Terahertz physical layer key generation method based on leaky-wave antenna

Abstract

The invention belongs to the technical field of terahertz communication physical layer security, and particularly relates to a terahertz physical layer key generation method based on leaky-wave antenna space-spectrum coupling characteristics. According to the invention, the sensitivity of the effective antenna gain of the leaky-wave antenna to the frequency and the spatial propagation direction is utilized to make up for the defect of low eavesdropping channel doubtful degree caused by lack of small-scale fading in terahertz communication, and the conditional generative adversarial network is designed to accurately speculate the equivalent channel power gain. A physical layer key is generated based on the estimated channel power gain. According to the method, the channel power gain can be accurately estimated from the received signal strength with noise and auxiliary information, the reciprocity of an equivalent channel is enhanced, and the method has a lower key mismatch rate and a higher key generation rate; under the condition of different signal-to-noise ratios, good key generation performance can be kept.

Description

Technical Field

[0001] The present invention belongs to the technical field of terahertz physical layer security, and in particular relates to a terahertz physical layer key generation method. Background Art

[0002] With the development of next-generation wireless communications, the terahertz (THz) frequency band has attracted considerable attention due to its abundant spectrum resources, promising to meet the future demands for ultra-high-speed, ultra-low-latency communications. However, the physical layer security of THz communications faces new challenges. Traditional physical layer key generation schemes rely on the small-scale fading characteristics of the channel as a source of channel randomness. However, in the THz band, narrow beams and high-gain antennas are typically used to overcome severe path loss, resulting in sparse propagation paths and very weak small-scale fading effects, making it difficult to provide sufficient channel randomness for physical layer key generation.

[0003] Leaky-wave antennas offer a novel antenna solution for terahertz systems due to their simple structure, low cost, and unique space-frequency coupled propagation characteristics, where the radiation beam direction is closely related to the operating frequency. This characteristic can be exploited to increase channel randomness. However, directly using received signal strength to generate physical layer keys is susceptible to severe noise interference, suffers from low reciprocity, and may lack randomness, resulting in poor key generation performance. Accurately estimating legitimate channel characteristics is crucial for enhancing the physical layer security of terahertz communications, but existing solutions cannot quickly and accurately determine the power gain of space-frequency coupled terahertz channels.

[0004] Deep learning methods have demonstrated strong potential for handling the complex nonlinear mapping relationships in wireless communication systems. Generative adversarial networks (GANs) and their variant, conditional generative adversarial networks (cGANs), can learn the underlying distribution of data and generate specific samples based on this conditional information, offering new solutions for accurately estimating channel characteristics. Existing research has demonstrated that cGAN-assisted channel estimation can achieve high accuracy. Therefore, it is necessary to explore a new terahertz physical layer key generation protocol that combines the space-frequency coupling characteristics of leaky wave antennas with deep learning techniques. Summary of the Invention

[0005] The purpose of the present invention is to propose a terahertz physical layer key generation method based on the space-spectrum coupling characteristics of leaky wave antennas, so as to generate physical layer keys in terahertz frequency band communications more quickly and accurately.

[0006] The terahertz physical layer key generation method based on leaky wave antenna provided by the present invention is described in detail in the following sections: Figure 1 、 Figure 2 The specific steps are as follows:

[0007] (1) Subchannel allocation and initial information acquisition: Based on the measured spatial transmission direction, the communicating parties use the space-frequency coupling transmission characteristics of the leaky-wave antenna to determine the corresponding maximum antenna gain frequency in the transmission direction of both parties. A set of subchannels with different center frequencies are divided equally around this frequency. Then, reference signals are transmitted to each subchannel within the coherence time to obtain the received signal strength measurement value of each subchannel.

[0008] (2) Equivalent channel power gain estimation using a conditional generative adversarial network (cGAN): Using the received signal strength and subchannel center frequency as conditional inputs, a locally pre-trained conditional generative adversarial network (cGAN) is used to accurately estimate the equivalent channel power gain of each subchannel.

[0009] (3) Dynamic quantization and initial bit sequence generation: The communicating parties use the equivalent channel power gain sequence as the channel entropy source and perform operations including dynamic determination of quantization level, dynamic determination of quantization interval, and quantization to convert it into an initial bit sequence.

[0010] (4) Preliminary bit negotiation and key formation based on run-length search: Using the preliminary bit negotiation mechanism based on run-length search, the same bit segments are screened out in the initial bit sequence and spliced ​​to form the same initial physical layer key.

[0011] Further:

[0012] The specific process of sub-channel allocation and initial information acquisition in step (1) is as follows:

[0013] Without loss of generality, assume that the legitimate communication sender is Alice and the receiver is Bob. During the coherence time, the reference signal θ is transmitted. Alice uses the space-frequency coupling characteristic formula of the leaky wave antenna:

[0014]

[0015] Calculate the frequency point where the maximum antenna gain can be obtained theoretically, where f co is the cutoff frequency of the leaky wave antenna; Alice uses f max (θ) as the center, set the total number of subchannels N and the frequency interval Δf between adjacent subchannels, and symmetrically generate a subchannel frequency sequence f=[f1,f2,f3,…,f N ], and ensure that all sub-channel frequencies f k Both higher than f co Alice and Bob send reference signals to each other on the determined N subchannels within the coherence time, and measure the received signal strength measurement value sequence of each of them on each subchannel k. and By utilizing the space-frequency coupling characteristics of leaky-wave antennas, the characteristics of the terahertz channel show significant changes with frequency and propagation angle, which makes up for the defect of low eavesdropping channel ambiguity caused by the lack of small-scale fading in terahertz secure communications. At the same time, by utilizing the ultra-large bandwidth resources of the terahertz frequency band, the reference signal is transmitted in multiple sub-channels, and a longer equivalent channel power gain sequence is obtained in a shorter time, which can significantly improve the key generation rate.

[0016] In step (2), the equivalent channel power gain estimation is performed using the conditional generative adversarial network (cGAN). The specific process is as follows:

[0017] Alice and Bob, the two legal communication parties, each set the center frequency f of each subchannel k obtained in step 1 to k , the corresponding received signal strength measurement value Combined into conditional information vector Using the locally pre-trained conditional generative adversarial network (cGAN), the randomly generated noise vector z and the conditional information vector y are used as input to accurately estimate the equivalent channel power gain of each subchannel k. After completing the estimation of all sub-channels, both parties obtain their respective equivalent channel power gain sequences Compared with directly using the measured received signal strength as the source for generating the physical layer key, the conditional generative adversarial network (cGAN) is used to process the noisy received signal strength to estimate the equivalent channel power gain that better reflects the inherent characteristics of the channel, avoiding the randomness of the loss channel and improving the reciprocity of the equivalent channel.

[0018] The specific process of dynamic quantization and initial bit sequence generation in step (3) is as follows:

[0019] Alice estimates the equivalent channel power gain sequence Calculate the estimated entropy ε:

[0020]

[0021] in, Indicates the probability of equivalent channel power gain value, and dynamically determines the number of quantization levels m based on this, m should satisfy the condition m≤2 ε And m≥2, after confirmation, Alice sends m to Bob, and both parties synchronize the quantization level number; then both parties calculate their respective equivalent channel power gain sequences based on the synchronized quantization level number. The approximate probability density function is obtained by kernel density estimation Combined with the preset total probability ratio of the guard band α, the range of the power gain value is divided into m quantization intervals I j =(q j ,q j+1 -gj+1 ] and m-1 guard bands g j , where j = 0, 1, ..., m-2, I m-1 =(q m-1 ,q m ], q0 and q m are the corresponding minimum and maximum channel power gain values, and the following conditions are met:

[0022]

[0023] After the communication parties complete the division of quantization interval and protection band, they calculate the equivalent channel power gain sequence of each other. Each consecutive value in To quantify: If Falls into the jth quantization interval I j , then quantized to level index j. Each valid quantization level index j is mapped to an n = log2(m)-bit binary bit string. After the entire sequence mapping is completed, the communicating parties exchange the positions of the remaining mapped values ​​within their respective sequences and retain only the quantized bit strings corresponding to these positions. These are then concatenated in sequence to form their respective initial bit sequences. This quantization scheme, dynamically adjusted based on the statistical characteristics of the sequence, maximizes the use of channel randomness while minimizing quantization errors between the communicating parties.

[0024] The specific process of preliminary bit negotiation and key formation based on run-length search in step (4) is as follows:

[0025] The two communicating parties pre-set the run length s, where a run is defined as a string of s consecutive identical quantized bits; Alice searches for all runs of length s in her initial bit sequence and records the starting position index of each run to form a position list POS A , and sends the list to Bob; Bob receives POS A After that, it checks whether there is a run of length s at the corresponding position in its own initial bit sequence, and records the starting position index of the run observed by both parties to form a common position list POS B , and POS B Send back to Alice; both communicating parties use the common location list POS B For each jointly confirmed run start position, both parties extract a quantized bit string from the run starting at that position in their respective sequences. Finally, both parties sort all the extracted quantized bit strings according to their position in POS BThe two parties connect the bits in the order in which they are sent to form their respective shared initial physical layer keys. This preliminary bit negotiation mechanism based on finding a common run is used to exchange run position information rather than bit content. Under the premise of strictly ensuring security, the consistent bit segments of both parties are efficiently screened out as the shared initial physical layer key, further reducing the key mismatch rate.

[0026] The terahertz physical layer key generation method based on conditional generative adversarial networks proposed in this invention has the following characteristics and advantages compared with existing physical layer security technologies:

[0027] (1) Characteristics of the new propagation channel: The spatial-frequency coupling characteristics of the leaky-wave antenna in the terahertz band and the large-scale path loss are effectively utilized as a random source for key generation. In terahertz physical layer security technology, the contradiction between the dependence of traditional physical layer key generation methods on small-scale fading and the fact that small-scale fading in the terahertz band is so weak that it can be ignored is overcome.

[0028] (2) High-precision channel feature estimation: It is proposed that the estimator trained using a conditional generative adversarial network can accurately estimate the channel power gain from the noisy received signal strength and auxiliary information, thereby enhancing the reciprocity of the equivalent channel.

[0029] (3) Improved key generation performance: Due to the more accurate estimated power gain and better reciprocity, the keys generated by this method have a lower key mismatch rate and a higher key generation rate than the traditional method that directly uses the received signal strength.

[0030] (4) Robustness and generalization ability: Under different signal-to-noise ratio conditions, this method can maintain good key generation performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a framework diagram of the terahertz physical layer key generation method based on the spatial spectrum coupling characteristics of the leaky wave antenna of the present invention.

[0032] Figure 2 This is a flow chart of the terahertz physical layer key generation method based on the spatial spectrum coupling characteristics of the leaky wave antenna of the present invention.

[0033] Figure 3 The performance comparison of the terahertz physical layer key generation method based on the spatial spectrum coupling characteristics of the leaky wave antenna of the present invention and the key generation method directly using the received signal strength is shown in FIG1.

[0034] Figure 4 This is a performance comparison between the terahertz physical layer key generation method based on the spatial spectrum coupling characteristics of the leaky wave antenna of the present invention and the key generation method directly using the received signal strength (2).

[0035] Figure 5This is a diagram illustrating the key generation performance of the terahertz physical layer key generation method based on the spatial spectrum coupling characteristics of the leaky wave antenna of the present invention under different signal-to-noise ratios of the receiver Bob. DETAILED DESCRIPTION

[0036] The present invention will be further described below with reference to the following embodiments and accompanying drawings, so as to more clearly understand the essence of the present invention, as well as the effectiveness and advantages of the present invention. However, it should be understood that the specific embodiments described herein are not intended to limit the present invention.

[0037] The embodiment is a specific implementation of the method of the present invention in a typical point-to-point terahertz communication scenario. The two parties in legitimate communication are Alice and Bob, where the sender Alice is equipped with a leaky-wave antenna in TE1 mode and the receiver Bob is equipped with an omnidirectional antenna.

[0038] System parameters and channel environment settings: A one-dimensional parallel plate leaky wave antenna is used, and the distance between the plates is d = 3×10-3m. The cutoff frequency is calculated from this (c is the speed of light). The number of sub-channels N = 30, and the frequency interval Δf between adjacent sub-channels is 2.0 GHz. The communication frequency range is set between 50 GHz and 450 GHz. The transmission angle θ between Alice and Bob obeys the uniform distribution U(15°, 85°), and the communication distance d obeys the uniform distribution U(3m, 25m). The transmission power used by both parties in the legal communication channel estimation is p = 21.76 dBm. Considering the atmospheric absorption effect in the terahertz band, the approximate model described in Recommendation ITU-RP.676 is used to calculate the atmospheric attenuation, and the water vapor relative humidity influence factor is set to 0.0157. In order to simulate the imperfect reciprocity and noise interference of the actual channel, Alice's sub-channel receives the noise power Bob's received noise power for each sub-channel

[0039] Conditional Generative Adversarial Network (cGAN) model parameter settings: The cGAN used in this paper includes a generator G(·) and a discriminator D(·). The length of the noise vector z in the input of the generator G(·) is set to 4. The generator learning rate is set to 0.0003, and the discriminator learning rate is set to 0.0002. Both the generator and the discriminator use the Adam optimizer with beta parameters of (0.5, 0.999). The cosine annealing learning rate scheduling strategy is T maxThe minimum learning rate is set to 500, and the minimum learning rate is set to 0.000001. The generator's L1 loss weight, λ1, is initially set to 1.0. When the generator's total loss falls below a preset threshold of 0.6, λ1 is automatically adjusted to 0.1. This adjustment occurs only once. The batch size of the training data input to the model is set to 64. In each epoch, the discriminator is trained once. The generator is trained 10 times. The model is trained for a total of 4500 epochs.

[0040] Dynamic quantization and preliminary bit negotiation parameter settings: The number of quantization levels m is dynamically determined based on the calculated estimated entropy, and the maximum allowed number of quantization levels is set to 16. The total probability of the guard band is α = 0.2, and the run length is s = 1.

[0041] Effect analysis: Figure 3 and Figure 4 This figure shows a performance comparison between the proposed solution and a method that directly uses received signal strength for physical layer key generation. It can be seen that the proposed solution significantly reduces the key mismatch rate and generally generates longer keys with the same number of reference signal measurements, demonstrating that the introduction of the cGAN estimator can more effectively exploit the randomness and reciprocity of the space-frequency coupling channel characteristics. Figure 5 It shows that under the influence of different noises on the receiver Bob, the key mismatch rate of the present invention can be maintained at a low value, and the key generation length is still maintained at a high level, which verifies that the proposed physical layer key generation scheme exhibits excellent robustness under different noise levels.

Claims

1. A terahertz physical layer key generation method based on leaky wave antenna, characterized in that: The specific steps are: (1) Subchannel allocation and initial information acquisition: Based on the measured spatial transmission direction, the communicating parties use the space-frequency coupling transmission characteristics of the leaky-wave antenna to determine the corresponding maximum antenna gain frequency in the transmission direction of both parties. A set of subchannels with different center frequencies are divided equally around this frequency. Then, reference signals are transmitted to each subchannel within the coherence time to obtain the received signal strength measurement value of each subchannel. (2) Equivalent channel power gain estimation using a conditional generative adversarial network (cGAN): Using the received signal strength and subchannel center frequency as conditional inputs, a locally pre-trained conditional generative adversarial network (cGAN) is used to accurately estimate the equivalent channel power gain of each subchannel. (3) Dynamic quantization and initial bit sequence generation: The communicating parties use the equivalent channel power gain sequence as the channel entropy source and perform operations including dynamic determination of quantization level, dynamic determination of quantization interval, and quantization to convert it into an initial bit sequence. (4) Preliminary bit negotiation and key formation based on run-length search: Using the preliminary bit negotiation mechanism based on run-length search, the same bit segments are screened out in the initial bit sequence and spliced ​​to form the same initial physical layer key.

2. The terahertz physical layer key generation method according to claim 1, characterized in that: The specific process of sub-channel allocation and initial information acquisition in step (1) is as follows: Assume that the legitimate communication sender is Alice and the receiver is Bob. During the coherence time, the reference signal is transmitted θ. Alice uses the space-frequency coupling characteristic formula of the leaky wave antenna: Calculate the frequency point where the maximum antenna gain can be obtained theoretically, where f co is the cutoff frequency of the leaky wave antenna; Alice uses f max (θ) as the center, set the total number of subchannels N and the frequency interval Δf between adjacent subchannels, and symmetrically generate a subchannel frequency sequence f=[f1,f2,f3,…,f N ], and ensure that all sub-channel frequencies f l Both higher than f co Alice and Bob send reference signals to each other on the determined N subchannels within the coherence time, and measure the received signal strength measurement value sequence of each of them on each subchannel k. and By utilizing the space-frequency coupling characteristics of leaky-wave antennas, the characteristics of the terahertz channel show significant changes with frequency and propagation angle, which makes up for the low degree of suspicion of eavesdropping channels caused by the lack of small-scale fading in terahertz secure communications. At the same time, by utilizing the ultra-large bandwidth resources of the terahertz frequency band, the reference signal is transmitted in multiple sub-channels, and a longer equivalent channel power gain sequence is obtained in a shorter time, thereby significantly improving the key generation rate.

3. The terahertz physical layer key generation method according to claim 2, characterized in that: In step (2), the equivalent channel power gain estimation is performed using the conditional generative adversarial network (cGAN). The specific process is as follows: Alice and Bob, the two legal communication parties, each set the center frequency f of each subchannel k obtained in step 1 to k , the corresponding received signal strength measurement value Combined into conditional information vector Using the locally pre-trained conditional generative adversarial network (cGAN), the randomly generated noise vector z and the conditional information vector y are used as input to accurately estimate the equivalent channel power gain of each subchannel k. After completing the estimation of all sub-channels, both parties obtain their respective equivalent channel power gain sequences 4. The terahertz physical layer key generation method according to claim 3, characterized in that: The dynamic quantization and initial bit sequence generation described in step (3) are as follows: Alice estimates the equivalent channel power gain sequence Calculate the estimated entropy ε: in, Indicates the probability of equivalent channel power gain value, and dynamically determines the number of quantization levels m based on this, m should satisfy the condition m≤2 ε And m≥2, after confirmation, Alice sends m to Bob, and both parties synchronize the quantization level number; then both parties calculate their respective equivalent channel power gain sequences based on the synchronized quantization level number. The approximate probability density function is obtained by kernel density estimation Combined with the preset total probability ratio of the guard band α, the range of the power gain value is divided into m quantization intervals I j =(q j ,q j+1 -g j+1 ] and m-1 guard bands g j , where j = 0, 1, ..., m-2, I m-1 =(q m-1 ,q m ], q0 and q m are the corresponding minimum and maximum channel power gain values, and meet the following conditions: After the communication parties complete the division of quantization interval and protection band, they calculate the equivalent channel power gain sequence of each other. Each consecutive value in To quantify: If Falls into the jth quantization interval I j , then quantize it to level index j; map each valid quantization level index j to a binary bit string of n=log2(m) bits; after the entire sequence mapping is completed, the communicating parties exchange the position information of the mapping values ​​retained in their respective sequences, and only retain the quantization bit strings corresponding to these positions, and splice them in order to form their respective initial bit sequences.

5. The terahertz physical layer key generation method according to claim 4, characterized in that: The specific process of preliminary bit negotiation and key formation based on run-length search in step (4) is as follows: The two communicating parties pre-set the run length s, where a run is defined as a string of s consecutive identical quantized bits; Alice searches for all runs of length s in her initial bit sequence and records the starting position index of each run to form a position list POS A , and sends the list to Bob; Bob receives POS A After that, it checks whether there is a run of length s at the corresponding position in its own initial bit sequence, and records the starting position index of the run observed by both parties to form a common position list POS B , and POS B Send back to Alice; Both communicating parties use a common location list POS B For each mutually confirmed run start position, both parties extract a quantized bit string from the run starting at that position in their respective sequences; Finally, both parties will extract all the quantized bit strings according to their POS B The shared initial physical layer key is formed.

Images