1. A task unloading method facing to location privacy protection is characterized by comprising the following steps:

s1, generating a virtual position with the current equipment position l (x, y) as the center of circle and r as the center of circle₁,r₂Forming a circular virtual position space S for the radius, selecting a virtual position l with a virtual position selection probability p₁'(x₁',y₂') and another virtual position l is selected on the other side of the circle with the center as the symmetry₂'(x₁',y₂') p is a real number from 0 to 1, r₁、r₂Are all real numbers greater than 0 and r₁＜r₂；

S2, constructing a server selection matrix, acquiring the positions of M nearby servers by the equipment, calculating the real distance between the equipment and the M servers according to the real positions of the equipment, and calculating the real distance between the equipment and the M servers according to the real positions of the equipmentDistance evaluates privacy level for each server, calculates transmission cost according to distance and bandwidth, and forms a server selection matrix Tc, wherein element Tc_(i)The privacy level and the transmission cost of the ith server and equipment are represented, M is a positive integer, and i is a positive integer not greater than M;

s3, selecting the server sending the unloading request, calculating the server which can effectively protect the position information and pay lower transmission cost according to the server selection matrix obtained in the step S2, and selecting the server according to the server selection probability p_chSelecting an optimal server to send a task unloading request;

s4, calculating a task matrix, firstly allocating bandwidth to waiting task unloading equipment in a current queue according to the distance, then calculating transmission delay, local calculation time, local calculation energy consumption, edge calculation time and edge calculation energy consumption, and forming the parameters into a task matrix Tt with N rows, wherein an element Tt_(i)Representing the correlation attribute of the ith task, wherein N is a positive integer, and i is a positive integer not greater than N;

and S5, using reinforcement learning modeling, and taking the decision parameters of the task as input to obtain a task unloading result.

2. The task offloading method for location privacy protection as recited in claim 1, wherein: step S1, initializing the virtual position selection probability p according to the distance between the position point of the virtual space and the real position, and conforming to normal distribution in distance from near to far;

mu is a position parameter, and sigma is a scale parameter; denotes p obedience is expected to be μ and variance is σ²Probability distribution of (2).

3. The task offloading method for location privacy protection as recited in claim 1, wherein: the specific step of constructing the server selection matrix in step S2 includes:

(1) calculating the real distance between the equipment and the server according to a longitude and latitude distance formula;

(2) judging the privacy protection level of each server according to a privacy judgment formula;

(3) the transmission cost is estimated from the distance and the bandwidth.

4. The task offloading method for location privacy protection as recited in claim 1, wherein: the step S3 of selecting a server that sends an offload request specifically includes:

(1) selecting a group of servers with low privacy disclosure risk and low transmission cost by using linear programming;

(2) calculating the server selection probability by using the current speed and the displacement direction of the user;

(3) and (3) selecting the server selected in the step (1) according to the server selection probability to be the most suitable server for submitting the task unloading request in the appointed time.

5. The task offloading method for location privacy protection as recited in claim 4, wherein: the step (1) specifically comprises the steps of firstly carrying out relaxation operation on integer constraint conditions of an integer linear programming problem, then carrying out filtering operation on a solved result, finally comparing each candidate solution, reserving a server with better performance, and omitting the rest parts.

6. The task offloading method for location privacy protection as recited in claim 1, wherein: the task matrix specifically calculating in step S4 includes:

(1) allocating bandwidth according to a distance between the device and the server;

(2) calculating task transmission delay according to the bandwidth and the distance;

(3) calculating transmission delay according to the transmission power of the equipment and the size of the task;

(4) calculating task local calculation time according to the calculation capacity of the equipment and the calculation power required by the task;

(5) calculating local calculation energy consumption according to the equipment power and the local calculation time;

(6) calculating edge calculation time according to the edge calculation resources and the calculation force required by the task;

(7) and calculating the edge calculation energy consumption according to the edge calculation power consumption and the calculation time.

7. The task offloading method for location privacy protection as recited in claim 1, wherein: the selection of the unloading task in the step S5 specifically includes the following processes:

(1) firstly, pre-trimming a task waiting for an unloading decision;

(2) taking the computing power and the resource quantity of the server as the server state;

(3) taking the tolerance time and the required computing power of each task as input, and outputting an unloading strategy by a neural network;

(4) and updating the server state.

8. The task offloading method for location privacy protection as recited in claim 7, wherein: the pruning processing in the step (1) comprises the steps of selecting and locally executing the tasks which exceed the tolerance time, and then calculating the tasks which cannot successfully return the result within the tolerance time and are locally executed under the current existing resource and calculation power state.

9. The task offloading method for location privacy protection as recited in claim 7, wherein: and (3) the neural network output unloading strategy is to carry out priority sequencing according to the task tolerance time and unload the tasks which save more energy consumption and time and benefit greatly to the edge nodes for processing.

Background

Location privacy protection refers to protecting a user's location information from being revealed when serving based on location, thereby protecting the user's other sensitive information from being inferred. More and more researchers are beginning to focus on the location privacy protection problem. The edge task unloading refers to that in the edge calculation, the edge server selects part of tasks to be processed on the edge node in the task unloading request submitted by the mobile equipment, and the rest tasks are processed locally.

The existing location privacy protection methods mainly comprise anonymity, location fuzziness and encryption. The anonymity method is to hide the identity of the user, and usually delete the identity attribute first, and then perform anonymization processing on the identity attribute. Anonymity may protect not only user location privacy but also user data privacy, but anonymity is easily de-anonymized. The position ambiguity refers to the fact that the accurate position of the user is blurred into a spatial range, and the server only knows that the user is in the ambiguous space, but does not know the exact position. The location ambiguity may degrade the quality of service of the LBS (location based service). Encryption is to encrypt data by using cryptography, and the processing and query of each position information are based on ciphertext, so that an illegal attacker cannot decrypt the real position and identity information of a user. And the encryption process increases time overhead and transmission overhead.

The advent of mobile edge networks in recent years has enabled some of the challenges related to computationally intensive tasks to be effectively addressed. When a user device begins running a compute-intensive task and is in the mobile edge network, the device may choose to send the compute task to a nearby common server, a process known as Offloading (Offloading). And the tasks are unloaded to the edge computing server for processing, and the computing resources and the electricity of the user equipment are not consumed. Therefore, the mobile edge calculation can effectively reduce the resource occupation and energy consumption of the equipment and simultaneously provide faster response speed. But we also face limitations in the process of computing offload. The limited computing resources and bandwidth resources do not ensure that all of the computing requests are processed at the first time, and the limited bandwidth makes the users sending and receiving data at the same time limited. If a large number of users are directly allocated to the same network access point or a large number of tasks are allocated to the same server without adopting a certain strategy, all the users cannot submit tasks at a normal rate and receive calculation results. How to allocate limited computing and bandwidth resources to a large number of offloading computing tasks, and plan a reasonable offloading policy and resource allocation policy for a computing task requested by a user equipment is an urgent problem to be solved in the mobile edge network technology. The existing task unloading method mainly comprises methods such as linear planning, resource matching, game theory and the like.

Disclosure of Invention

The invention aims to solve the problems of position privacy and task unloading in the existing edge calculation, and provides a method for realizing task unloading decision and resource allocation by using a reinforcement learning method while protecting position privacy.

The technical scheme adopted by the invention for solving the technical problems is as follows: a task unloading method facing to location privacy protection comprises the following steps:

s1, generating a virtual position with the current equipment position l (x, y) as the center of circle and r as the center of circle₁,r₂Forming a circular virtual position space S for the radius, selecting a virtual position l with a virtual position selection probability p₁'(x₁',y₂') and another virtual position l is selected on the other side of the circle with the center as the symmetry₂'(x₁',y₂') p is a real number from 0 to 1, r₁、 r₂Are all real numbers greater than 0 and r₁＜r₂。

S2, constructing a server selection matrix, obtaining positions of M nearby servers by equipment, calculating real distances between the equipment and the M servers according to the real positions of the equipment, evaluating privacy level of each server according to the distances, calculating transmission cost according to the distances and the bandwidth, and forming a server selection matrix Tc, wherein element Tc_(i)Representing privacy level and transmission cost of the ith server and device, M being a positive integer, and i being not greater than MA positive integer.

S3, selecting the server sending the unloading request, calculating the server capable of effectively protecting the position information and paying lower transmission cost according to the server selection matrix obtained in the step S2, and selecting the probability p according to the server_chAnd selecting an optimal server to send a task unloading request.

S4, calculating a task matrix, firstly allocating bandwidth to waiting task unloading equipment in a current queue according to the distance, then calculating transmission delay, local calculation time, local calculation energy consumption, edge calculation time and edge calculation energy consumption, and forming the parameters into a task matrix Tt with N rows, wherein an element Tt_(i)Representing the correlation property of the ith task, N is a positive integer, and i is a positive integer not greater than N.

And S5, using reinforcement learning modeling, taking decision parameters (including the calculation transmission delay, the sending delay, the local calculation time, the local calculation energy consumption, the edge calculation time and the edge calculation energy consumption in the step S4) of the task as input, and obtaining a task unloading result.

The invention has the following advantages and beneficial effects:

1. the invention considers the position privacy protection of the user in the unloading of the edge computing task, and obtains the position of the server by generating the virtual position by utilizing the position fuzzy. The method comprises the steps of considering the position leakage risk when a server is selected to send a task unloading request for the first time, balancing privacy and transmission cost, and selecting the server which can protect the user position privacy and has low transmission cost. This approach avoids exposing the user's location directly to the server and reduces the risk of location awareness by the server.

2. When the user bandwidth is distributed, the channel transmission delay of the user is considered, and because the user selects the server not to be the nearest target, the bandwidth is reasonably distributed according to the distance and the channel quality by considering the delay problem.

3. Compared with the traditional cloud, the edge cloud computing is closer to the user equipment, network transmission delay can be reduced, the application program running speed is increased, the energy consumption of the user equipment is saved, and the user experience is improved. And (3) a reinforcement learning online learning task unloading strategy is used, delay and server resource conditions are fully considered, the goal of maximizing benefits and saving energy consumption and time is taken as a target, the optimal unloading strategy is found, and the performance of the mobile edge cloud computing is greatly improved.

Drawings

FIG. 1 is a schematic diagram of the location privacy protection oriented task offload policy of the present invention;

FIG. 2 is a schematic diagram of a location privacy preserving framework;

FIG. 3 is a server selection matrix;

FIG. 4 is a task offloading diagram of reinforcement learning.

Detailed Description

Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.

The invention discloses a task unloading method facing to position privacy protection, which comprises the following steps as shown in figure 1:

firstly, generating a virtual position, taking the current position l (x, y) as the center of a circle and r as the center of a circle₁,r₂Forming a circle for the radius, selecting a virtual location l with an initial random probability p₁'(x₁',y₂') and another virtual position l is selected at the other side symmetrical to the center of the circle₂'(x₁',y₂'). The p is a real number from 0 to 1, and r₁、r₂Are all real numbers greater than 0 and r₁＜r₂

In the embodiment, the above obtained virtual position is used to initiate a server request based on LBS, and the position of the server deployed in the base station is obtained to obtain the position information set of the server.

The method for generating a virtual position using a blurred actual position comprises the following steps:

(1) taking the equipment position l (x, y) as a circle center, solving a virtual position space S according to a distance formula:

S＝{s(x',y')|x+r₁＜x'＜x+r₂,y+r₁＜y'＜y+r₂}

where x ', y' represent the latitude and longitude coordinates of the virtual location, r₁，r₂Representing the difference between warp and weft coordinates in virtual space and a real position; and in order to make the difference between the virtual space and the real position not to affect the service quality too much, the present embodiment sets 0.005 < r1 < r2 < 0.01.

(2) Initializing a virtual position selection probability p according to the distance between the position point of the virtual space and the real position, and conforming to normal distribution in distance from near to far;

μ is a position parameter and σ is a scale parameter. Denotes p obedience is expected to be μ and variance is σ²Probability distribution of (2).

And secondly, constructing a server selection matrix. The device obtains the positions of M nearby servers, the real distances between the device and the M servers are calculated according to the real positions of the device, privacy levels of all the servers are evaluated according to the distances, transmission costs are calculated according to the distances and the bandwidths, and a server selection matrix Tc is formed. Wherein the element Tc_(i)And the privacy level and the transmission cost of the ith server and equipment are represented, wherein M is a positive integer, and i is a positive integer not greater than M.

In the present embodiment, the device location information l (x, y) and the server location information l are used_i(x_i,y_i) Calculating the distance dis according to the latitudinal distance formula:

C＝sin(y)sin(y_i)cos(x-x_i)+cos(y)cos(y_i)

wherein R represents the average radius of the earth.

And judging the privacy protection level of each server by using a privacy judgment formula according to the distance dis:

wherein dis_i,mjRepresenting devices i and BS_mjThe distance between them; r_mj,rangeRepresents BS_mjThe maximum communication range of; BS_mjJ-th device representing that device i can communicate with base station

Estimating transmission cost from distance and bandwidth

Ts_i,mj＝B_mj*η_i

η_i＝log(1+S/N)

Wherein, Ts_i,mjRepresenting devices i and BS_mjTransmission rate of B_mjRepresenting bandwidth resources of the server, η_iThe uplink spectral efficiency of the device is shown, the S/N is the signal-to-noise ratio of the device, and the Task is the size of the calculation Task of the device.

Thirdly, selecting a server sending the unloading request, calculating the server capable of paying lower transmission cost while effectively protecting the position information by using linear programming according to the server selection matrix obtained in the second step, and selecting the probability p according to the server_chAnd selecting the optimal server.

In this embodiment, a matrix is selected according to the server obtained in the second step, as shown in fig. 3. (1) A linear programming problem is formulated, and the objective function is as follows: maxz ═ (c1 privy + c2/delay) p, the associated constraint is as follows: delay is less than or equal to 20, and privacy is equal to {2,3,4 }. And selecting a server meeting the conditions through linear programming, and firstly relaxing the integer constraint conditions of the integer linear programming problem so as to convert the integer constraint conditions into solvable linear programming problems. Solving the linear programming problem according to the constraint to obtain an optimal solution (privacy, delay), then performing filtering operation on the solved result, because the position can be perceived by being too close to the server, filtering the servers meeting the conditions that c3 is not more than dis/range is not more than c4, the more devices connected with the servers means that the resource distribution of the servers is limited, if the number of the currently selectable servers is more than one, filtering the servers with the connection device N more than 10, finally, comparing each candidate solution, only keeping the server with better performance as the final solution, and discarding the rest.

(2) And then uses the current speed v of the user_tAnd the direction of displacement (x)_a,y_a) Computing a server selection probability p_ch：

Wherein R is_mj,rangeIndicating the communication range of the base station; t is t_rRepresenting the constrained time of the device computing task. x is the number of_i， y_iA horizontal and vertical coordinate representing the server location.

(3) And selecting the server selected by the linear programming according to the selection probability, and selecting the server which is most suitable for unloading in the appointed time.

And fourthly, calculating a task matrix. First, allocating bandwidth to waiting task unloading equipment in a current queue according to distance. And then evaluating the offloading cost and benefit of all the tasks requesting offloading. And calculating energy consumption and delay according to basic parameters of the task in the task unloading request, including parameters such as distance, data size and required calculation force. Specifically, the transmission delay is calculated according to the bandwidth, and the sending delay, the transmission energy consumption, the local calculation time, the local calculation energy consumption, the edge calculation time and the edge calculation energy consumption are calculated according to the task unloading request. These parameters form a task matrix Tt of N rows, where the elements Tt_(i)And representing the correlation attribute of the ith task, wherein N is a positive integer, and i is a positive integer not greater than N.

(1) Allocating bandwidth B according to distance between device and server_i：

Wherein, B_mjRepresenting all bandwidth resources of the server, d_i,mjIndicating the distance between the device and the base station, D indicating the connection to the BS_mjThe sum of the device distances of (a).

(2) Calculating task transmission delay according to bandwidth and distance

(3) According to the device transmission power p^sendAnd task size L_taskiCalculating the transmission delay:

T_i ^send＝L_taskip^send

(4) according to the device transmission power p_t,iAnd transmission delayCalculating transmission energy consumption:

(4) calculating the task local calculation time according to the calculation capacity of the equipment and the calculation power required by the task:

wherein, C_taskiThe computational power required for the task is represented,representing local computing power.

(5) Calculating local calculation energy consumption according to the equipment power and the local calculation time:

wherein the content of the first and second substances,representing the device power.

(6) Calculating edge calculation time according to edge calculation resources and the calculation force required by the task:

wherein the content of the first and second substances,representing edge computing resources.

(7) And calculating the edge calculation energy consumption according to the edge calculation power consumption and the calculation time.

Wherein the content of the first and second substances,the edge calculation power is represented by the edge calculation power,indicating the calculation time.

And fifthly, using reinforcement learning modeling, and taking decision parameters of the task as input to obtain a task unloading result. Firstly, a solving process of a calculation task unloading strategy based on a reinforcement learning theory is given: first, according to the reinforcement learning theory, several important parts of the problem to be solved are defined.

The reinforcement learning process needs to convert the original problem into a Markov decision process < S, A, R >, namely a process consisting of a state S, an action A and a reward R. The system starts from a certain state, selects and executes the action according to the current state, then reaches a new state, and obtains the reward corresponding to the new state.

Here we define the remaining computing power and resources of each server over time t as the state it was in. In each state, the selectable action of the user equipment is taken as three decision actions which respectively represent no action, the computing task is locally operated, and the computing task is unloaded to the edge cloud server to be operated. Reward information R for each state_tDefined as a weighted sum W that saves energy and time and revenue when this state is reached, if taking an offload action results in less energy consumption, then offload to the edge cloud server as a computing task T_i,tThe output action a is 1; otherwise, the operation is performed locally on the user equipment as an unloading strategy, and the output action a is 0, that is, the unloading is not performed. And considering the task tolerance time, performing priority sequencing according to the task tolerance time, unloading the tasks which save more energy and time and benefit greatly to the edge nodes for processing, and using the income obtained by unloading the tasks and the saved energy and time as rewards.

In this embodiment, the task waiting for the unload decision is pruned in advance. And selecting the task which is executed locally and cannot successfully return a result within the tolerance time after the task exceeds the tolerance time is executed locally, and then calculating the current existing resources and calculation power state to select the task to be unloaded locally. Secondly, the computing power and the resource quantity of the server are used as the server state; the tolerance time and the required computing power of each task are used as input, the neural network outputs an unloading strategy, and the strategy of the neural network output is obtained, and meanwhile, the reward R brought by the strategy is obtained_tHere, the benefitsDefining the weighted sum of the obtained benefit, the saved energy consumption and the time of the unloading task; then the state information S of the decision making process is processed_tDecision result A_tReward R_tAnd the new state information S reached_t+1And storing the experience in an experience replay cache as historical experience. Finally, updating the state of the server; in the subsequent decision making process, a batch of historical experience training neural network parameters are randomly extracted from the cache every N decision making processes. The method is to use a random gradient descent method to adjust the neural network parameters in the direction of increasing the yield.

The task unloading strategy facing to the position privacy protection can effectively protect the position privacy of the user; and the multi-objective optimization method can accumulate more benefits while reducing energy consumption and delay, and considers the benefits of service providers while considering the quality of service of users.

While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.