A covert communication method of UAV cooperative jamming for cognitive radio network

Abstract

The present invention discloses a method for concealed communication of unmanned aerial vehicle cooperative interference for cognitive radio network. The method comprises the following steps: taking concealed rate maximization as optimization goal, under the conditions that the probability of eavesdropper detection error satisfies the minimum detection error probability constraint, the primary user transmission rate satisfies the minimum transmission rate constraint, the trajectory of the unmanned aerial vehicle satisfies the flight speed and flight altitude constraints, and the transmission power of the unmanned aerial vehicle satisfies the maximum transmission power constraint, the transmission power of the unmanned aerial vehicle and the flight trajectory are jointly optimized to form an iterative update data stream of the transmission power scheme and the flight trajectory scheme, construct a generative adversarial migration network, train and generate the optimal transmission power and flight trajectory of the unmanned aerial vehicle according to the environmental state information and the position information of each node, and operate the action of the unmanned aerial vehicle in the communication area. The present invention makes full use of the advantages of big data, and effectively shortens the execution time of unmanned aerial vehicle cooperative interference resource optimization by migrating the source domain features with similar communication environment as the target domain.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a UAV cooperative interference covert communication method for a cognitive radio network. Background Art

[0002] As an important technical means to alleviate the problem of spectrum scarcity in wireless communications and provide effective spectrum access, cognitive radio has been widely used in various occasions such as the Internet of Things, medical health, vehicle-mounted networks, and emergency systems, and has achieved remarkable results. However, the openness of wireless transmission channels and the dynamic access characteristics of cognitive radio make cognitive radio vulnerable to malicious attacks, and its security performance faces severe challenges.

[0003] At present, covert communication, as a low-detection / low-interception communication method, can provide a higher level of security performance compared to physical layer security and encryption technology. Specifically, covert communication attempts to make it impossible for eavesdroppers to detect the occurrence of communication behavior, thereby completely hiding the communication information. In particular, collaborative interference, as an effective auxiliary means of covert communication, sends interference signals to eavesdroppers, thereby effectively confusing eavesdroppers. Among them, drones are used as mobile interference sources due to their strong maneuverability and controllability. By dynamically adjusting their motion trajectory and transmission power to effectively improve the performance of covert communications, they have gradually attracted the attention of domestic and foreign researchers.

[0004] X. Jiang et al. proposed that when implementing covert communications in the ground-to-air communication network, drones can be used as aerial base stations, relays or mobile interference sources to effectively improve the performance of covert communications. They also pointed out that when drones are used as interference sources, they will interfere with both eavesdroppers and legitimate users. Therefore, it is necessary to optimize the interference position and power of drones to enhance the concealment of legitimate users' transmissions without affecting the transmission quality of legitimate users.

[0005] W. Liang et al. gave an approximate expression for the covert rate of drones as mobile interference sources in Nakagami-m fading channels and additive Gaussian white noise channels. The experimental simulation results show that the interference signal sent by the drone can increase the noise uncertainty of the eavesdropper, thereby confusing the eavesdropper and effectively enhancing the security performance of covert communications.

[0006] X. Zhou et al. proposed to use a full-duplex drone that can send interference signals to eavesdroppers while communicating with legitimate users, and maximize the minimum concealment rate by jointly optimizing the drone's trajectory and interference power. However, in the process of solving the optimization problem, the assumptions are too idealistic, that is, it is necessary to assume that the eavesdropper has an optimal detection threshold and can obtain the eavesdropper's relevant channel information.

[0007] Z.Li et al. used drones as interference sources to assist cognitive users in covert communication. Assuming that the eavesdropper detection threshold is unknown and some channel state information is known, a model-driven generative adversarial network is used to jointly optimize the drone's three-dimensional flight trajectory and interference power. Although the assumptions are close to actual application scenarios, they do not fully utilize the advantages of big data samples and have a high computational complexity. Summary of the invention

[0008] In order to solve the above technical problems, the present invention proposes a UAV cooperative interference covert communication method for cognitive radio networks, which realizes the joint optimization of UAV cooperative interference power and flight trajectory in the covert communication of cognitive radio networks.

[0009] The technical solution of the present invention to solve the above technical problem is: a UAV cooperative interference covert communication method for cognitive radio network, comprising the following steps:

[0010] With the maximization of concealment rate as the optimization goal, the UAV transmit power and flight trajectory are jointly optimized under the conditions that the eavesdropper's detection error probability meets the minimum detection error probability constraint, the main user's transmission rate meets the minimum transmission rate constraint, the UAV trajectory meets the flight speed and flight altitude constraints, and the UAV transmit power meets the maximum transmit power constraint. An iterative update data stream of the transmit power scheme and the flight trajectory scheme is formed, and a generative adversarial migration network is constructed. According to the environmental state information and the position information of each node, the optimal UAV transmit power and flight trajectory scheme is trained and generated, and the UAV's actions in the communication area are operated according to the UAV transmit power and flight trajectory.

[0011] Furthermore, the steps of iteratively updating the data flow for forming the transmission power scheme and the flight trajectory scheme are as follows:

[0012] 2a) Convert the joint optimization problem of UAV transmission power and flight trajectory into UAV transmission power optimization sub-problem and UAV flight trajectory optimization sub-problem;

[0013] 2b) Iteratively solve the two optimization sub-problems to form the following iterative update data flow:

[0014]

[0015] Among them, I represents the location information of each node, S represents the environmental status information, and L represents the total number of iterative updates of the data stream. and They represent the UAV transmission power schemes generated at the first, second, lth iteration and the end of the iteration respectively. and They represent the UAV flight trajectory plans generated at the first, second, lth iteration and the end of the iteration respectively.

[0016] Furthermore, the steps of constructing a generative adversarial transfer network are:

[0017] 3a) Based on the transmission power scheme P UAV and flight trajectory scheme T UAV Iteratively update the data stream, construct a source generator with the channel parameters related to the eavesdropper as weight parameters, the source generator includes a data input layer, a multi-layer power generation layer, a multi-layer trajectory update layer and a power and trajectory output layer;

[0018] 3b) construct a target generator with the same structure as the source generator;

[0019] 3c) Construct a discriminator consisting of a three-layer neural network, which includes a data input layer, a hidden layer, and a detection error probability output layer.

[0020] Furthermore, the training steps for generating the optimal UAV transmission power and flight trajectory are as follows:

[0021] 4a) collecting training data samples in a source domain with a similar wireless communication environment as the target domain, and pre-training the source generator and discriminator using supervised learning;

[0022] 4b) copying the pre-trained first half of the network weight parameters in the source generator to the first half of the network weight parameters in the target generator, and initializing the second half of the network weight parameters in the target generator according to the channel distribution information of the target domain;

[0023] 4c) Update the current environment status information and the location information of each node, generate the UAV transmission power and flight trajectory along the direction of the target generator data flow, and obtain the detection error probability of the eavesdropper along the direction of the discriminator data flow;

[0024] 4d) Calculate the loss function LF of the target generator separately G And the loss function LF of the discriminator D ;

[0025] 4e) Set the threshold ε. During the tth training process, if the target generator loss function value And the discriminator loss function value The UAV transmission power and flight trajectory generated by 4c) are used as the optimal UAV cooperative interference resource optimization solution, otherwise, execute 4f);

[0026] 4f) Along LF D Update the discriminator weight parameters along the gradient descent direction of LF GThe weight parameters of the second half of the target generator are updated in the direction of the gradient ascent. When the number of weight parameter updates reaches the maximum number of times T max When , the UAV transmission power and flight trajectory generated by 4c) are used as the optimal UAV cooperative interference resource optimization solution, otherwise, return to 4c).

[0027] Furthermore, the source generator in the step 3a) has a data input layer including location information I and environmental status information S, a power generation layer including a UAV transmission power plan generated in each iteration of the iterative update data stream, a trajectory update layer including a UAV flight trajectory plan generated in each iteration of the iterative update data stream, and a power and trajectory output layer including a UAV transmission power plan and a flight trajectory plan generated at the end of the iteration in the iterative update data stream. The number of power generation layers and trajectory update layers is L-1, where L represents the total number of iterative updates of the iterative update data stream.

[0028] Furthermore, in the step 3c), the discriminator has a data input layer including the UAV transmission power scheme and flight trajectory scheme output by the source generator or the target generator and the UAV transmission power scheme and flight trajectory scheme when there is no covert communication, the detection error probability output layer includes the detection error probability of the eavesdropper, and the number of neurons m in the hidden layer is determined by the following formula:

[0029] m=2k+1

[0030] Where k is the number of neurons in the discriminator data input layer.

[0031] Furthermore, the loss function LF of the target generator in step 4d) is G , calculated using the following formula:

[0032]

[0033] Among them, N represents the total number of time slots divided by the UAV flight time, C[n] represents the covert communication rate of the nth time slot, G(I,S)[n] represents the UAV transmission power plan and flight trajectory plan generated by the generator when the current node position information I and environmental state information S are input in the nth time slot, D(G(I,S)[n]) represents the UAV transmission power plan and flight trajectory plan generated by the generator when the generator is input in the nth time slot, and the eavesdropper detection error probability output by the discriminator.

[0034] Furthermore, the loss function LF of the discriminator in step 4d) is D , calculated using the following formula:

[0035]

[0036] Wherein, N represents the total number of time slots divided into UAV flight time, x represents the transmission power scheme and flight trajectory scheme of the UAV when no covert communication is performed, D(x) represents the error probability of eavesdropper detection output by the discriminator when the transmission power scheme and flight trajectory scheme of the UAV without covert communication are input, G(I,S)[n] represents the transmission power scheme and flight trajectory scheme of the UAV generated by the generator when the current node position information I and environmental state information S are input in the nth time slot, and D(G(I,S)[n]) represents the error probability of eavesdropper detection output by the discriminator when the transmission power scheme and flight trajectory scheme of the UAV generated by the generator are input in the nth time slot.

[0037] The technical effects of the present invention are:

[0038] 1. The present invention constructs a generative adversarial transfer network based on the joint optimization problem of drone transmission power and flight trajectory, and trains to generate the optimal drone transmission power and flight trajectory, thereby increasing the noise uncertainty of the eavesdropper. It not only effectively reduces the computational complexity, but also provides a theoretical basis for the constructed generative adversarial transfer network.

[0039] 2. The present invention makes full use of the advantages of big data samples. By extracting and migrating the data features of the source domain that have a similar wireless communication environment to the target domain, it effectively shortens the execution time of the optimization of UAV collaborative interference resources in the covert communication of the cognitive radio network and effectively enhances the security performance of covert communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart for realizing the present invention;

[0041] Figure 2 Generate an adversarial transfer network for the present invention;

[0042] Figure 3 Comparison of the convergence speed between the generative adversarial transfer network of the present invention and the model-driven generative adversarial network. DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the present invention more clear and explicit, the present invention is further described in detail below with reference to the accompanying drawings.

[0044] like Figure 1 As shown, Figure 1 The specific steps of the UAV cooperative interference covert communication method for cognitive radio network of the present invention are as follows:

[0045] Step 1: Form an iterative update data stream for the power transmission plan and trajectory planning plan.

[0046] Taking into account the constraints such as the probability of eavesdropper detection error, the primary user transmission rate, the UAV flight trajectory and the transmission power, the concealment rate is maximized by jointly optimizing the UAV transmission power and flight trajectory. The optimization problem is expressed as

[0047]

[0048] std[n]≥1-ε,

[0049]

[0050]

[0051] H min ≤z UAV [n]≤H max ,

[0052]

[0053] Among them, P UAV [n] and T UAV [n] represents the transmission power and flight trajectory of the UAV in the nth time slot, C[n] represents the concealment rate of the cognitive user in the nth time slot, d[n] represents the detection error probability of the eavesdropper in the nth time slot, ε∈[0,1] is the preset value, P PR [n] represents the transmission rate of the primary user in the nth time slot, Indicates the minimum transmission rate that the primary user meets, T UAV [n] and T UAV [n-1] represents the flight trajectory of the UAV at the nth and n-1th time slots, respectively, v max represents the maximum flight speed of the drone, T represents the flight time of the drone, N represents the total number of time slots into which the flight time T of the drone is divided, and z UAV [n] represents the flight altitude of the UAV at the nth time slot, H max Indicates the maximum flight altitude of the drone, H min Indicates the minimum flight altitude of the drone. Indicates the maximum transmission power of the drone.

[0054] The block coordinate descent method is used to transform the joint optimization problem of UAV transmission power and flight trajectory into the UAV transmission power optimization sub-problem and the UAV flight trajectory optimization sub-problem. The UAV transmission power optimization sub-problem refers to fixing the UAV flight trajectory and optimizing the UAV transmission power; the UAV flight trajectory optimization sub-problem refers to fixing the UAV transmission power and optimizing the UAV flight trajectory. By iteratively solving the UAV transmission power optimization sub-problem and the UAV flight trajectory optimization sub-problem, the power transmission scheme P is formed. UAVand trajectory planning scheme T UAV The iterative update data flow is:

[0055]

[0056] Among them, I represents the location information of each node, S represents the environmental status information, and L represents the total number of iterative updates of the data stream. and They represent the UAV transmission power schemes generated at the first, second, lth iteration and the end of the iteration respectively. and They represent the UAV flight trajectory plans generated at the first, second, lth iteration and the end of the iteration respectively.

[0057] Step 2: Construct a generative adversarial transfer network.

[0058] 2.1) Construct a source generator based on the iterative update data stream, including a data input layer, a multi-layer power generation layer, a multi-layer trajectory update layer, and a power and trajectory output layer, wherein the weight parameters of each layer are set as channel parameters related to the eavesdropper. Specifically, the data input layer includes the location information I and the environmental state information S, the power generation layer includes the drone transmission power plan generated by each iteration in the iterative update data stream, the trajectory update layer includes the drone flight trajectory plan generated by each iteration in the iterative update data stream, the power and trajectory output layer includes the drone transmission power plan and flight trajectory plan generated at the end of the iteration in the iterative update data stream, and the number of power generation layers and trajectory update layers is L-1, where L represents the total number of iterative updates of the iterative update data stream.

[0059] 2.2) Construct a target generator with the same structure as the source generator.

[0060] 2.3) Construct a discriminator consisting of a three-layer neural network, including a data input layer, a hidden layer, and a detection error probability output layer. Specifically, the data input layer includes the UAV transmission power scheme and flight trajectory scheme output by the source generator or the target generator, as well as the UAV transmission power scheme and flight trajectory scheme when there is no covert communication. The detection error probability output layer includes the detection error probability of the eavesdropper. The number of neurons m in the hidden layer is determined by the following formula:

[0061] m=2k+1

[0062] Where k is the number of neurons in the discriminator data input layer.

[0063] Step 3: Train the Generative Adversarial Transfer Network.

[0064] 3.1) Collect training data samples in the source domain that has a similar wireless communication environment as the target domain, and use supervised learning to pre-train the source generator and discriminator.

[0065] 3.2) The first half of the network weight parameters in the pre-trained source generator are copied to the first half of the network weight parameters in the target generator, and the second half of the network weight parameters in the target generator are initialized according to the channel distribution information of the target domain. For example, in the second half of the network weight parameters in the target generator, if the channel between the drone and the eavesdropper obeys the Rice distribution, the channel parameters between the drone and the eavesdropper are initialized to the Rice distribution; if the channel between the eavesdropper and the cognitive user obeys the Rayleigh distribution, the channel parameters between the eavesdropper and the cognitive user are initialized to the Rayleigh distribution; if a part of the additive white Gaussian noise is superimposed on the signal received by the eavesdropper, the noise variance at the eavesdropper receiving end is initialized to the additive white Gaussian noise.

[0066] 3.3) Update the current environment status information and the location information of each node, generate the drone transmission power and flight trajectory along the direction of the target generator data flow, and obtain the detection error probability of the eavesdropper along the direction of the discriminator data flow.

[0067] 3.4) Calculate the loss function LF of the target generator separately G And the loss function LF of the discriminator D Among them, the loss function LF of the target generator is G And the loss function LF of the discriminator D , calculated using the following formulas:

[0068]

[0069]

[0070] Wherein, N represents the total number of time slots divided by the UAV flight time, C[n] represents the covert communication rate of the nth time slot, G(I,S)[n] represents the UAV transmission power plan and flight trajectory plan generated by the generator when the current node position information I and environmental state information S are input in the nth time slot, D(G(I,S)[n]) represents the eavesdropper detection error probability output by the discriminator when the UAV transmission power plan and flight trajectory plan generated by the generator are input in the nth time slot, x represents the UAV transmission power plan and flight trajectory plan when no covert communication is performed, and D(x) represents the eavesdropper detection error probability output by the discriminator when the UAV transmission power plan and flight trajectory plan without covert communication are input.

[0071] 3.5) Set the threshold ε. During the tth training process, if the target generator loss function value And the discriminator loss function value The UAV transmission power and flight trajectory generated by 2c) are used as the optimal UAV collaborative interference resource optimization plan, and the UAV is operated according to the UAV transmission power and flight trajectory as the optimal UAV collaborative interference resource optimization plan, otherwise, execute 2f).

[0072] 3.6) Along LF D Update the discriminator weight parameters along the gradient descent direction of LF G The weight parameters of the second half of the target generator are updated in the direction of the gradient ascent. When the number of weight parameter updates reaches the maximum number of times T max When the UAV transmission power and flight trajectory generated by 2c) are used as the optimal UAV collaborative interference resource optimization plan, and the UAV is operated according to the UAV transmission power and flight trajectory as the optimal UAV collaborative interference resource optimization plan, otherwise, return to 2c).

[0073] Step 4: Operate the UAV's movements in the communication area according to the UAV's transmission power and flight trajectory.

[0074] The effect of the present invention can be further illustrated by simulation:

[0075] 1. Simulation conditions: Assume that the cognitive radio network contains four nodes: the primary user transmitter and receiver, the cognitive user transmitter and receiver. The underlay spectrum sharing mode is mainly considered. There is a randomly distributed eavesdropper in the network, trying to steal the communication information of the cognitive user. In order to achieve covert communication of the cognitive user, the drone sends interference signals to confuse the eavesdropper.

[0076] 2. Simulation content: The convergence speed of the UAV jamming trajectory and power joint optimization was simulated and compared using the generative adversarial transfer network of the present invention and the model-driven generative adversarial network proposed by Z. Li et al. The results are as follows: Figure 3 .

[0077] It can be seen from the simulation results that the model-driven generative adversarial network and the generative adversarial transfer network converge to output the UAV interference trajectory and power joint optimization scheme after 89 and 35 iterations respectively, indicating that the generative adversarial transfer network proposed in the present invention can effectively shorten the execution time of UAV collaborative interference resource optimization in covert communications of cognitive radio networks.

Claims

1. A UAV cooperative jamming covert communication method for cognitive radio networks, characterized in that: The following steps are involved: Taking the maximization of concealment rate as the optimization goal, under the conditions that the probability of eavesdropper detection error meets the minimum detection error probability constraint, the primary user transmission rate meets the minimum transmission rate constraint, the UAV trajectory meets the flight speed and flight altitude constraints, and the UAV transmission power meets the maximum transmission power constraint, the UAV transmission power and flight trajectory are jointly optimized to form an iterative update data stream of the transmission power scheme and the flight trajectory scheme, and construct a generative adversarial migration network. According to the environmental state information and the position information of each node, the optimal UAV transmission power and flight trajectory scheme is trained and generated, and the UAV's actions in the communication area are controlled according to the UAV transmission power and flight trajectory; The training steps for generating the optimal UAV transmission power and flight trajectory are as follows: a) Collect training data samples in the source domain with a similar wireless communication environment as the target domain, and use supervised learning to pre-train the source generator and discriminator; b) The first half of the network weight parameters in the pre-trained source generator are copied to the first half of the network weight parameters in the target generator, and the second half of the network weight parameters in the target generator are initialized according to the channel distribution information of the target domain; c) Update the current environment status information and the location information of each node, generate the UAV transmission power and flight trajectory along the target generator data flow direction, and obtain the detection error probability of the eavesdropper along the discriminator data flow direction; d) Calculate the loss function of the target generator separately And the loss function of the discriminator ; e) Setting thresholds , in During the training process, if the target generator loss function value And the discriminator loss function value , then the UAV transmission power and flight trajectory generated by c) are used as the optimal UAV cooperative interference resource optimization solution, otherwise, execute f); f) Along Update the discriminator weight parameters along the gradient descent direction The weight parameters of the second half of the target generator are updated in the direction of the gradient ascent. When the number of weight parameter updates reaches the maximum number of times specified When , the UAV transmission power and flight trajectory generated by c) are used as the optimal UAV cooperative interference resource optimization solution, otherwise, return to c).

2. The method for cooperative interference covert communication of unmanned aerial vehicles in cognitive radio networks according to claim 1 is characterized in that: The steps of iteratively updating the data flow for forming the transmission power scheme and the flight trajectory scheme are: 2a) The joint optimization problem of UAV transmission power and flight trajectory is transformed into the UAV transmission power optimization sub-problem and the UAV flight trajectory optimization sub-problem; 2b) Iteratively solve the two optimization sub-problems to form the following iterative update data flow: ； in, Indicates the location information of each node. Indicates environmental status information. Indicates the total number of iterative updates of the data stream, , , and Respectively represent sequence sequence The UAV transmission power scheme generated by the iteration and the end of the iteration is , , and Respectively represent sequence sequence The UAV flight trajectory plan generated at the end of the iteration.

3. The method for cooperative interference covert communication of unmanned aerial vehicles in cognitive radio networks according to claim 1 is characterized in that: The steps of constructing a generative adversarial transfer network are: 3a) Based on the transmit power scheme and flight trajectory plan Iteratively update the data stream, construct a source generator with the channel parameters related to the eavesdropper as weight parameters, the source generator includes a data input layer, a multi-layer power generation layer, a multi-layer trajectory update layer and a power and trajectory output layer; 3b) Construct a target generator with the same structure as the source generator; 3c) Construct a discriminator consisting of a three-layer neural network, which includes a data input layer, a hidden layer, and a detection error probability output layer.

4. The method for cooperative interference covert communication of unmanned aerial vehicles in cognitive radio networks according to claim 3 is characterized in that: In step 3a), the source generator, whose data input layer includes position information and environmental status information The power generation layer includes the UAV transmission power plan generated by each iteration in the iterative update data stream, the trajectory update layer includes the UAV flight trajectory plan generated by each iteration in the iterative update data stream, and the power and trajectory output layer includes the UAV transmission power plan and flight trajectory plan generated at the end of the iteration in the iterative update data stream. The number of layers of the power generation layer and the trajectory update layer is ,in Indicates the total number of iterations of the iterative update data stream.

5. The method for cooperative interference covert communication of unmanned aerial vehicles in cognitive radio networks according to claim 3 is characterized in that: The discriminator in step 3c) has a data input layer including the UAV transmission power scheme and flight trajectory scheme output by the source generator or the target generator and the UAV transmission power scheme and flight trajectory scheme when there is no covert communication, and the detection error probability output layer includes the detection error probability of the eavesdropper, the number of neurons in the hidden layer Determined by the following formula: ； in is the number of neurons in the discriminator data input layer.

6. The method for cooperative interference covert communication of unmanned aerial vehicles in cognitive radio networks according to claim 1 is characterized in that: The loss function of the target generator in step d) , calculated using the following formula: ； in, Indicates the total number of time slots divided by the drone flight time, Indicates The covert communication rate of time slots is Indicated in Input the current location information of each node in each time slot and environmental status information When the generator generates the UAV transmission power plan and flight trajectory plan, Indicated in The generator generates the UAV transmission power plan and flight trajectory plan as input at each time slot, and the discriminator outputs the probability of eavesdropper detection error.

7. The method for cooperative interference covert communication of unmanned aerial vehicles in cognitive radio networks according to claim 1 is characterized in that: The loss function of the discriminator in step d) , calculated using the following formula: ； in, Indicates the total number of time slots divided by the drone flight time, Indicates the transmission power plan and flight trajectory plan of the drone when no covert communication is performed. It represents the probability of eavesdropper detection error output by the discriminator when the transmission power plan and flight trajectory plan of the drone that is not performing covert communication are input. Indicated in Input the current location information of each node in each time slot and environmental status information When the generator generates the UAV transmission power plan and flight trajectory plan, Indicated in The generator generates the UAV transmission power plan and flight trajectory plan as input at each time slot, and the discriminator outputs the probability of eavesdropper detection error.

Images