A method, device and medium for RSMA-assisted MEC system secure uninstallation

Abstract

The present invention belongs to the field of wireless communication technology, and specifically discloses a method, device and medium for RSMA-assisted MEC system secure offloading; the method comprises: establishing an RSMA-assisted secure MEC system model containing an internal eavesdropper; establishing a target optimization model with the goal of minimizing the maximum delay of the secure MEC system model and taking the common rate allocation vector, precoding matrix and offloading ratio in the secure MEC system model as optimization variables; obtaining a sub-problem of jointly optimizing the precoding matrix and the common rate allocation vector by fixing the offloading ratio, and approximating the sub-problem of the first problem to a convex optimization problem by using a continuous convex approximation algorithm; obtaining a sub-problem of optimizing the offloading ratio by fixing the precoding matrix and the common rate allocation vector, and the sub-problem of the second problem being a convex optimization problem; and alternately iterating the approximated sub-problem of the first problem and the sub-problem of the second problem by using a block coordinate descent method until the iterative error of the iterative objective function is less than a threshold value, thereby obtaining the optimal delay of the target optimization model.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technologies, and in particular to a method, device, and medium for securely unloading an RSMA-assisted MEC system. Background Art

[0002] Mobile Edge Computing (MEC) shifts computing, storage, and network services from traditional centralized cloud service centers to the edge of the network—access points close to demand points (such as base stations, user devices, vehicles, and drones). This architecture aims to improve system responsiveness, reduce bandwidth pressure and latency, and better support large-scale real-time data processing needs, as well as the computationally intensive and latency-sensitive applications required by modern mobile devices. By moving computing and storage resources from cloud service centers to the network edge, MEC significantly shortens data transmission distances, alleviates network congestion, improves responsiveness, and reduces bandwidth consumption and latency. This makes it effective for applications requiring low latency and high bandwidth. Unlike traditional cloud computing, which requires data to be transmitted to remote data centers for processing, increasing network burden, latency, and limiting real-time responsiveness, MEC decentralizes cloud service functions to edge nodes, reducing reliance on the core network, optimizing the allocation of computing resources, and meeting the efficient computing needs of mobile devices and IoT devices.

[0003] Existing secure mobile edge computing (MEC) networks mostly use three communication methods. One method provides services to users through multi-carrier channels. Although this method can effectively avoid signal interference between users, the bandwidth allocated to each user is significantly reduced, significantly reducing the utilization of communication resources. Another method uses spatial division multiple access (SDMA) for signal transmission. Although this method allows all users to share the same time-frequency resources, it inevitably causes inter-user interference during transmission, which is an extreme interference management strategy that treats interference as noise. Finally, the use of NOMA for signal transmission also allows all users to share the same time-frequency resources, but it causes a complete user message to be mapped into a common stream, resulting in an extreme interference management strategy of complete decoding interference. This, in turn, allows the entire message to be decoded by other users, making secure communication impossible in the case of internal eavesdroppers and only guaranteed in the case of external eavesdroppers.

[0004] Therefore, the three communication methods mentioned above either fail to further improve the system's latency performance or are unable to resist internal eavesdroppers. In other words, although current secure mobile edge computing (MEC) networks can use traditional communication methods to ensure certain security performance, they often cannot simultaneously meet the system's latency requirements. Summary of the Invention

[0005] In order to overcome the shortcomings of the existing technology, the present invention provides a method for RSMA-assisted MEC system security offloading, which enables all users in the mobile edge computing network to share the same time and frequency resources, while reducing interference between users while ensuring communication security during transmission, thereby further improving the system's delay performance.

[0006] A second object of the present invention is to provide a corresponding electronic device.

[0007] A third object of the present invention is to provide a computer-readable storage medium.

[0008] The technical solution of the present invention to solve the above technical problems is:

[0009] A method for RSMA-assisted MEC system secure uninstallation includes the following steps:

[0010] S1: Establish a RSMA-assisted secure MEC system model with internal eavesdroppers;

[0011] S2: With the goal of minimizing the maximum latency of the secure MEC system model, a target optimization model is established using the common rate allocation vector, precoding matrix, and offload ratio in the secure MEC system model as optimization variables.

[0012] S3: By fixing the offloading ratio β, we obtain subproblem 1 of jointly optimizing the precoding matrix W and the common rate allocation vector r. We then use the continuous convex approximation algorithm to approximate subproblem 1 as a convex optimization problem.

[0013] S4: By fixing the precoding matrix W and the common rate allocation vector r, the second sub-problem of optimizing the offloading ratio β is obtained. This sub-problem is a convex optimization problem.

[0014] S5: Use the block coordinate descent method to alternately iterate the approximate sub-problem 1 in step S3 and the sub-problem 2 in step S4 until the iterative error of the iterative objective function is less than the threshold ∈, ∈, and the optimal delay of the system is obtained.

[0015] In a preferred embodiment of the present invention, in step S1, the steps of constructing the secure MEC system model are as follows:

[0016] Step S101: Assume that the edge user needs to perform a computing task with D bits; where β k The computation task of D bits needs to be offloaded to auxiliary user k for computation. remaining The computing task of bits is performed at the edge user; wherein the edge user is equipped with N tAntenna; The auxiliary user is equipped with a single antenna and integrated with an MEC computing module; It is assumed that the task segment offloaded to auxiliary user k must be kept confidential from the other K-1 auxiliary users, that is, each auxiliary user not only decodes the message it needs as a legitimate receiver, but also acts as a potential internal eavesdropper to intercept messages intended for other auxiliary users;

[0017] Step S102: The message sent by the edge user to the auxiliary user k is represented as W k , message W k Including the corresponding assigned computing tasks β k D bits; the message W k Split into a common message W c,k and a private message W p,k ; Among them, the public message W from all auxiliary users c,1 ,...,W c,K Merged into a super public message W c ; Encode it into a common stream x using the codebook shared by all auxiliary users c , all auxiliary users need to decode the common stream x c ; Private messages W from all auxiliary users p,1 ,...,W p,K are encoded separately as private streams x1,...,x K ,Each private stream is decoded independently only by its corresponding auxiliary user;

[0018] Step S103: Place the computing tasks that need to be securely transmitted in private messages, and place the remaining computing tasks in public messages for transmission;

[0019] Step S104: Definition And its linear preprocessing satisfies E(xx H )=I;wherein,

[0020] The signal sent by the edge user is:

[0021]

[0022] Where: W is the precoding matrix, Precoders corresponding to the public stream and the private stream of auxiliary user k respectively;

[0023] The signal received by auxiliary user k is:

[0024]

[0025] Where, represents the channel coefficient between the edge user and the auxiliary user k; zk is the corresponding additive white Gaussian noise with zero mean and variance σ 2 ,Right now

[0026] In a preferred embodiment of the present invention, in step S1, K-1 auxiliary users acting as internal eavesdroppers intercept the private message of auxiliary user k in a non-cooperative manner, or, K-1 auxiliary users acting as internal eavesdroppers collude with each other to intercept the private message of auxiliary user k.

[0027] In a preferred embodiment of the present invention, in step S2, K-1 auxiliary users acting as internal eavesdroppers intercept the private message of auxiliary user k in a non-cooperative manner. The target optimization model is constructed as follows:

[0028] Step S201: Calculate the auxiliary user k’s decoding of the public stream x c and private stream x k The achievable rate is

[0029] Auxiliary user k decodes the public stream x c The achievable rate R ck for:

[0030]

[0031] Assist user k in decoding private stream x k The achievable rate is:

[0032]

[0033] Step S202: To ensure that all auxiliary users can successfully decode the super public message W c , the following conditions must be met:

[0034]

[0035] r k ≥0;

[0036] Where: r k is the actual transmission rate of the public message assigned to auxiliary user k;

[0037] Step S203: Calculate the total achievable rate of auxiliary user k:

[0038] R k =r k +R p,k ;

[0039] Step S204: Calculate the achievable eavesdropping rate of auxiliary user j decoding the private stream of auxiliary user k:

[0040]

[0041] Step S205: Calculate the confidentiality rate of the private stream transmitted by auxiliary user k:

[0042]

[0043] Where:

[0044] Step S206: Through the transmission delay of auxiliary user k, the processing delay of edge user local calculation and the processing delay offloaded to auxiliary user k k The total delay is calculated by calculating the processing delay of the D-bit computing task; where

[0045] The transmission delay of auxiliary user k is:

[0046]

[0047] Where: B represents the system bandwidth;

[0048] The processing delay of local computing by edge users is:

[0049]

[0050] Where: φ0 is the number of CPU cycles required for edge users to calculate 1 bit of data; f0 is the computing resources of edge users;

[0051] The computational tasks β offloaded to auxiliary user k k The processing delay of D is:

[0052]

[0053] Where: φ k represents the number of CPU cycles required to assist the user in calculating 1 bit of data, f k Computing resources to assist users;

[0054] The total delay is:

[0055]

[0056] Step S207: Establishing an initial target optimization model (P1):

[0057]

[0058] C2:tr(WW H )≤P,

[0059]

[0060] Then the max function of the total delay in the target optimization model is processed to obtain an equivalent target optimization model (P2):

[0061]

[0062] C1,C2,C3,C4.

[0063] Where: P and They represent the maximum transmission power budget of the edge user and the minimum secure transmission rate threshold of each private message respectively; constraint C1 requires that each auxiliary user can successfully decode the public message; constraint C2 requires that the transmission power allocation of the edge user cannot exceed the maximum power budget; constraint C3 requires that the offloaded computing tasks cannot exceed the total computing tasks owned by the edge user; constraint C4 requires that each private message can be securely offloaded.

[0064] In a preferred embodiment of the present invention, in step S3, given the task offloading ratio β, the first sub-problem of jointly optimizing the precoding matrix W and the common rate allocation r is obtained as follows:

[0065]

[0066] C10:tr(WW H )≤P,

[0067]

[0068] Where: Constraint C7, Constraint C9 and Constraint C11 are non-convex constraints.

[0069] In a preferred embodiment of the present invention, the steps of using a continuous convex approximation algorithm to approximate subproblem 1 as a convex optimization problem are as follows:

[0070] Step S301: Introduce non-negative auxiliary variables Replace R respectively c,k 、R p,k and Rewrite constraints C7, C9, and C11 as follows:

[0071]

[0072]

[0073] Where: Constraint C12a, Constraint C12b, Constraint C12c, and Constraint C13b are non-convex constraints;

[0074] Introducing non-negative auxiliary variables and Instead of (r k +α p,k -α e,k ) and Rc,k 、R p,k 、 In the signal-to-interference-plus-noise ratio part, rewrite constraints C12a, C12b, C12c, and C13b as follows:

[0075]

[0076]

[0077]

[0078]

[0079] Where: Constraint C16b, Constraint C17a, Constraint C17b, and Constraint C18b are non-convex constraints, and the other constraints are convex constraints;

[0080] Step S302: Constraints C16b, C17a, C17b, and C18b are linearly approximated as convex constraints using first-order Taylor expansion; wherein,

[0081] Constraint C16b is linearly approximated as a convex constraint:

[0082]

[0083] Where:

[0084] Constraint C18b is linearly approximated as a convex constraint:

[0085]

[0086] Where:

[0087] Constraint C17a is linearly approximated as a convex constraint:

[0088]

[0089] Where:

[0090] Constraint C17b is linearly approximated as a convex constraint:

[0091]

[0092] Where:

[0093] Step S303: Subproblem 1 is approximately a convex optimization problem (P4);

[0094]

[0095] stC8,C10,C13a,C15a,C15b,

[0096] C16a,C18a,C19,C20,C21,C22.

[0097] The convex optimization problem (P4) is solved using CVX.

[0098] In a preferred embodiment of the present invention, in step S4, by fixing the precoding matrix W and the common rate allocation vector r, a second sub-problem of optimizing the offloading ratio β is obtained. This second sub-problem is a convex optimization problem (P5):

[0099]

[0100] The convex optimization problem (P5) is solved using CVX.

[0101] In a preferred embodiment of the present invention, in step S5, the specific steps of the block coordinate descent method are:

[0102] Step S501: Set the number of iterations n = 0, the convergence accuracy threshold ∈ = 10 -6 , given a feasible initial point: precoding matrix W [n] , non-negative auxiliary variable η [n] , uninstall ratio β [n] ;

[0103] Step S502: Given a precoding matrix W [n] , non-negative auxiliary variable η [n] , uninstall ratio β [n] , update the common rate allocation vector r by solving problem (P4) [n+1] , precoding matrix W [n+1] , non-negative auxiliary variable η [n+1] ;

[0104] Step S503: Given a common rate allocation vector r [n+1] , precoding matrix W [n+1] , update the unloading ratio β by solving problem (P5) [n+1] variable;

[0105] Step S504: Execute the operation n←n+1 and calculate the objective function value t [n] ;

[0106] Step S505: Determine the convergence of the target optimization model:

[0107] When the absolute value of the difference between the objective function of two consecutive iterations is greater than the convergence accuracy threshold ∈=10 -6 When , return to step S502;

[0108] When the absolute value of the objective function difference between two consecutive iterations is not greater than the convergence accuracy threshold ∈=10 -6 When , the iteration is exited and the optimal system delay t is obtained [n] And its corresponding optimal public rate allocation vector r [n] , precoding matrix W [n] , uninstall ratio β [n] .

[0109] In a preferred embodiment of the present invention, the number of edge users is single or multiple.

[0110] An electronic device includes a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the RSMA-assisted MEC system secure unloading method.

[0111] A computer-readable storage medium stores a computer program implemented according to the RSMA-assisted MEC system security unloading method in the form of computer-readable instructions. When the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

[0112] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0113] 1. The RSMA-assisted MEC system security offloading method of the present invention can achieve soft bridging by utilizing the splitting of user messages and the non-orthogonal transmission of public messages decoded by multiple users and private messages decoded by corresponding users, thereby reconciling the two extreme interference management strategies of complete decoding interference and treating interference as noise. This ensures secure communication even in the presence of strong internal eavesdropping, and enables all users in the mobile edge computing (MEC) network to share the same time-frequency resources. At the same time, during the transmission process, interference between users can be minimized to further improve the system's latency performance.

[0114] 2. In the target optimization model established by the present invention, it is necessary to continuously adjust the task offloading ratio unique to MEC, so as to adjust the sum of the processing delay of the local calculation of the edge user, the transmission delay offloaded from the edge user to the auxiliary user, and the processing delay of the auxiliary user to process its assigned computing task, so as to make the delay equal to minimize the delay of the system; therefore, in order to further improve the delay performance of the system, it is necessary to continuously adjust the public rate allocation vector and precoding matrix in RSMA, so as to balance the two extreme interference management strategies of fully decoding interference and treating interference as noise, so as to improve the transmission rate and reduce the delay from offloading from edge users to auxiliary users. The transmission delay of the user, the reduction of the transmission delay will change the offloading ratio, thereby offloading more computing tasks to the auxiliary user for calculation; however, due to the broadcast nature of the wireless channel, the computing tasks are easily intercepted by eavesdroppers. Therefore, the present invention takes into account the constraints of secure communication. In order to ensure that the corresponding task segment is decoded by its corresponding auxiliary user, it is necessary to adaptively adjust the parameters of the public rate allocation vector and the precoding matrix. If too much power is allocated to public messages to reduce interference between users, the secure transmission of messages may not be guaranteed; if too much power is allocated to private messages to ensure the secure transmission of information, the delay performance of the system will be reduced. Therefore, the present invention needs to adaptively adjust the task offloading ratio, the public rate allocation vector and the precoding matrix, so as to improve the delay performance of the system while meeting the requirements of secure communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0115] Figure 1 1 is a flow chart of the RSMA-assisted MEC system secure unloading method of the present invention;

[0116] Figure 2 This is a model diagram of a secure MEC system with RSMA assistance and internal eavesdroppers.

[0117] Figure 3 This is the first simulation result graph;

[0118] Figure 4 This is the second simulation result diagram. DETAILED DESCRIPTION

[0119] The present invention will be described in further detail below with reference to the embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0120] Example 1

[0121] See also Figure 1 and Figure 2 The RSMA-assisted MEC system secure uninstallation method of the present invention comprises the following steps:

[0122] S1: Establish a RSMA-assisted secure MEC system model with internal eavesdroppers;

[0123] In this embodiment, it is assumed that the edge user needs to perform a computing task with D bits; let β k ∈[0,1] represents the proportion of computing tasks offloaded to auxiliary user k; β in the computing tasks k The computation task of D bits needs to be offloaded to auxiliary user k for computation, where: remaining The computing task of bits is performed at the edge user; wherein the edge user is equipped with N t Antenna; The auxiliary user is equipped with a single antenna and integrated with an MEC computing module; It is assumed that the task segment offloaded to auxiliary user k must be kept confidential from the other K-1 auxiliary users, that is, each auxiliary user not only decodes the message it needs as a legitimate receiver, but also acts as a potential internal eavesdropper to intercept messages intended for other auxiliary users;

[0124] The message sent by the edge user to auxiliary user k is represented as W k , message W k Including the corresponding assigned computing tasks β k D bits; according to the 1-layer RSMA principle, the message W k Split into a common message W c,k and a private message W p,k ; Among them, the public message W from all auxiliary users c,1 ,...,W c,K Merged into a super public message W c ; Encode it into a common stream x using the codebook shared by all auxiliary users c , all auxiliary users need to decode the common stream x c ; Private messages W from all auxiliary users p,1 ,...,W p,K are encoded separately as private streams x1,...,x K , each private stream is independently decoded only by its corresponding auxiliary user; since private messages are very suitable for secure transmission, the computing tasks that need to be securely transmitted are placed in private messages, while the remaining computing tasks are placed in public messages for transmission;

[0125] Next, define And its linear preprocessing satisfies E(xx H )=I; Therefore, the signal s sent by the edge user is:

[0126]

[0127] Where: W is the precoding matrix, Precoders corresponding to the public stream and the private stream of auxiliary user k respectively;

[0128] The signal received by auxiliary user k is:

[0129]

[0130] Where, represents the channel coefficient between the edge user and the auxiliary user k; z k is the corresponding additive white Gaussian noise with zero mean and variance σ 2 ,Right now

[0131] In this embodiment, K-1 auxiliary users acting as internal eavesdroppers intercept the private message of auxiliary user k in a non-cooperative manner.

[0132] S2: With the goal of minimizing the maximum latency of the secure MEC system model, a target optimization model is established using the common rate allocation vector, precoding matrix, and offload ratio in the secure MEC system model as optimization variables.

[0133] In this embodiment, the specific steps of constructing the target optimization model are:

[0134] Step S201: For auxiliary user k, firstly transfer private stream x k Decode the public stream x as interference c After decoding the public message, auxiliary user k extracts its own part and removes the public message through successive interference cancellation (SIC). After that, auxiliary user k considers the remaining K-1 private streams as interference to decode its own private stream x. k ;therefore,

[0135] Auxiliary user k decodes the public stream x c The achievable rate R ck for:

[0136]

[0137] Assist user k in decoding private stream x k The achievable rate is:

[0138]

[0139] Step S202: Adopting the adaptive rate transmission mode, the actual transmission rate of the public message allocated to the auxiliary user k is set to r k , which is a variable that needs to be optimized later. To ensure that auxiliary user k can successfully decode the public message, the following conditions must be met:

[0140]

[0141] Where: r k is the actual transmission rate of the public message assigned to auxiliary user k;

[0142] Step S203: Calculate the total achievable rate of auxiliary user k:

[0143] R k =r k +R p,k ;

[0144] Step S204: The remaining K-1 auxiliary users eavesdrop on the private message of auxiliary user k by removing their own private flow using an additional layer of ideal SIC. Therefore, the auxiliary user The achievable eavesdropping rate for decoding the private stream of auxiliary user k is:

[0145]

[0146] Step S205: Under the constraints of the strongest eavesdropping capability, assist user k in transmitting the confidentiality rate of the private stream:

[0147]

[0148] Where:

[0149] Step S206: Through the transmission delay of auxiliary user k, the processing delay of edge user local calculation and the processing delay offloaded to auxiliary user k k The total delay is calculated by calculating the processing delay of the D-bit computing task; where

[0150] The transmission delay of auxiliary user k is:

[0151]

[0152] Where: B represents the system bandwidth;

[0153] The processing delay of local computing by edge users is:

[0154]

[0155] Where: φ0 is the number of CPU cycles required for edge users to calculate 1 bit of data; f0 is the computing resources of edge users;

[0156] The computational tasks β offloaded to auxiliary user k k The processing delay of D is:

[0157]

[0158] Where: φ krepresents the number of CPU cycles required to assist the user in calculating 1 bit of data, f k Computing resources to assist users;

[0159] Ignoring the time required to transmit the computation results from the auxiliary users to the edge users, the total delay is:

[0160]

[0161] Step S207: Taking into account the effectiveness and reliability of the communication system, the optimization goal is to minimize the total system delay while ensuring secure transmission from edge users to auxiliary users. The optimization variables include the computing task allocation ratio (i.e., the offload ratio). Common Rate Allocation Vector and the precoding matrix

[0162] Establish the initial target optimization model (P1):

[0163]

[0164] Where: P and They represent the maximum transmit power budget of the edge user and the minimum secure transmission rate threshold of each private message respectively; Constraint C1 requires that each auxiliary user can successfully decode the public message; Constraint C2 requires that the transmission power allocation of the edge user cannot exceed the maximum power budget; Constraint C3 requires that the offloaded computing task cannot exceed the total computing task owned by the edge user; Constraint C4 requires that each private message can be securely offloaded;

[0165] It is important to note that the objective function, constraint C1, and constraint C4 are non-convex. Furthermore, the three optimization variables—the common rate allocation vector, the offloading ratio, and the precoding matrix—are highly coupled in the objective function and the constraints. Therefore, the optimization problem (P1) is a non-convex optimization problem that is generally difficult to solve.

[0166] Then the max function of the total delay in the target optimization model is processed to obtain an equivalent target optimization model (P2):

[0167]

[0168]

[0169]

[0170] C1,C2,C3,C4.

[0171] Without affecting the optimality of the optimization problem (P1), each constraint in constraint C5 and constraint C6 can satisfy the equality condition; if the condition is not satisfied, t can be reduced without improving the objective function value; this also means that the time for the edge user to offload to each auxiliary user k is equal and is also equal to the time required for local calculation by the edge user.

[0172] S3: By fixing the offloading ratio β, we obtain subproblem 1 of jointly optimizing the precoding matrix W and the common rate allocation vector r, and use the continuous convex approximation algorithm (SCA) to approximate subproblem 1 as a convex optimization problem.

[0173] Given the task offloading ratio β, which is considered as a known constant, the first sub-problem of jointly optimizing the precoding matrix W and the public rate allocation r is:

[0174]

[0175] C10:tr(WW H )≤P,

[0176]

[0177] Where: Constraint C7, Constraint C9 and Constraint C11 are non-convex constraints.

[0178] Introducing non-negative auxiliary variables Replace R respectively c,k 、R p,k and Rewrite constraints C7, C9, and C11 as follows:

[0179]

[0180]

[0181]

[0182] Among them, the operator [·] + It has been deleted from the left side of constraint C14, and the two are equivalent; constraint C12a, constraint C12b, constraint C12c, and constraint C13b are all non-convex constraints;

[0183] For constraints C12a, C12b, C12c, and C13b, introduce non-negative auxiliary variables and Instead of (r k +α p,k -α e,k ) and R c,k 、R p,k 、 In the signal-to-interference-plus-noise ratio part, rewrite constraints C12a, C12b, C12c, and C13b as follows:

[0184]

[0185] Where: Constraint C16b, Constraint C17a, Constraint C17b, and Constraint C18b are non-convex constraints, and the other constraints are convex constraints;

[0186] Constraints C16b, C17a, C17b, and C18b are all linearly approximated as convex constraints using first-order Taylor expansion;

[0187] For constraints C16b and C17a, the non-negative variable η p,k ,η c,k Move to the denominator on the right side of the inequality; the right side of the inequality is now Both are quadratic linear fractional forms and are convex functions; since the first-order Taylor expansion of any convex function at any point is its global lower bound, The first-order Taylor expansion is:

[0188]

[0189] Therefore, constraint C16b is linearly approximated as a convex constraint:

[0190]

[0191] Constraint C18b is linearly approximated as a convex constraint:

[0192]

[0193] where the superscript [n] represents the value of the variable during the nth iteration.

[0194] For constraint C17a, the right side of the inequality log2(1+η k,j ) is a concave function; since the first-order Taylor expansion of any concave function at any point is its global upper bound, log2(1+η k,j )The first-order Taylor expansion is:

[0195]

[0196] That is, constraint C17a is linearly approximated as a convex constraint:

[0197]

[0198] For constraint C17b, the non-negative variable η k,j Move to the denominator on the right side of the inequality; for convex functions The first-order Taylor expansion is:

[0199]

[0200] That is, constraint C17b is linearly approximated as a convex constraint:

[0201]

[0202] Finally, subproblem 1 is approximated as a convex optimization problem (P4);

[0203]

[0204] stC8,C10,C13a,C15a,C15b,

[0205] C16a,C18a,C19,C20,C21,C22.

[0206] The convex optimization problem (P4) is solved using CVX.

[0207] S4: By fixing the precoding matrix W and the common rate allocation vector r, the second sub-problem of optimizing the offloading ratio β is obtained. This sub-problem is a convex optimization problem.

[0208] By fixing the precoding matrix W and the common rate allocation vector r, we can obtain the second sub-problem of optimizing the offloading ratio β, which is already a convex optimization problem (P5):

[0209]

[0210] The convex optimization problem (P5) is a linear programming problem (LP) that can be solved efficiently using optimization tools such as CVX.

[0211] S5: Use the block coordinate descent method to alternately iterate the approximated subproblems 1 and 2 until the iterative error of the iterative objective function is less than the threshold ∈, and the optimal delay of the system is obtained; the specific steps are:

[0212] Step S501: Set the number of iterations n = 0, the convergence accuracy threshold ∈ = 10 -6 , given a feasible initial point: precoding matrix W [n] , non-negative auxiliary variable η [n] , uninstall ratio β [n] ;

[0213] Step S502: Given a precoding matrix W [n] , non-negative auxiliary variable η [n] , uninstall ratio β [n] , update the common rate allocation vector r by solving problem (P4) [n+1] , precoding matrix W [n+1], non-negative auxiliary variable η [n+1] ;

[0214] Step S503: Given a common rate allocation vector r [n+1] , precoding matrix W [n+1] , update the unloading ratio β by solving problem (P5) [n+1] variable;

[0215] Step S504: Execute the operation n←n+1 and calculate the objective function value t [n] ;

[0216] Step S505: Determine the convergence of the target optimization model:

[0217] When the absolute value of the difference between the objective function of two consecutive iterations is greater than the convergence accuracy threshold ∈=10 -6 When , return to step S502;

[0218] When the absolute value of the objective function difference between two consecutive iterations is not greater than the convergence accuracy threshold ∈=10 -6 When , the iteration is exited and the optimal system delay t is obtained [n] And its corresponding optimal public rate allocation vector r [n] , precoding matrix W [n] , uninstall ratio β [n] .

[0219] In this embodiment, the edge user may be single or multiple.

[0220] In order to verify the performance of the RSMA-assisted MEC system security offloading method of the present invention, the following four benchmark schemes are compared with the RSMA-assisted MEC system security offloading method of the present invention, and their performance is evaluated using numerical results:

[0221] (1) SDMA-partial: Edge users partially offload their computational tasks, and each auxiliary user handles a portion of the tasks, using SDMA for secure transmission;

[0222] (2) RSMA-full: The edge user completely offloads its entire computing task, and each auxiliary user handles a part of the task, using RSMA for secure transmission;

[0223] (3) SDMA-full: The edge user completely offloads its entire computing task, and each auxiliary user handles a part of the task, using SDMA for secure transmission;

[0224] (4) Local-computing: The computing task is completed by the edge user alone.

[0225] In the simulations, the following parameters were used unless otherwise stated:

[0226] The average power gain of the channel uses the path loss model ρ(d / d0) -λ Modeling, where ρ = -40dB represents the path loss at the reference distance d0 = 1m, d represents the distance between the auxiliary user and the edge user, and λ = 3 represents the path loss exponent; set K = 4, Nt = 4, P = 100mW, B = 1MHz, σ 2 =-110dBm, D = 10Mbits, φ k =φ0=100CPU / bit,f k =f0=1GHz,∈=10 -6 , Considering that the angular spacing between adjacent auxiliary users and edge users is 10 degrees, and the distance between each auxiliary user and the edge user is d = 75m; the simulation results are as follows Figure 3 and Figure 4 shown.

[0227] See also Figure 3 , Figure 3 The relationship between the total system delay and the maximum transmission power P of the edge user is given in.

[0228] It can be seen that in the "Local-computing" scheme, the total delay is independent of the maximum transmission power P at the edge user; on the contrary, for all other strategies, the total delay decreases with the increase of the transmission power P at the edge user; compared with SDMA, the RSMA network allows the transmission power of public messages to be optimized while adjusting the transmission power of private messages to reduce interference, thereby significantly reducing the total delay of the system; compared with full offloading, partial offloading reduces the computational burden of auxiliary users by leaving part of the computing tasks to be processed locally, thereby reducing the delay of the entire system.

[0229] See also Figure 4 , Figure 4 The relationship between the total system delay and the computing task volume D is given in .

[0230] For all schemes, a larger computational workload corresponds to longer system latency. As the computational workload increases, the latency gap between the RSMA-partial and SDMA-partial schemes widens. This is because the dramatic increase in the total computational workload leads to a larger difference in the latency required for offloading to the auxiliary user for processing between the two schemes. Similarly, this also contributes to the latency performance gap between RSMA-full and SDMA-full. Due to the additional computing resources of edge users and the lack of transmission latency, the latency of partial offloading is always better than that of its full offloading counterpart.

[0231] Example 2

[0232] This embodiment differs from Embodiment 1 in that: the K-1 auxiliary users acting as internal eavesdroppers collude with each other to intercept the private messages of auxiliary user k, and they can collude perfectly; it is assumed that a central node integrates all the information collected by the colluding eavesdroppers, and the K-1 colluding eavesdroppers can be effectively regarded as a single eavesdropper equipped with K-1 distributed antennas; that is, the K-1 colluding eavesdroppers can utilize maximum ratio combining (MRC) to enhance the signal interception capability; it is also assumed that the colluding eavesdroppers can decode each other's private messages; in other words, the private message transmitted to one colluding eavesdropper will not interfere with the communications of other colluding eavesdroppers.

[0233] In this embodiment, the steps for constructing the target optimization model are as follows:

[0234] Step S201: Calculate the auxiliary user k’s decoding of the public stream x c and private stream x k The achievable rate is

[0235] Auxiliary user k decodes the public stream x c The achievable rate R ck for:

[0236]

[0237] Assist user k in decoding private stream x k The achievable rate is:

[0238]

[0239] Step S202: To ensure that all auxiliary users can successfully decode the super public message W c , the following conditions must be met:

[0240]

[0241] r k ≥0;

[0242] Where: r k is the actual transmission rate of the public message assigned to auxiliary user k;

[0243] Step S203: Calculate the total achievable rate of auxiliary user k:

[0244] R k =r k +R p,k ;

[0245] Step S204: Calculate the achievable eavesdropping rate of the other K-1 auxiliary users attempting to collaboratively eavesdrop on the private flow of auxiliary user k:

[0246]

[0247] Step S205: Calculate the confidentiality rate of the private stream transmitted by auxiliary user k:

[0248] R s,k =[R p,k -R k,e ] + ;

[0249] Where:

[0250] Step S206: Through the transmission delay of auxiliary user k, the processing delay of edge user local calculation and the processing delay offloaded to auxiliary user k k The total delay is calculated by calculating the processing delay of the D-bit computing task; where

[0251] The transmission delay of auxiliary user k is:

[0252]

[0253] Where: B represents the system bandwidth;

[0254] The processing delay of local computing by edge users is:

[0255]

[0256] Where: φ0 is the number of CPU cycles required for edge users to calculate 1 bit of data; f0 is the computing resources of edge users;

[0257] The computational tasks β offloaded to auxiliary user k k The processing delay of D is:

[0258]

[0259] Where: φ k represents the number of CPU cycles required to assist the user in calculating 1 bit of data, f k Computing resources to assist users;

[0260] The total delay is:

[0261]

[0262] Step S207: Establishing an initial target optimization model (P1):

[0263]

[0264] C2:tr(WW H )≤P,

[0265]

[0266] Where: P and They represent the maximum transmission power budget of the edge user and the minimum secure transmission rate threshold of each private message respectively; constraint C1 requires that each auxiliary user can successfully decode the public message; constraint C2 requires that the transmission power allocation of the edge user cannot exceed the maximum power budget; constraint C3 requires that the offloaded computing tasks cannot exceed the total computing tasks owned by the edge user; constraint C4 requires that each private message can be securely offloaded.

[0267] In this embodiment, the solution process of the above-mentioned target optimization model can refer to Example 1.

[0268] Example 3

[0269] The electronic device of the present invention includes a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the RSMA-assisted MEC system secure unloading method.

[0270] Example 4

[0271] The computer-readable storage medium of the present invention stores a computer program implemented according to the RSMA-assisted MEC system security unloading method in the form of computer-readable instructions. When the computer program is called and executed by the computer, the steps included in the corresponding method are executed.

[0272] The above is a preferred embodiment of the present invention, but the embodiment of the present invention is not limited to the above content. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A method for RSMA-assisted MEC system secure unloading, characterized in that: The following steps are involved: S1: Establishing a RSMA-assisted secure MEC system model, wherein the secure MEC system model contains an internal eavesdropper; S2: With the goal of minimizing the maximum delay of the secure MEC system model, the public rate allocation vector r, precoding matrix W, and offload ratio β in the secure MEC system model are used as optimization variables to establish a target optimization model; S3: By fixing the offloading ratio β, we obtain subproblem 1 of jointly optimizing the precoding matrix W and the common rate allocation vector r. We then use the continuous convex approximation algorithm to approximate subproblem 1 as a convex optimization problem. S4: By fixing the precoding matrix W and the common rate allocation vector r, the second sub-problem of optimizing the offloading ratio β is obtained. This sub-problem is a convex optimization problem. S5: Use the block coordinate descent method to alternately iterate the approximate sub-problem 1 in step S3 and the sub-problem 2 in step S4 until the iterative error of the iterative objective function is less than the threshold Get the optimal system delay.

2. The RSMA-assisted MEC system secure uninstallation method according to claim 1, characterized in that: In step S1, the steps for constructing the secure MEC system model are: Step S101: Assume that the edge user needs to perform a computing task with D bits; where β k The computation task of D bits needs to be offloaded to auxiliary user k for computation. remaining The computing task of bits is performed at the edge user; wherein the edge user is equipped with N t Antenna; The auxiliary user is equipped with a single antenna and integrated with an MEC computing module; It is assumed that the computing task segment offloaded to auxiliary user k must be kept confidential from the other K-1 auxiliary users, that is, each auxiliary user not only decodes the message it needs as a legitimate receiver, but also intercepts the message intended to be sent to other auxiliary users as a potential internal eavesdropper; Step S102: The message sent by the edge user to the auxiliary user k is represented as W k , message W k Including the corresponding assigned computing tasks β k D bits; the message W k Split into a common message W c，k and a private message W p，k ; Among them, the public message W from all auxiliary users c，1 ,...,W c,K Merged into a super public message W c ; Use the codebook shared by all auxiliary users to convert the super public message W c Encoded as a public stream x c , all auxiliary users need to decode the common stream x c ; Private messages W from all auxiliary users p，1 ,...,W p，K are encoded separately as private streams x1, ..., x K， Each private stream is decoded independently only by its corresponding auxiliary user; Step S103: Place the computing tasks that need to be securely transmitted in private messages, and place the remaining computing tasks in public messages for transmission; Step S104: Definition And its linear preprocessing satisfies E(xx H )=I;wherein, The signal sent by the edge user is: Where: W is the precoding matrix, Precoders corresponding to the public stream and the private stream of auxiliary user k respectively; The signal received by auxiliary user k is: Where, represents the channel coefficient between the edge user and the auxiliary user k; z k is the corresponding additive white Gaussian noise with zero mean and variance σ 2 ,Right now 3. The RSMA-assisted MEC system secure uninstallation method according to claim 2, characterized in that: In step S1, K-1 auxiliary users acting as internal eavesdroppers intercept the private message of auxiliary user k in a non-cooperative manner, or the K-1 auxiliary users acting as internal eavesdroppers collude with each other to intercept the private message of auxiliary user k.

4. The RSMA-assisted MEC system secure uninstallation method according to claim 3 is characterized in that: In step S2, K-1 auxiliary users acting as internal eavesdroppers intercept the private message of auxiliary user k in a non-cooperative manner. The steps for constructing the target optimization model are as follows: Step S201: Calculate the auxiliary user k’s decoding of the public stream x c and private stream x k The achievable rate is Auxiliary user k decodes the public stream x c The achievable rate R c，k for: Assist user k in decoding private stream x k The achievable rate is: Step S202: To ensure that all auxiliary users can successfully decode the super public message W c , the following conditions must be met: r k ≥0； Where: r k is the actual transmission rate of the public message assigned to auxiliary user k; Step S203: Calculate the total achievable rate of auxiliary user k: R k =r k +R p，k ; Step S204: Calculate the achievable eavesdropping rate of auxiliary user j decoding the private stream of auxiliary user k: Step S205: Calculate the confidentiality rate of the private stream transmitted by auxiliary user k: Where: Step S206: Through the transmission delay of auxiliary user k, the processing delay of edge user local calculation and the processing delay offloaded to auxiliary user k k The total delay is calculated by calculating the processing delay of the D-bit computing task; where The transmission delay of auxiliary user k is: Where: B represents the system bandwidth; The processing delay of local computing by edge users is: Where: The number of CPU cycles required to calculate 1 bit of data for the edge user; f0 is the computing resources of the edge user; The computational tasks β offloaded to auxiliary user k k The processing delay of D is: Where: represents the number of CPU cycles required to assist the user in calculating 1 bit of data, f k Computing resources to assist users; The total delay is: Step S207: Establishing an initial target optimization model (P1): C2:tr(WW H )≤P, Then the max function of the total delay in the target optimization model is processed to obtain an equivalent target optimization model (P2): C1,C2,C3,C4 Where: P and They represent the maximum transmission power budget of the edge user and the minimum secure transmission rate threshold of each private message respectively; constraint C1 requires that each auxiliary user can successfully decode the public message; constraint C2 requires that the transmission power allocation of the edge user cannot exceed the maximum transmission power budget; constraint C3 requires that the offloaded computing tasks cannot exceed the total computing tasks owned by the edge user; constraint C4 requires that each private message can be securely offloaded.

5. The RSMA-assisted MEC system secure uninstallation method according to claim 4 is characterized in that: In step S3, given the task offloading ratio β, the first sub-problem of jointly optimizing the precoding matrix W and the common rate allocation vector r is: C10:tr(WW H )≤P, Where: Constraint C7, Constraint C9 and Constraint C11 are non-convex constraints.

6. The RSMA-assisted MEC system secure uninstallation method according to claim 5, characterized in that: The steps to approximate subproblem 1 to a convex optimization problem using the continuous convex approximation algorithm are as follows: Step S301: Introduce non-negative auxiliary variables Replace R respectively c，k 、R p，k and Rewrite constraints C7, C9, and C11 as follows: Where: Constraint C12a, Constraint C12b, Constraint C12c, and Constraint C13b are non-convex constraints; Introducing non-negative auxiliary variables and Instead of (r k +α p，k -α e，k ) and R c，k 、R p，k 、 In the signal-to-interference-plus-noise ratio part, rewrite constraints C12a, C12b, C12c, and C13b as follows: Where: Constraint C16b, Constraint C17a, Constraint C17b, and Constraint C18b are non-convex constraints, and the other constraints are convex constraints; Step S302: Constraints C16b, C17a, C17b, and C18b are linearly approximated as convex constraints using first-order Taylor expansion; wherein, Constraint C16b is linearly approximated as a convex constraint: Where: Constraint C18b is linearly approximated as a convex constraint: Where: Constraint C17a is linearly approximated as a convex constraint: Where: Constraint C17b is linearly approximated as a convex constraint: Where: Here, the superscript [n] represents the value of the variable during the nth iteration; Step S303: Subproblem 1 is approximately a convex optimization problem (P4); st C8,C10,C13a,C15a,C15b, C16a, C18a, C19, C20, C21, C22 The convex optimization problem (P4) is solved using CVX.

7. The RSMA-assisted MEC system secure uninstallation method according to claim 6, characterized in that: In step S4, by fixing the precoding matrix W and the common rate allocation vector r, subproblem 2 for optimizing the offloading ratio β is obtained. Subproblem 2 is a convex optimization problem (P5): The convex optimization problem (P5) is solved using CVX.

8. The RSMA-assisted MEC system secure uninstallation method according to claim 7, characterized in that: In step S5, the specific steps of the block coordinate descent method are: Step S501: Set the number of iterations n = 0, the convergence accuracy threshold Given a feasible initial point: precoding matrix W [n] , non-negative auxiliary variable η [n] , uninstall ratio β [n] ; Step S502: Given a precoding matrix W [n] , non-negative auxiliary variable η [n] , uninstall ratio β [n] , update the common rate allocation vector r by solving problem (P4) [n+1] , precoding matrix W [n+1] , non-negative auxiliary variable η [n+1] ; Step S503: Given a common rate allocation vector r [n+1] , precoding matrix W [n+1] , update the unloading ratio β by solving problem (P5) [n+1] variable; Step S504: Execute the operation n←n+1 and calculate the objective function value t [n] ; Step S505: Determine the convergence of the target optimization model: When the absolute value of the difference between the objective function of two consecutive iterations is greater than the convergence accuracy threshold When , return to step S502; When the absolute value of the objective function difference between two consecutive iterations is not greater than the convergence accuracy threshold When , the iteration is exited and the optimal system delay t is obtained [n] And its corresponding optimal public rate allocation vector r [n] , precoding matrix W [n] , uninstall ratio β [n] .

9. An electronic device comprising a central processing unit and a memory, characterized in that: The central processing unit is used to call and run the computer program stored in the memory to perform the steps of the RSMA-assisted MEC system secure unloading method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that It stores a computer program implemented by the RSMA-assisted MEC system security uninstallation method according to any one of claims 1 to 8 in the form of computer-readable instructions. When the computer program is called and executed by the computer, the steps included in the corresponding method are executed.

Images