A data processing method combining security and energy load balancing

Abstract

The present invention discloses a data processing method for joint security and energy load balancing, comprising the following steps: Each user terminal sends the allocated pilot sequence to the access point AP-ES, where the access point AP-ES is an access point AP equipped with an independent edge server ES; The access point AP-ES determines whether the user terminal UE is attacked by the attacker Eve by detecting abnormal conditions of the pilot signal; If the user terminal UE is not attacked, it is marked as a normal user terminal UE, otherwise, it is marked as a user terminal UE under attack; The access point AP-ES, as an agent, respectively executes the current service policy H<supgt;*< / supgt> on their respective local observations; Based on the result of the current service policy H<supgt;*< / supgt>, the user terminal UE calculates the optimal task offloading ratio Λ<supgt;*< / supgt>, and obtains the optimal interference strategy J<supgt;*< / supgt> based on the service policy H<supgt;*< / supgt> and the optimal task offloading ratio Λ<supgt;*< / supgt>. A data processing method based on security and load balancing is proposed, which can achieve an efficient and secure transmission scheme against uplink passive eavesdropping and downlink pilot contamination attacks.

Description

Technical Field

[0001] The present invention relates to the field of network security, and in particular to a data processing method for combining security and energy load balancing. Background Art

[0002] At present, the research on computational offloading under the non-cellular massive MIMO (multiple-input-multiple-output) architecture mainly focuses on three aspects: minimizing energy consumption, optimizing latency, and allocating computing resources. These works are all committed to solving the problem of how to ensure the QoS (quality of experience) of UE (user equipment) in load calculation, and the load balancing performance of edge servers in the network is not fully considered. The load imbalance of edge servers will significantly affect the service quality and reduce the service life of edge servers.

[0003] In the non-cellular massive MIMO scenario, data security threats are a key issue in the task offloading process. Considering the significant directional gain enhancement brought by massive MIMO, it is feasible to achieve high-precision beamforming in the non-cellular massive MIMO scenario. However, active eavesdroppers can send spoofed pilot sequences to reduce the downlink beamforming accuracy, which is called PCA (pilot contamination attack). As a supplement to traditional cryptography, PLS (physical layer security) aims to exploit the characteristics and defects of wireless channels, including noise, fading, interference, dispersion, and diversity. Therefore, to solve this problem, PLS technology has gradually become one of the key technologies to resist the active attacker Eve. In addition, in order to counter the passive eavesdropping of the attacker Eve, existing research mainly focuses on CJ (cooperative jamming). This technology involves deploying a friendly jammer to selectively interfere with the attacker Eve, thereby preventing data leakage by transmitting jamming signals targeting the attacker Eve. While physical layer secure transmission has been well studied, simultaneous joint transmission of uplink and downlink guarantees for computational offloading in cell-free massive MIMO networks has not been fully explored. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a data processing method combining security and energy load balancing, which jointly considers network load balancing performance and service quality issues of user terminals.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] A data processing method for joint security and energy load balancing is provided, which comprises the following steps:

[0007] S1: Each user terminal sends the assigned pilot sequence to the access point AP-ES, which is an access point AP equipped with an independent edge server ES; the access point AP-ES determines whether the user terminal UE is attacked by the attacker Eve by detecting abnormal conditions of the pilot signal;

[0008] S2: If the user terminal UE is not attacked, it is marked as a normal user terminal UE, otherwise, it is marked as a user terminal UE under attack;

[0009] For the user terminal UE under attack, determine whether the arrival angle between the user terminal UE and the access point AP-ES meets the security transmission condition. If so, the access point AP-ES maintains the connection with the user terminal UE under attack, otherwise, disconnects the connection;

[0010] S3: Access points AP-ES act as intelligent agents and execute the current service strategy H on their respective local observations. * ;

[0011] S4: Based on the current service strategy H * As a result, the user terminal UE calculates the optimal task offloading ratio Λ * , and based on the service strategy H * and the optimal task offloading ratio Λ * Get the optimal interference strategy J * .

[0012] Furthermore, the method for screening normal user terminals UE in step S2 is:

[0013]

[0014] Wherein, A is a set of access points AP-ES in the network system, A={1,2,…,L}, l′ is the number of the access point AP-ES in the set A, and L is the total number of access points AP-ES in the set A;

[0015] The large-scale fading coefficient β of the channel between all access points AP-ES and user terminals UE lk Sort in descending order to get the set after descending sorting After sorting the collection in descending order The elements in After sorting the collection in descending order The minimum number of elements in l is the set after sorting in descending order The number of the corresponding access point AP-ES;

[0016] C U U p is the subset of user terminals UE that have not been actively polluted by the attacker Eve, U p is the user terminal UE subset that Eve takes active pilot pollution attack on, and p is the subset U p The number of the user terminal UE in β l′k is the large-scale fading coefficient of the channel between the l′th access point and the kth user terminal UE;

[0017] 0<δ<1, δ is a predefined threshold for screening whether the user terminal UE is under attack, k is the number of the user terminal UE in the network system, A subset of user terminals UE that are attacked;

[0018] The method for screening the user terminal UE under attack is:

[0019]

[0020] Furthermore, in step S2, the method for determining whether the access point AP-ES maintains the connection with the user terminal UE under attack is:

[0021] The access point AP-ES obtains the arrival angle information of the user terminal UE and the attacker Eve respectively, and the access point AP-ES distinguishes the user terminal UE and the attacker Eve by transmitting a narrow directional beam;

[0022] Maximize the difference between the narrow directional beam from the access point AP-ES to the user terminal UE and the narrow directional beam from the access point AP-ES to the attacker Eve, thereby eliminating the attacker Eve and aligning the beam to the user terminal UE;

[0023] The antenna of the access point AP-ES adopts an ideal sector-based antenna pattern model, so that the kth user terminal UE is in the main lobe area of ​​the antenna, and the attacker Eve is outside the main lobe area of ​​the antenna. The lth access point AP-ES can provide services to the kth user terminal UE and meet the following conditions:

[0024]

[0025] Among them, θ lk is the arrival angle from the access point AP-ES to the user terminal UE, θ le is the arrival angle from access point AP-ES to attacker Eve, is the beam width of beamforming at the lth access point AP-ES.

[0026] Furthermore, the optimal task offloading ratio Λ in step S4* The calculation method is:

[0027] S41: For a given service policy H * and the optimal jamming strategy J * , derive the feasible solution of the offloading ratio Λ of the user terminal UE;

[0028] The objective function of the task offloading decision of the user terminal UE is:

[0029] The constraints of the objective function of task offloading decision are:

[0030]

[0031] in, is the total energy consumption of the kth user terminal UE, is the maximum energy consumption of the kth user terminal UE, is the energy consumption calculated locally by the user terminal UE, λ k is the uninstall rate of the user terminal UE, is the uninstallation time of the user terminal UE, is the time calculated locally by the user terminal UE, is the computing time of the edge server ES, is the time delay limit of the kth user terminal UE, P t k is the transmission power of the kth user terminal UE;

[0032] S42: Constraint Transformed into two linear constraints:

[0033] S43: Constraints and Merge to get:

[0034]

[0035] Among them, a k is the energy consumption during task offloading, b k The energy consumption generated by the local computing process, P t k is the transmit power of the user terminal UE, T k is the total task calculated at the user terminal UE, is the uplink confidentiality transmission rate in uplink transmission, is the transmission rate in the uplink transmission that is not attacked by the attacker Eve, is the effective capacitance coefficient of the CPU architecture of the user terminal UE, C k The number of CPU cycles required for the user terminal UE to calculate the task, f k The frequency of the CPU architecture of the user terminal UE;

[0036] S44: Energy consumption generated by the local calculation process of the user terminal UE k Greater than maximum energy consumption Right now Energy consumption generated during task offloading of user terminal UE k Less than the energy consumption of the local computing process b k , that is, a k k ,get:

[0037]

[0038] S45: Constructing the second-order partial derivative of the objective function And the objective function E LIB It is a convex problem, which can be derived;

[0039]

[0040] Among them, m le is the association element between the access point AP-ES and the attacker Eve, p lk is the equivalent substitution parameter of the second-order partial derivative process, η lk It is an association element between the access point AP-ES and the user terminal UE. is the effective capacitance coefficient of the CPU architecture of the access point AP-ES, C l The number of CPU cycles required for the AP-ES calculation task, f l k The frequency of the CPU architecture of the access point AP-ES. is the transmit power of the access point AP-ES;

[0041] S45: Then obtain the optimal task offloading ratio Λ * For the convex problem of task offloading ratio Λ, the CVX optimization tool is used to optimize the objective function Solve and get the optimal task offloading ratio Λ * .

[0042] Furthermore, the service strategy H * The optimization method is:

[0043] S31: For a given optimal task offloading ratio Λ * and the optimal jamming strategy J * , establish and calculate the optimal service strategy H * ​The objective function is: Optimal service strategy H * The association between the user terminal UE and the access point AP-ES is represented by the derived matching matrix H=L×K;

[0044] Objective Function The constraints are:

[0045]

[0046] Among them, η lk An element that associates the user terminal UE with the access point AP-ES, is the downlink confidentiality transmission rate that can be achieved in downlink transmission, The uplink confidentiality transmission rate that can be achieved in uplink transmission, R s ,R u are the transmission rate thresholds of the downlink transmission link and the uplink transmission link, respectively, and c is the subset The number of the user terminal UE, f l,k is the computing resource allocated by the lth access point AP-ES to the kth user terminal UE associated with it, and ε is the constraint value of the total energy consumption of K user terminals UE;

[0047] S32: Define thresholds As the lower bound of the achievable confidentiality transmission rate in uplink transmission, the constraint and Carry out a merger;

[0048]

[0049] in, is the optimal offloading strategy for the kth user terminal UE, T k is the total task of the kth user terminal UE, C c The number of CPU cycles required for the access point AP-ES;

[0050] S33: The lower limit of the achievable confidentiality transmission rate in uplink transmission is Under the conditions, The maximum tolerable delay is greater than the lower limit Then update the objective function The constraints are:

[0051]

[0052] S34: Construct Markov decision (S, D, P, R, γ), S is the state space, D is the action space, P is the penalty space, R is the immediate reward, γ is the discount factor, the access point AP-ES interacts with the mobile edge network as an intelligent agent, learns the optimal matching relationship between AP-ES and the user terminal UE, based on the environment of the MADRL algorithm, the state at time t The energy consumption of each access point AP-ES and calculation delay Decision; the state space of each agent is:

[0053]

[0054] S35: Each agent’s action is defined as The agent selects an action for each time t. The agent decides the user terminal UE to provide service based on the current state. The action space of each agent is:

[0055]

[0056] in, is the element associated with the user terminal UE and the access point AP-ES at time t;

[0057] S36: Since each agent needs to decide which user terminal UE to provide service to, the dimension of each agent action space D is 2 K , the immediate reward of the lth agent is recorded as

[0058]

[0059] Among them, p i is the penalty factor, i is the number of the penalty factor;

[0060] S37: Setting the service policy H of the access point AP-ES according to the penalty factor * ;

[0061]

[0062] Furthermore, the optimal interference strategy J * The calculation method is:

[0063] A41: For a given service policy H * and the optimal task offloading ratio Λ * , the objective function of the interference strategy J is: The constraints of the objective function of the interference strategy J are:

[0064]

[0065] A42: Select a set J of candidate access points AP-ES based on the energy load imbalance degree according to a predefined threshold δ c , set J c It contains several interference strategies J of candidate access points AP-ES, each interference strategy The selection method is:

[0066]

[0067] Among them, e is the set J c The number of the interference strategy in the

[0068] A43: Set J c Each interference strategy Substitute into the objective function of the interference strategy J and output the optimal interference strategy J according to the constraints * , so that the attacker Eve is located in the main lobe area of ​​the artificial noise beam transmitted by the access point AP-ES, and the user terminal UE is located in the side lobe area.

[0069] The beneficial effects of the present invention are as follows: in view of the particularity of the distribution of non-cellular large-scale MIMO network systems, the present invention proposes a data processing method based on security and load balancing under its architecture, which can realize an efficient and secure transmission solution against uplink passive eavesdropping and downlink pilot pollution attacks.

[0070] Using the complete offloading of the SAP algorithm, each user terminal UE is considered to load all tasks to the access point AP-ES to ensure the minimum energy consumption of the user terminal UE to perform tasks locally; the security of downlink transmission is guaranteed, and the optimized interference strategy is adopted to ensure the security of uplink transmission.

[0071] The present invention mainly considers the optimization of user terminal UE task offloading decision, interference strategy optimization and access point AP-ES service selection optimization. The service selection of access point AP-ES no longer considers the security of user terminal UE in downlink transmission. Access point AP-ES only selects user terminal UE to provide service according to the current load status.

[0072] The present invention also considers that each user terminal UE offloads tasks to the access point AP-ES with a specific probability, and performs numerical optimization to minimize e-LoBaR (energy-based load imbalance), thereby ensuring the security of uplink and downlink transmissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 Flowchart of a data processing method for joint security and energy load balancing.

[0074] Figure 2 This is the schematic diagram of the SPA algorithm. DETAILED DESCRIPTION

[0075] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0076] like Figure 1 As shown, a data processing method for combining security and energy load balancing comprises the following steps:

[0077] S1: Each user terminal sends the assigned pilot sequence to the access point AP-ES, which is an access point AP equipped with an independent edge server ES; the access point AP-ES determines whether the user terminal UE is attacked by the attacker Eve by detecting abnormal conditions of the pilot signal.

[0078] S2: If the user terminal UE is not attacked, it is marked as a normal user terminal UE; otherwise, it is marked as a user terminal UE under attack.

[0079] For the user terminal UE under attack, it is determined whether the arrival angle between the user terminal UE and the access point AP-ES meets the security transmission condition. If so, the access point AP-ES maintains the connection with the user terminal UE under attack, otherwise, the connection is disconnected.

[0080] like Figure 2 As shown, steps S1 and S2 are performed based on the SAP algorithm. Since the number of access points AP-ES in the network is greater than the number of user terminals UE, the present invention considers finding the access point AP-ES with better anti-attacker Eve security performance in the downlink data transmission process from the perspective of spatial position through the difference in the angle of arrival AOA. Since the attacker Eve has limited interference capability, the attacker UE in the network is divided into two categories: normal and attacked.

[0081] The method for screening normal user terminals UE in step S2 is:

[0082]

[0083] Wherein, A is a set of access points AP-ES in the network system, A={1,2,…,L}, l′ is the number of the access point AP-ES in the set A, and L is the total number of access points AP-ES in the set A;

[0084] The large-scale fading coefficient β of the channel between all access points AP-ES and user terminals UE lkSort in descending order to get the set after descending sorting After sorting the collection in descending order The elements in After sorting the collection in descending order The minimum number of elements in l is the set after sorting in descending order The number of the corresponding access point AP-ES;

[0085] C U U p is the subset of user terminals UE that have not been actively polluted by the attacker Eve, U p is the user terminal UE subset that Eve takes active pilot pollution attack on, and p is the subset U p The number of the user terminal UE in β l′k is the large-scale fading coefficient of the channel between the l′th access point and the kth user terminal UE;

[0086] 0<δ<1, δ is a predefined threshold for screening whether the user terminal UE is under attack, k is the number of the user terminal UE in the network system, A subset of user terminals UE that are attacked;

[0087] The method for screening the user terminal UE under attack is:

[0088]

[0089] Since the access point AP-ES can obtain the arrival angle AOA information of the user terminal UE and the attacker Eve respectively, the access point AP-ES can distinguish the user terminal UE from the attacker Eve by emitting a narrow directional beam. Therefore, the main idea of ​​the SAP algorithm is to select a suitable access point AP-ES so that the narrow directional beam from the access point AP-ES to the user terminal UE is significantly different from the narrow directional beam of the attacker Eve, so as to achieve beam alignment to the user terminal UE while excluding the attacker Eve.

[0090] The method for determining whether the access point AP-ES maintains the connection with the attacked user terminal UE in step S2 is:

[0091] The access point AP-ES obtains the arrival angle information of the user terminal UE and the attacker Eve respectively, and the access point AP-ES distinguishes the user terminal UE and the attacker Eve by transmitting a narrow directional beam;

[0092] Maximize the difference between the narrow directional beam from the access point AP-ES to the user terminal UE and the narrow directional beam from the access point AP-ES to the attacker Eve, thereby eliminating the attacker Eve and aligning the beam to the user terminal UE;

[0093] The antenna of the access point AP-ES adopts an ideal sector-based antenna pattern model, so that the kth user terminal UE is in the main lobe area of ​​the antenna, and the attacker Eve is outside the main lobe area of ​​the antenna. The lth access point AP-ES can provide services to the kth user terminal UE and meet the following conditions:

[0094]

[0095] Among them, θ lk is the arrival angle from the access point AP-ES to the user terminal UE, θ le is the arrival angle from access point AP-ES to attacker Eve, is the beam width of beamforming at the lth access point AP-ES.

[0096] S3: Access points AP-ES act as intelligent agents and execute the current service strategy H on their respective local observations. * .

[0097] S4: Based on the current service strategy H * As a result, the user terminal UE calculates the optimal task offloading ratio Λ * , and based on the service strategy H * and the optimal task offloading ratio Λ * Get the optimal interference strategy J * .

[0098] The optimal task offloading ratio Λ in step S4 * The calculation method is:

[0099] S41: For a given service policy H * and the optimal jamming strategy J * , derive the feasible solution of the offloading ratio Λ of the user terminal UE;

[0100] The objective function of the task offloading decision of the user terminal UE is:

[0101] The constraints of the objective function of task offloading decision are:

[0102]

[0103] in, is the total energy consumption of the kth user terminal UE, is the maximum energy consumption of the kth user terminal UE, is the energy consumption calculated locally by the user terminal UE, λ k is the uninstall rate of the user terminal UE, is the uninstallation time of the user terminal UE, is the time calculated locally by the user terminal UE, is the computing time of the edge server ES, is the time delay limit of the kth user terminal UE, P t k is the transmission power of the kth user terminal UE;

[0104] S42: Constraint Transformed into two linear constraints:

[0105] S43: Constraints and Merge to get:

[0106]

[0107] Among them, a k is the energy consumption during task offloading, b k The energy consumption generated by the local computing process, P t k is the transmit power of the user terminal UE, T k is the total task calculated at the user terminal UE, is the uplink confidentiality transmission rate in uplink transmission, is the transmission rate in the uplink transmission that is not attacked by the attacker Eve, is the effective capacitance coefficient of the CPU architecture of the user terminal UE, C k The number of CPU cycles required for the user terminal UE to calculate the task, f k The frequency of the CPU architecture of the user terminal UE;

[0108] S44: Energy consumption generated by the local calculation process of the user terminal UE k Greater than maximum energy consumption Right now Energy consumption generated during task offloading of user terminal UE k Less than the energy consumption of the local computing process b k , that is, a k k ,get:

[0109]

[0110] S45: Constructing the second-order partial derivative of the objective function​ And the objective function E LIB It is a convex problem, which can be derived;

[0111]

[0112] Among them, m le is the association element between the access point AP-ES and the attacker Eve, p lk is the equivalent substitution parameter of the second-order partial derivative process, η lk It is an association element between the access point AP-ES and the user terminal UE. is the effective capacitance coefficient of the CPU architecture of the access point AP-ES, C l The number of CPU cycles required for the AP-ES calculation task, f l k The frequency of the CPU architecture of the access point AP-ES. is the transmit power of the access point AP-ES;

[0113] S45: Then obtain the optimal task offloading ratio Λ * For the convex problem of task offloading ratio Λ, the CVX optimization tool is used to optimize the objective function Solve and get the optimal task offloading ratio Λ * .

[0114] Service Strategy * The optimization method is:

[0115] S31: For a given optimal task offloading ratio Λ * and the optimal jamming strategy J * , establish and calculate the optimal service strategy H * The objective function is: Optimal service strategy H * The association between the user terminal UE and the access point AP-ES is represented by the derived matching matrix H=L×K;

[0116] Objective Function The constraints are:

[0117]

[0118] Among them, η lk An element that associates the user terminal UE with the access point AP-ES, is the downlink confidentiality transmission rate that can be achieved in downlink transmission, The uplink confidentiality transmission rate that can be achieved in uplink transmission, R s ,R u are the transmission rate thresholds of the downlink transmission link and the uplink transmission link, respectively, and c is the subset The number of the user terminal UE, f l,k is the computing resource allocated by the lth access point AP-ES to the kth user terminal UE associated with it, and ε is the constraint value of the total energy consumption of K user terminals UE;

[0119] S32: Define thresholds As the lower bound of the achievable confidentiality transmission rate in uplink transmission, the constraint and Carry out a merger;

[0120]

[0121] in, is the optimal offloading strategy for the kth user terminal UE, T k is the total task of the kth user terminal UE, C c The number of CPU cycles required for the access point AP-ES;

[0122] S33: The lower limit of the achievable confidentiality transmission rate in uplink transmission is Under the conditions, The maximum tolerable delay is greater than the lower limit Then update the objective function The constraints are:

[0123]

[0124] It should be noted that since the objective function Complex, Service Strategy H * The optimization of is still non-convex. In addition, the distributed nature of the non-cellular massive MIMO system provides an optimal matching matrix H * This brings unique challenges. Since the access point AP-ES does not have a fixed user terminal UE, the sub-feasible solution is more complex and dynamic. Therefore, a feasible distributed solution method based on the MADRL (multi-agent deep reinforcement learning) framework is proposed.

[0125] S34: Construct Markov decision (S, D, P, R, γ), S is the state space, D is the action space, P is the penalty space, R is the immediate reward, γ is the discount factor, the access point AP-ES interacts with the mobile edge network as an intelligent agent, learns the optimal matching relationship between AP-ES and the user terminal UE, based on the environment of the MADRL algorithm, the state at time t The energy consumption of each access point AP-ES and calculation delay Decision; the state space of each agent is:

[0126]

[0127] S35: Each agent’s action is defined as The agent selects an action for each time t. The agent decides the user terminal UE to provide service based on the current state. The action space of each agent is:

[0128]

[0129] in, is the element associated with the user terminal UE and the access point AP-ES at time t;

[0130] S36: Since each agent needs to decide which user terminal UE to provide service to, the dimension of each agent action space D is 2 K , the immediate reward of the lth agent is recorded as

[0131]

[0132] Among them, p i is the penalty factor, i is the number of the penalty factor;

[0133] S37: Setting the service policy H of the access point AP-ES according to the penalty factor * ;

[0134]

[0135] Optimal jamming strategy J * The calculation method is:

[0136] A41: For a given service policy H * and the optimal task offloading ratio Λ * , the objective function of the interference strategy J is: The constraints of the objective function of the interference strategy J are:

[0137]

[0138] A42: Select a set J of candidate access points AP-ES based on the energy load imbalance degree according to a predefined threshold δ c , set J c It contains several interference strategies J of candidate access points AP-ES, each interference strategy The selection method is:

[0139]

[0140] Among them, e is the set Jc The number of the interference strategy in the

[0141] A43: Set J c Each interference strategy Substitute into the objective function of the interference strategy J and output the optimal interference strategy J according to the constraints * , so that the attacker Eve is located in the main lobe area of ​​the artificial noise beam transmitted by the access point AP-ES, and the user terminal UE is located in the side lobe area.

[0142] Aiming at the particularity of the distribution of non-cellular large-scale MIMO network system, the present invention proposes a data processing method based on security and load balancing under its architecture, which can realize an efficient and secure transmission solution against uplink passive eavesdropping and downlink pilot pollution attacks.

[0143] Using the complete offloading of the SAP algorithm, each user terminal UE is considered to load all tasks to the access point AP-ES to ensure the minimum energy consumption of the user terminal UE to perform tasks locally; the security of downlink transmission is guaranteed, and the optimized interference strategy is adopted to ensure the security of uplink transmission.

[0144] The present invention mainly considers the optimization of user terminal UE task offloading decision, interference strategy optimization and access point AP-ES service selection optimization. The service selection of access point AP-ES no longer considers the security of user terminal UE in downlink transmission. Access point AP-ES only selects user terminal UE to provide service according to the current load status.

[0145] The present invention also considers that each user terminal UE offloads tasks to the access point AP-ES with a specific probability, and performs numerical optimization to minimize e-LoBaR (energy-based load imbalance), thereby ensuring the security of uplink and downlink transmissions.

Claims

1. A data processing method for joint security and energy load balancing, characterized in that: The following steps are involved: S1: Each user terminal sends the assigned pilot sequence to the access point AP-ES, which is an access point AP equipped with an independent edge server ES; the access point AP-ES determines whether the user terminal UE is attacked by the attacker Eve by detecting abnormal conditions of the pilot signal; S2: If the user terminal UE is not attacked, it is marked as a normal user terminal UE, otherwise, it is marked as a user terminal UE under attack; For the user terminal UE under attack, determine whether the arrival angle between the user terminal UE and the access point AP-ES meets the security transmission condition. If so, the access point AP-ES maintains the connection with the user terminal UE under attack, otherwise, disconnects the connection; S3: Access points AP-ES act as intelligent agents and execute the current service strategy H on their respective local observations. * ; S4: Based on the current service strategy H * As a result, the user terminal UE calculates the optimal task offloading ratio Λ * , and based on the service strategy H * and the optimal task offloading ratio Λ * Get the optimal interference strategy J * ; The method for screening normal user terminals UE in step S2 is: Wherein, A is a set of access points AP-ES in the network system, A={1,2,…,L}, l′ is the number of the access point AP-ES in the set A, and L is the total number of access points AP-ES in the set A; The large-scale fading coefficient β of the channel between all access points AP-ES and user terminals UE lk Sort in descending order to get the set after descending sorting After sorting the collection in descending order The elements in After sorting the collection in descending order The minimum number of elements in l is the set after sorting in descending order The number of the corresponding access point AP-ES; C U U p is the subset of user terminals UE that have not been actively polluted by the attacker Eve, U p is the user terminal UE subset that Eve takes active pilot pollution attack on, and p is the subset U p The number of the user terminal UE in β l′k is the large-scale fading coefficient of the channel between the l′th access point and the kth user terminal UE; 0<δ<1, δ is a predefined threshold for screening whether the user terminal UE is under attack, k is the number of the user terminal UE in the network system, A subset of user terminals UE that are attacked; The method for screening the user terminal UE that is attacked is: The optimal task offloading ratio Λ in step S4 * The calculation method is: S41: For a given service policy H * and the optimal jamming strategy J * , derive the feasible solution of the offloading ratio Λ of the user terminal UE; The objective function of the task offloading decision of the user terminal UE is: The constraints of the objective function of task offloading decision are: in, is the total energy consumption of the kth user terminal UE, is the maximum energy consumption of the kth user terminal UE, is the energy consumption calculated locally by the user terminal UE, λ k is the uninstall rate of the user terminal UE, is the uninstallation time of the user terminal UE, is the time calculated locally by the user terminal UE, is the computing time of the edge server ES, is the time delay limit of the kth user terminal UE, P t k is the transmission power of the kth user terminal UE; S42: Constraint Transformed into two linear constraints: S43: Constraints and Merge to get: Among them, a k is the energy consumption during task offloading, b k The energy consumption generated by the local computing process, P t k is the transmit power of the user terminal UE, T k is the total task calculated at the user terminal UE, is the uplink confidentiality transmission rate in uplink transmission, is the transmission rate in the uplink transmission that is not attacked by the attacker Eve, is the effective capacitance coefficient of the CPU architecture of the user terminal UE, C k The number of CPU cycles required for the user terminal UE to calculate the task, f k The frequency of the CPU architecture of the user terminal UE; S44: Energy consumption generated by the local calculation process of the user terminal UE k Greater than maximum energy consumption Right now Energy consumption generated during task offloading of user terminal UE k Less than the energy consumption of the local computing process b k , that is, a k k ,get:​ S45: Constructing the second-order partial derivative of the objective function And the objective function E LIB It is a convex problem, which can be derived; Among them, m le is the association element between the access point AP-ES and the attacker Eve, p lk is the equivalent substitution parameter of the second-order partial derivative process, η lk It is an association element between the access point AP-ES and the user terminal UE. is the effective capacitance coefficient of the CPU architecture of the access point AP-ES, C l The number of CPU cycles required for the AP-ES calculation task, f l k The frequency of the CPU architecture of the access point AP-ES. is the transmit power of the access point AP-ES; S45: Then obtain the optimal task offloading ratio Λ * For the convex problem of task offloading ratio Λ, the CVX optimization tool is used to optimize the objective function Solve and get the optimal task offloading ratio Λ * ; The service strategy H * The optimization method is: S31: For a given optimal task offloading ratio Λ * and the optimal jamming strategy J * , establish and calculate the optimal service strategy H * The objective function is: Optimal service strategy H * The association between the user terminal UE and the access point AP-ES is represented by the derived matching matrix H=L×K; Objective Function The constraints are: Among them, η lk An element that associates the user terminal UE with the access point AP-ES, is the downlink confidentiality transmission rate that can be achieved in downlink transmission, The uplink confidentiality transmission rate that can be achieved in uplink transmission, R s ,R u are the transmission rate thresholds of the downlink transmission link and the uplink transmission link, respectively, and c is the subset The number of the user terminal UE in f l,k is the computing resource allocated by the lth access point AP-ES to the kth user terminal UE associated with it, and ε is the constraint value of the total energy consumption of K user terminals UE; S32: Define thresholds As the lower bound of the achievable confidentiality transmission rate in uplink transmission, the constraint and Carry out a merger; in, is the optimal offloading strategy for the kth user terminal UE, T k is the total task of the kth user terminal UE, C c The number of CPU cycles required for the access point AP-ES; S33: The lower limit of the achievable confidentiality transmission rate in uplink transmission is Under the conditions, The maximum tolerable delay is greater than the lower limit Then update the objective function The constraints are: S34: Construct Markov decision (S, D, P, R, γ), S is the state space, D is the action space, P is the penalty space, R is the immediate reward, γ is the discount factor, the access point AP-ES interacts with the mobile edge network as an intelligent agent, learns the optimal matching relationship between AP-ES and the user terminal UE, based on the environment of the MADRL algorithm, the state at time t The energy consumption of each access point AP-ES and calculation delay Decision; the state space of each agent is: S35: The action of each agent is defined as The agent selects an action for each time t. The agent decides the user terminal UE to provide service based on the current state. The action space of each agent is: in, is the element associated with the user terminal UE and the access point AP-ES at time t; S36: Since each agent needs to decide which user terminal UE to provide service to, the dimension of each agent action space D is 2 K , the immediate reward of the lth agent is recorded as Among them, p i is the penalty factor, i is the number of the penalty factor; S37: Setting the service policy H of the access point AP-ES according to the penalty factor * ; 2. The data processing method for joint security and energy load balancing according to claim 1, characterized in that: The method for determining whether the access point AP-ES maintains the connection with the user terminal UE under attack in step S2 is: The access point AP-ES obtains the arrival angle information of the user terminal UE and the attacker Eve respectively, and the access point AP-ES distinguishes the user terminal UE and the attacker Eve by transmitting a narrow directional beam; Maximize the difference between the narrow directional beam from the access point AP-ES to the user terminal UE and the narrow directional beam from the access point AP-ES to the attacker Eve, thereby eliminating the attacker Eve and aligning the beam to the user terminal UE; The antenna of the access point AP-ES adopts an ideal sector-based antenna pattern model, so that the kth user terminal UE is in the main lobe area of ​​the antenna, and the attacker Eve is outside the main lobe area of ​​the antenna. The lth access point AP-ES can provide services to the kth user terminal UE and meet the following conditions: Among them, θ lk is the arrival angle from the access point AP-ES to the user terminal UE, θ le is the arrival angle from access point AP-ES to attacker Eve, is the beam width of beamforming at the lth access point AP-ES.

3. The data processing method for joint security and energy load balancing according to claim 1, characterized in that: The optimal interference strategy J * The calculation method is: A41: For a given service policy H * and the optimal task offloading ratio Λ * , the objective function of the interference strategy J is: The constraints of the objective function of the interference strategy J are: A42: Select a set J of candidate access points AP-ES based on the energy load imbalance degree according to a predefined threshold δ c , set J c It contains several interference strategies J of candidate access points AP-ES, each interference strategy The selection method is: Among them, e is the set J c The number of the interference strategy in A43: Set J c Each interference strategy Substitute into the objective function of the interference strategy J and output the optimal interference strategy J according to the constraints * , so that the attacker Eve is located in the main lobe area of ​​the artificial noise beam transmitted by the access point AP-ES, and the user terminal UE is located in the side lobe area.

Images