between the discrete variable set Y and the task set S is expressed as:

[0070]

[0071] Among them, y1,y2,…,y n Represents each sample in the discrete variable set Y, S1, S2, ..., S n Represents each sample in the task set S.

[0072] Within the coverage of the edge server, the sent task data to be processed is divided into M categories, where M represents the number of categories of tasks to be processed at different locations. represents the probability of leakage of the task to be processed at the mth position, and the value range of m is [1, M]. Therefore, the possibility of leakage at the corresponding position of each task to be processed is expressed as:

[0073]

[0074] Among them, the probability of location leakage prob n,m Satisfies [0,1] and Therefore, the expression of the task privacy entropy of the task to be processed is:

[0075]

[0076] Among them, H(y n ) represents the task privacy entropy.

[0077] Data privacy entropy is an indicator used to measure the amount of information that may be leaked during the transmission of a task. The relationship between the optimal offloading strategy X (i.e., the strategy after coupling the task offloading strategy and the network selection) and the task set S is expressed as:

[0078]

[0079] Among them, x1,x2,…,x n Represents each sample in the optimal offloading strategy X.

[0080] The data of the pending task to be processed is divided into G categories, where G represents the number of data categories of the pending task during the transmission process. represents the probability of the g-th data leakage of the task to be processed, the value range of g is [1, G], λ is the average number of leakages during data transmission, e is the natural base, λ gThe probability weight representing the average number of leaks in the case of the g-th type of data leak.

[0081] Therefore, the number of pending tasks and prob n,g The correlation can be expressed as:

[0082]

[0083] Among them, prob n,g Satisfies [0,1] and Therefore, the data privacy entropy value is defined as H = H(S1), H(S2), ..., H(S n ), then the data privacy entropy of the task to be processed is expressed as:

[0084]

[0085] Within the coverage of the edge server, there will be different probabilities of information leakage when users choose different specific locations, which will affect the privacy risks of ultimately choosing different networks for data transmission.

[0086] The embodiment of the present application comprehensively considers that the unloading location of the task to be processed and the selection of the data transmission network have a dual impact on privacy protection. For this association optimization model, the joint privacy entropy of each task to be processed is expressed as:

[0087]

[0088] The correlation between the offloading strategy X and the joint privacy entropy value H is expressed as:

[0089]

[0090] For example, based on the above formula, the maximum joint privacy entropy H in the joint privacy entropy model can be obtained: max , where the maximum joint privacy entropy H max The calculation process is as follows:

[0091] H max =max{H(S1y1),H(S2y2),...,H(S n y n )}.

[0092] To summarize, the embodiment of the present application obtains the network selection of the task to be processed based on the communication model, obtains the task unloading ratio, resource allocation ratio, task processing time and minimized task processing time of the task to be processed at different unloading locations through the calculation model, and obtains the joint privacy entropy of the task to be processed and the maximized joint privacy entropy according to the joint privacy entropy model.

[0093] Afterwards, an associated optimization model among task offloading ratio, network selection, resource allocation ratio, task processing time, and joint privacy entropy is established as shown below.

[0094] The maximum tolerable time for a given pending task The time for the task to be processed to be calculated locally on the terminal and offloaded to the edge server and cloud server through multi-network transmission cannot exceed Normalized time utility function for pending tasks It is expressed as:

[0095]

[0096] in, represents the standardized time utility function of the task to be processed, t n Indicates the time to process pending tasks.

[0097] Use H max Represents the maximum joint privacy entropy of the joint privacy entropy model and the standardized privacy utility function of the task to be processed for:

[0098]

[0099] in, Represents the normalized privacy utility function of the task to be processed.

[0100] Linearly weight the time utility function and the joint privacy entropy utility, and use the linear weighted value U normalized by the processing time of the task to be processed and the joint privacy entropy n To describe the overall utility of the tasks to be processed in the entire system, it is expressed as:

[0101]

[0102] Where δ is the weight factor between time and joint privacy entropy.

[0103] In the embodiment of the present application, the association optimization model consists of an objective function and eleven constraints. The objective function takes minimizing task processing time and maximizing privacy entropy as optimization goals, normalizes these two goals, and converts them into a function value using a linear weighted method, expressed as:

[0104]

[0105] Among them, U n represents the processing time of the task to be processed and the normalized linear weighted value of the joint privacy entropy, N represents the number of tasks, X is the strategy after the task offloading strategy and the network selection strategy are coupled, X = [α n ,β n,k,γ n,k ] n×(2k+1) , element α n Indicates the proportion of pending tasks left to be executed locally on the terminal, and the element β n,k represents the proportion of pending tasks offloaded to edge servers through network k, and the element γ n,k represents the proportion of pending tasks offloaded to the cloud server through network k for execution, in, Indicates the resources allocated when pending tasks are processed locally on the terminal. represents the resources allocated when the pending tasks are offloaded to the edge server through network k, Indicates the resources allocated when pending tasks are offloaded to the cloud server for processing.

[0106] For example, the eleven constraints are:

[0107] (C1) This constraint ensures that pending tasks are processed within their maximum tolerable time. Completed within.

[0108] (C2) This constraint limits the energy consumption of the task to be processed when it is executed locally on the terminal to no more than

[0109] in, Indicates the energy consumption of the pending task when it is executed on the terminal. represents the energy consumption when the pending tasks are offloaded from the terminal to the edge server, Indicates the maximum energy consumption of the terminal when executing locally.

[0110] (C3) This constraint limits the energy consumption of the task to be processed when it is executed on the edge server to no more than

[0111] in, Represents the energy consumption of the pending task when it is executed on the edge server. Represents the energy consumption when the tasks to be processed are offloaded from the edge server to the cloud server.

[0112] (C4) This constraint is the network selection indicator matrix. If the unloading is performed through network k, then w n,k =1, otherwise w n,k =0.

[0113] (C5) This constraint ensures that the allocated network bandwidth does not exceed the total bandwidth B, and K represents the maximum number of communication resources.

[0114] (C6)0≤δ≤1, this constraint specifies the range of the weight factor δ.

[0115] (C7) This constraint ensures that the local computing resources allocated to the task to be processed by the terminal do not exceed

[0116] in, Indicates the maximum computing resources that the terminal locally allocates to the task to be processed.

[0117] (C8) This constraint ensures that the computing resources allocated to the edge server for the task to be processed do not exceed

[0118] in, Indicates the maximum computing resources allocated when the pending tasks are offloaded to the edge server for processing.

[0119] (C9) This constraint ensures that the computing resources allocated by the cloud server to the task to be processed do not exceed

[0120] in, Indicates the maximum computing resources allocated when the pending tasks are offloaded to the cloud server for processing.

[0121] (C10) This constraint ensures that the sum of the proportion of tasks to be processed that remain in the terminal, the proportion that are offloaded to the edge server, and the proportion that are offloaded to the cloud server is 1.

[0122] (C11) This constraint limits the transmission power of each task to be processed to no more than Indicates the maximum transmission power of each task to be processed, which is a preset value.

[0123] Step 103: Solve the association optimization model to obtain the task offloading strategy, network selection strategy and resource allocation strategy. Specifically:

[0124] A two-layer offloading algorithm based on the improved dung beetle optimizer solves the associated optimization model. The two-layer offloading algorithm is divided into two layers. The upper layer of the two-layer offloading algorithm is based on the improved dung beetle optimizer to solve the optimal task offloading strategy and optimal network selection strategy of the associated optimization model. The lower layer of the two-layer offloading algorithm is based on the improved dung beetle optimizer to solve the optimal resource allocation strategy of the associated optimization model.

[0125] Exemplarily, step 103 includes:

[0126] The association optimization model is decoupled into two sub-models through a two-layer offloading algorithm. The upper sub-model is responsible for the task offloading and network selection process of the tasks to be processed. The upper goal is to optimize the coupling of the task offloading strategy and the network selection strategy, which is expressed as the objective function and constraints (C1)-(C6), (C10)-(C11); the lower sub-model allocates resources to the tasks to be processed according to the task offloading strategy and network selection strategy given by the upper sub-model, and finds the optimal resource allocation strategy, which is expressed as the objective function and constraints (C1)-(C3), (C6)-(C9).

[0127] Specifically, the upper algorithm in the two-layer offloading algorithm solves the upper sub-model of the associated optimization model, that is, the optimal task offloading strategy and network selection strategy are solved based on the improved dung beetle optimizer. The lower algorithm in the two-layer offloading algorithm solves the lower sub-model of the associated optimization model, that is, the optimal resource allocation strategy is solved based on the improved dung beetle optimizer.

[0128] like Figure 3 As shown in the figure, the two-layer unloading algorithm based on the improved dung beetle optimizer to solve the associated optimization model mainly has six steps, which are summarized as follows:

[0129] Bernoulli chaotic mapping is used to initialize the dung beetle population and algorithm parameters.

[0130] Calculate the fitness values ​​of all agents according to the objective function.

[0131] Update the positions of all dung beetles according to the adaptive parameter adjustment strategy.

[0132] Determine whether each agent has crossed the boundary.

[0133] Update the fitness value of the current optimal solution.

[0134] Repeat the above steps until iter meets the termination criteria and outputs the global optimal solution and fitness value.

[0135] Exemplarily, the initialization population of the Bernoulli chaotic map unloading algorithm is adopted to improve the diversity of the initial population. The specific process is as follows:

[0136]

[0137] Among them, ψ is the control parameter, z1 is the mapping output, and z0 represents the initialization generation of the chaotic map.

[0138] Initialize the dung beetle population, set the initial position for each dung beetle, and expand it to D-dimensional space. In the boundary range, the initialization of the dung beetle population after adding the chaotic map is expressed as:

[0139] D0=Lb+z1(Ub-Lb).

[0140] Among them, D0 represents the initial position of the dung beetle population, Ub and Lb represent the upper and lower bounds of the problem respectively.

[0141] In addition, an adaptive parameter adjustment strategy is used to dynamically adjust the number of producers (i.e., the number of dung beetles) during the operation of the two-layer unloading algorithm to effectively balance the convergence speed and exploration ability. The specific operations are as follows:

[0142] When updating producers, the adaptive parameter adjustment strategy will set the producer ratio P percent Reduce it as follows:

[0143]

[0144] Among them, P u Iter represents the upper bound of the producer ratio, P1 represents the lower bound of the producer ratio, iter represents the current number of iterations, and Iter represents the max Indicates the maximum number of iterations. In the embodiment of the present application, P u The value of is set to 0.4 and the value of P1 is set to 0.2.

[0145] The dung beetle optimizer based on the dung beetle in nature is as follows.

[0146] In nature, there are four types of dung beetles, including the ball-bearing dung beetle, the female dung beetle, the small dung beetle and the robber dung beetle.

[0147] The specific operations of the four dung beetle behaviors are as follows:

[0148] (1) Rolling behavior

[0149] When the ball-rolling dung beetle rolls its dung ball backward, it can navigate based on moonlight or sunlight to move the dung ball in a straight line. When there is no light source, its trajectory is curved. The intensity of the light source affects the movement path of the ball-rolling dung beetle, and its position update in the search space can be expressed as:

[0150]

[0151] Among them, iter represents the current iteration number; X i (iter) represents the position of the i-th ball beetle in the iter-th iteration; v∈(0,0.2] is a constant value, indicating the deflection coefficient; b∈(0,1] is a constant value, ω is a natural coefficient, where the value is 1 or -1 (usually a random number is used to generate the positive and negative values ​​of the natural coefficient); X w represents the global worst position; ΔX represents the change in light intensity.

[0152] When ball-rolling dung beetles encounter an obstacle and cannot move, they will jump on top of the dung ball, reposition themselves, and obtain a new route. Assuming that the ball-rolling dung beetle will immediately continue to roll the ball backwards after determining the new orientation, this process is represented as a dance behavior, as follows:

[0153] X i (iter+1)=X i (iter)+tanθ|X i (iter)-X i (iter-1)|.

[0154] Among them, θ∈[0,π] represents the azimuth angle. When θ is equal to 0, or π, the position of the rolling dung beetle does not change.

[0155] (2) Reproductive behavior

[0156] In nature, after the ball-rolling dung beetle transports the dung ball to a safe place and hides it, the female dung beetle will lay eggs in the dung ball. The regional boundary selection strategy used to simulate the dung beetle's egg laying is defined as follows:

[0157]

[0158] Among them, X * is the current local optimal spawning location; Lb * and Ub * are the lower and upper limits of the spawning area respectively; Lb and Ub are the lower and upper limits of the optimization problem; and iter indicates the current number of iterations. max Represents the maximum number of iterations, where R is a natural number.

[0159] The boundary range of the egg-laying area changes dynamically, which mainly depends on the R value. Therefore, once the egg-laying area is determined, the female dung beetle will choose the dung ball in this area to lay eggs, which is called a brooding ball. In the unloading algorithm, each female dung beetle lays only one egg in each iteration. The location of the egg can be expressed as follows:

[0160] B i (iter+1)=X * +b1(B i (iter)-Lb * )+b2(B i (iter)-Ub * ).

[0161] Among them, B i(iter) is the position information of the i-th brooding ball in the iter-th iteration, b1 and b2 represent two independent random vectors of size 1×D, and D represents the dimension of the optimization problem. Note that the position of the brooding ball is strictly limited to a certain range, namely the spawning area.

[0162] (3) Foraging behavior

[0163] Some mature dung beetles will come out of the ground to forage, and it is necessary to establish an optimal foraging area to guide the dung beetles to forage. The boundaries of the optimal foraging area are defined as follows:

[0164]

[0165] Among them, Lb b and Ub b are the lower and upper bounds of the optimal foraging area. At this point, the position of the dung beetle is updated as follows:

[0166] X i (iter+1)=X i (iter)+C1(X i (iter)-Lb b )+C2(X i (iter)-Ub b ).

[0167] Among them, C1 is a random number that obeys the normal distribution, and C2 is a random vector within the range of 1×D (0.1).

[0168] (4) Theft

[0169] Some dung beetles, called thieves, steal dung balls from other dung beetles. The two-tier unloading algorithm places the stealing behavior at the best foraging location X b , and the position of the dung beetle is updated as follows:

[0170]

[0171] in is a random vector of size 1×D, and S is a constant value.

[0172] After one iteration, the positions of the ball-rolling dung beetle, egg-hatching ball, small dung beetle and thief dung beetle are updated.

[0173] Step 104, allocate the unloading locations of the tasks to be processed in proportion according to the task unloading strategy, select the corresponding network transmission data based on the network selection strategy, and allocate the computing resources required by the tasks to be processed at the unloading locations through the resource allocation strategy.

[0174] After determining the task offloading strategy, network selection strategy and resource allocation strategy through step 104, the processing ratio of the pending tasks in the offloading location is determined according to the offloading strategy, which networks are used simultaneously for data transmission is determined according to the network selection strategy, and resources are allocated to the pending tasks in the offloading location according to the resource allocation strategy.

[0175] It should be noted that steps 102 to 104 are not performed in sequence, and the task offloading strategy, network selection strategy, and resource allocation strategy obtained in step 104 can be allocated to step 102 or step 103 for use.

[0176] In the embodiment of the present application, the parameter settings of the multi-access edge computing system are shown in Table 1:

[0177] Table 1

[0178]

[0179] like Figures 4 to 7 As shown, the method of MEC task offloading with joint resource allocation and privacy protection provided in an embodiment of the present application is used to offload computing tasks, and the results and effects produced are as follows.

[0180] like Figure 4 As shown in the figure, in order to evaluate the impact of system privacy risk on the utility function during offloading, the change of the target value under different numbers of tasks is studied by adjusting the value of the average number of leakages (λ). The joint privacy entropy is used to measure the potential probability of privacy leakage. As the joint privacy entropy increases, the security of task offloading is also improved.

[0181] Higher joint privacy entropy makes computing tasks more decentralized, thereby protecting the security of important data in the task. As λ increases, the probability mass function of the Poisson distribution shifts to the right, that is, the probability mass moves to the right. At the same time, the peak of the Poisson distribution gradually becomes wider. This change makes the probability of more events increase with the adjustment of λ. When λ = 1, the target value is at the top; when λ = 9, the target value tends to the middle; and when λ = 9, the target value is at the bottom. This is because the change of λ causes the change of entropy value. The larger λ is, the larger the entropy value is, and the smaller the target value is. As the number of tasks increases, the overall target value shows an upward trend. However, when the number of tasks reaches 15, the overall target value decreases slightly. This is because when the number of tasks is greater than 15, the tasks will be limited by resources, resulting in a significant increase in latency, and the optimal solution cannot be achieved, thus satisfying the suboptimal solution. In this case, the target value -1.22 corresponding to λ = 5 is 43.81% lower than the target value -0.85 corresponding to λ = 1, and 48.45% higher than the target value -2.39 of λ = 9. These differences are mainly due to the effect of different λ on the entropy value which leads to an increase in the target value.

[0182] In order to evaluate the multi-network transmission communication model during offloading, Figure 5 The objective function values ​​under various network conditions are compared. Figure 5 As shown in the figure, the change trend of the target value is similar. When the number of networks increases from 1 to 3, the target value decreases from -0.50 to -1.11, a decrease of 54.95%. When the number of networks increases from 3 to 5, the target value decreases from -1.11 to -1.75, a decrease of 57.66%. This is due to the fact that as the number of networks increases, users will try to choose as many networks as possible for transmission. Since the computing power of the edge is much greater than the local computing power, the task processing time is reduced, thereby reducing the target value.

[0183] Analyze the impact of the target weight factor on the total cost and the objective function, and observe the impact of different values ​​on the above indicators by changing the target weight. Figure 6 As shown, in Figure 6 In the figure, the horizontal axis represents the number of tasks and the vertical axis represents the target value.

[0184] Given the same number of tasks, as the target weight factor δ increases, the target value also increases.

[0185] The target weight factor δ is used to achieve a balance between latency and privacy security. The value is different when different requirements are met. When δ = 0.3, it means that the security requirements for the entire system are higher. When the number of tasks increases from 5 to 25, the target value decreases from -3.14 to -12.79. This shows that as the number of tasks increases, the joint privacy entropy value also increases, resulting in a decrease in the target value.

[0186] When δ=0.8, it indicates that the real-time requirement for the task is high. When the number of tasks increases from 5 to 25, the target value increases from 0.35 to 7.55, indicating that as the number of tasks increases, the delay increases significantly, which leads to an increase in the target value.

[0187] For the changes of different weight factors δ under the same number of tasks, it is found that the target value also increases when the weight factor δ increases from 0.3 to 0.8. When the number of tasks is 15, δ increases from 0.3 to 0.6, and the target value increases from -8.91 to -1.46, an increase of 83.61%. This is because a larger weight factor δ indicates a higher real-time requirement and a lower privacy requirement. Different users can choose different weight parameters according to their needs, and can even dynamically select weights according to the needs of actual applications to better meet their needs.

[0188] Figure 7 The following is a schematic diagram showing the performance comparison between the unloading method of the present application and other technologies at different iteration times provided in the embodiments of the present application. Figure 7As shown in the figure, the performance of the algorithm (D-IDBO) proposed in this application is compared with the existing double-layer unloading algorithm based on longhorn beetle whiskers (D-DBO), double-layer unloading algorithm based on longhorn beetle whiskers (D-BAS), double-layer unloading algorithm based on particle swarm (D-PSO), double-layer unloading algorithm based on improved particle swarm of longhorn beetle whiskers (D-BSO), greedy double-layer unloading algorithm (D-GA) and random double-layer unloading algorithm (D-RA). The results are as follows Figure 7 shown.

[0189] D-RA exhibits the worst search performance due to its lack of strategy selection. D-GA always selects the currently optimal solution at each step, but it is easy to fall into the local optimal solution, so a large number of iterations are usually required and the results are averaged. D-PSO originates from the simulation of the foraging process of flocks of birds. It can find the global optimal solution from the individuals of the entire flock of birds. However, due to the large number of particles, the time complexity of the two-layer solution problem is high, which makes it difficult to converge in short-term iterations. D-BAS imitates the strategy of longhorn beetles to forage according to the strength of food odor. It does not need to model the specific form of the function, so it performs well in efficient optimization. However, when dealing with multi-dimensional complex problems, search failures may occur.

[0190] from Figure 7 It can be seen that the algorithm D-IDBO proposed in this application has the smallest target value and a faster convergence speed, and its performance is better than other algorithms in the prior art.

[0191] Although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in this embodiment is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in sequence or in parallel according to the method shown in this embodiment or the accompanying drawings (for example, in a parallel processor or multi-threaded processing environment).

[0192] Some modules in the apparatus described in the present application can be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0193] The devices or modules described in the above application embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described in various modules according to their functions. When implementing the embodiments of the present application, the functions of each module can be implemented in the same or multiple software and / or hardware. Of course, the module that implements a certain function can also be implemented by combining multiple sub-modules or sub-units.

[0194] The methods, devices or modules described in this application can be implemented in the form of computer-readable program codes. The controller can be implemented in any appropriate manner. For example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program codes (such as software or firmware) that can be executed by the (micro) processor, logic gates, switches, application-specific integrated circuits (English: Application Specific Integrated Circuit; Abbreviation: ASIC), programmable logic controllers and embedded microcontrollers. Examples of controllers include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in a purely computer-readable program code, the controller can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded microcontrollers by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included in it for implementing various functions can also be regarded as structures within the hardware component. Or even, the means for realizing various functions may be regarded as both a software module for realizing the method and a structure within a hardware component.

[0195] An embodiment of the present application also provides an apparatus for performing a method for MEC task offloading with joint resource allocation and privacy protection, the apparatus comprising: a processor; a memory for storing processor executable instructions; when the processor executes the executable instructions, the method described in the embodiment of the present application is implemented.

[0196] The embodiments of the present application also provide a non-volatile computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed, the method described in the embodiments of the present application is implemented.

[0197] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist independently, or two or more modules may be integrated into one module.

[0198] The above storage media include but are not limited to random access memory (RAM), read-only memory (ROM), cache, hard disk (HDD) or memory card. The memory can be used to store computer program instructions.

[0199] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary hardware. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, or it can be reflected in the implementation process of data migration. The computer software product can be stored in a storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0200] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. All or part of this application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, mobile communication terminals, multi-processor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0201] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit the present application. Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present application.

Claims

1. A method for joint resource allocation and privacy protection of MEC task offloading, characterized in that: include: Acquire information of the task to be processed, the information of the task to be processed including the amount of input data of the task to be processed, the number of CPU cycles required to process the task to be processed, the level of leakage information of the task to be processed, the maximum tolerance time of the task to be processed, and the maximum energy consumption of the task to be processed; Establish an association optimization model through communication model, computation model and joint privacy entropy model; Solving the associated optimization model to obtain a task offloading strategy, a network selection strategy, and a resource allocation strategy; Allocate the unloading locations of the tasks to be processed in proportion according to the task unloading strategy, select the corresponding network to transmit data based on the network selection strategy, and allocate the computing resources required by the tasks to be processed at the unloading locations through the resource allocation strategy; The establishment of the association optimization model through the communication model, the calculation model and the joint privacy entropy model includes: Constructing the communication model, the calculation model and the joint privacy entropy model; Based on the communication model, the network selection of the task to be processed is obtained; the task unloading ratio, resource allocation ratio, task processing time and minimized task processing time of the task to be processed at the unloading location are obtained through the calculation model; and the joint privacy entropy and the maximized joint privacy entropy of the task to be processed are obtained according to the joint privacy entropy model; Integrate the task offloading ratio, the network selection, the resource allocation ratio and the task processing time, and establish an associated optimization model between the task processing time and the joint privacy entropy; The association optimization model includes an objective function and a plurality of constraints; After the association optimization model is established, the following steps are also included: Taking the minimization of task processing time and the maximization of joint privacy entropy as optimization targets of the objective function, and normalizing the optimization targets; The normalized optimization objective is transformed using the linear weighted method to obtain the function value: Among them, U n represents the normalized linear weighted value of the optimization target, N represents the number of tasks to be processed, X represents the strategy after the task offloading strategy and the network selection strategy of the tasks to be processed are coupled, X=[α n ,β n,k ,γ n,k ] n×(2k+1) , where α n Indicates the proportion of tasks to be processed executed on the terminal, β n,k represents the proportion of the pending tasks executed on the edge server, γ n,k represents the proportion of the pending tasks executed in the cloud server, k represents the kth network, F is the resource allocation strategy for the pending tasks, in, represents the computing resources allocated when the pending task is processed at the terminal, represents the computing resources allocated when the pending task is processed on the edge server, represents the computing resources allocated when the pending task is processed on the cloud server, l represents that the pending task is executed on the terminal, es represents that the pending task is executed on the edge server, and c represents that the pending task is executed on the cloud server; The solving of the association optimization model comprises: Solving the association optimization model based on a double-layer offloading algorithm of an improved dung beetle optimizer; The double-layer offloading algorithm is divided into two layers, and the upper layer of the double-layer offloading algorithm solves the optimal task offloading strategy and the optimal network selection strategy of the associated optimization model based on the improved dung beetle optimizer; The lower layer of the two-layer offloading algorithm solves the optimal resource allocation strategy of the associated optimization model based on the improved dung beetle optimizer; The double-layer unloading algorithm based on the improved dung beetle optimizer solves the associated optimization model, including: The initialization population of the two-layer unloading algorithm using Bernoulli chaotic mapping is: Among them, ψ is the control parameter, z1 is the mapping output, and z0 represents the initialization generation of the chaotic map; Initialize the dung beetle population, set an initial position for each dung beetle, and expand to a D-dimensional space. In the boundary range, the initialization of the dung beetle population after adding the chaotic map is expressed as: D0=Lb+z1(Ub-Lb); wherein D0 represents the initialization position of the dung beetle population, Ub represents the upper bound of the problem to be solved, and Lb represents the lower bound of the problem to be solved; Adopt an adaptive parameter adjustment strategy to dynamically adjust the number of producers during the operation of the offloading algorithm; When the adaptive parameter adjustment strategy updates the number of producers, the producer ratio will be reduced. The expression for reducing the producer ratio is: Among them, P percent represents the producer ratio, P u represents the upper bound of the producer ratio, P l Iter represents the lower bound of the producer ratio, iter represents the current number of iterations, and Iter max Indicates the maximum number of iterations.

2. The method according to claim 1, characterized in that The unloading locations include terminals, edge servers and cloud servers.

3. The method according to claim 2, characterized in that The communication model supports using multiple wireless networks to simultaneously perform data transmission between the terminal and the edge server and data transmission between the edge server and the cloud server.

4. The method according to claim 1, characterized in that: The joint privacy entropy model includes task privacy entropy and data privacy entropy; The task privacy entropy is used to obtain the leakage probability and leakage level of the to-be-processed task at different unloading locations; The data privacy entropy is used to obtain the leakage probability and leakage level of the data of the task to be processed during the data transmission process.

5. A device for performing a method for MEC task offloading with joint resource allocation and privacy protection, characterized in that: include: processor; a memory for storing processor-executable instructions; When the processor executes the executable instructions, the method according to any one of claims 1 to 4 is implemented.

6. A non-volatile computer-readable storage medium, characterized in that: The device comprises a computer program or an instruction for storing the computer program or the instruction, which, when executed, enables the method according to any one of claims 1 to 4 to be implemented.

Images