RF fingerprint PUF authentication system and method based on PCB antenna

Abstract

This invention discloses a radio frequency fingerprint (PUF) authentication system and method based on a PCB antenna, comprising: a device to be authenticated, including a wireless communication chip and a PCB antenna, wherein the PCB antenna includes a toothed conductive edge, the toothed conductive edge comprising multiple tooth structures, and the shape and size of each tooth structure being random within a critical tolerance; and a verification device, including a main control unit and a challenge sequence generation module, a radio frequency transceiver module, a feature fusion processing module, a security database, and a matching comparison unit connected to the main control unit. The output terminal of the challenge sequence generation module is connected to the transmitting input terminal of the radio frequency transceiver module, the receiving output terminal of the radio frequency transceiver module is connected to the input terminal of the feature fusion processing module, and both the feature fusion processing module and the matching comparison unit are connected to the security database. Wireless communication exists between the radio frequency transceiver module and the wireless communication chip. This invention has a simple structure, requires no dedicated circuitry, is low in cost, and offers high security and stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of communication security and Internet of Vehicles technology, and particularly relates to a radio frequency fingerprint PUF authentication system and method based on a PCB antenna. BACKGROUND

[0002] Existing radio frequency authentication systems, such as vehicle Bluetooth keys, rely on static keys stored in non-volatile memory for authentication, which has the defect that the keys are easy to extract and clone at low cost. Although existing PUF technologies, such as SRAM PUF, can solve the key storage problem, they often require special circuit design and do not fully utilize the multi-dimensional characteristics of the radio frequency communication link. In addition, the complex multipath effect of the vehicle-mounted environment makes it difficult for traditional single-parameter radio frequency fingerprint technology to meet the vehicle-grade stability requirements. It is extremely important to provide a more convenient and simple, more practical, and lower cost solution under the conditions of ensuring high security and high stability. SUMMARY

[0003] The present application provides a radio frequency fingerprint PUF authentication system and method based on a PCB antenna to solve the problems of complex circuit and high cost in the prior art.

[0004] Technical scheme: A radio frequency fingerprint PUF authentication system based on a PCB antenna, comprising:

[0005] An authentication device, comprising a wireless communication chip and a PCB antenna, the PCB antenna is electrically connected with the wireless communication chip, the PCB antenna contains a tooth-shaped conductive edge, the tooth-shaped conductive edge contains a plurality of tooth structures, the shape and size of each tooth structure have randomness within a critical tolerance;

[0006] A verification device, comprising a master control unit and a challenge sequence generation module, a radio frequency transceiver module, a feature fusion processing module, a security database, and a matching comparison unit connected with the master control unit, the output end of the challenge sequence generation module is connected with the sending input end of the radio frequency transceiver module, the receiving output end of the radio frequency transceiver module is connected with the input end of the feature fusion processing module, the feature fusion processing module and the matching comparison unit are both connected with the security database, and there is wireless communication between the radio frequency transceiver module and the wireless communication chip.

[0007] Further, the PCB antenna contains a long side and a bending part, and the outer side of the bending part and the outer side of the long side are both provided with a tooth-shaped conductive edge.

[0008] Further, the tooth structure includes one or more of a triangular tooth, a trapezoidal tooth, a rectangular tooth, a curved tooth, and an irregular tooth.

[0009] Further, the PCB antenna includes a snake-shaped structure.

[0010] Further, the challenge sequence generation module is configured to generate a challenge sequence with multi-dimension parameter adjustment, including at least two of the following dimensions: transmission power dimension, communication channel dimension, and communication rate dimension, and each dimension contains at least two parameters.

[0011] Further, the feature fusion processing module is configured to extract a feature vector, and the feature vector contains local microscopic features of the PCB antenna, and the local microscopic features include phase noise features of intermodulation components.

[0012] Further, the feature fusion processing module is configured to receive a feature matrix containing at least two of the following: received signal strength, carrier phase offset, and instantaneous frequency offset; the feature fusion processing module extracts a local feature tensor using a convolution unit, and the local microscopic features include phase noise features of intermodulation components; a global feature tensor is extracted using a Transformer self-attention mechanism; and a high-dimensional feature vector is output by using a MobileViT lightweight neural network model.

[0013] Further, for vehicle authentication, the radio frequency transceiver module adopts a distributed multi-antenna system layout, and contains a plurality of distributed antennas located at at least two of the following positions of the vehicle: left front door, right front door, center console, and trunk.

[0014] An authentication method of a PCB antenna-based radio frequency fingerprint PUF authentication system, comprising the following steps:

[0015] Registration phase: the verification device sends a full traversal challenge sequence in a shielding environment, and receives a radio frequency response signal of the to-be-authenticated device; a high-dimensional feature vector R is extracted from the radio frequency response signal, auxiliary data W and an original key K are generated by a double error correction fuzzy extractor, and a hash value of K and W are stored in a secure database;

[0016] Authentication phase: the verification device sends a challenge sequence, receives a real-time radio frequency response signal, and extracts a to-be-tested high-dimensional feature vector R' from the real-time radio frequency response signal; the W stored in the secure database is used to perform error correction operation on R', and a to-be-tested key K' is reconstructed; a hash value of K' is calculated, and is matched with the hash value of K stored in the secure database to determine whether the authentication is successful.

[0017] Further, the method for extracting a feature vector from a radio frequency response signal comprises:

[0018] Step one: the feature fusion processing module first pre-processes the radio frequency response signal, extracts multi-dimensional physical features, and integrates the multi-dimensional physical features into an initial feature matrix;

[0019] Step two: input the initial feature matrix into the convolution unit of the feature fusion processing module, and the convolution unit extracts the local features of the PCB antenna from the feature matrix, wherein the local features include the phase noise features of the intermodulation components, and a local feature tensor is obtained;

[0020] Step three: input the local feature tensor into the Transformer self-attention mechanism unit of the feature fusion processing module, for correlation analysis of the local features under different challenge dimensions, extraction of global fingerprint invariance features across various challenge parameters, filtering of non-antenna ontology feature information, and formation of a global feature tensor;

[0021] Step four: fuse the local feature tensor and the global feature tensor, and perform dimension reduction and feature condensation on the fused features through the fully connected layer of the MobileViT lightweight neural network model, and finally output a high-dimensional feature vector.

[0022] Compared with the prior art, the radio frequency fingerprint PUF authentication system and method based on the PCB antenna provided by the application has the following beneficial effects:

[0023] (1) Based on the high-frequency skin effect and the transmission line impedance perturbation theory, the PCB antenna with a critical tolerance serrated conductive edge forms a physically unclonable microscopic topographic entropy source due to its unique structure, which is used as the only radio frequency fingerprint, without the need for a dedicated circuit, with a simple structure, easy to implement and low cost.

[0024] (2) The verification device fully stimulates the nonlinear response of the PCB antenna by sending a challenge sequence adjusted in multiple dimensions. Due to the microscopic structure of the serrated conductive edge of the PCB antenna, the equivalent inductance is changed, resulting in a unique distortion of the input impedance circle trajectory of the antenna, specifically, the zero-crossing frequency of the antenna changes from inductive to capacitive, and when the challenge sequence emitted by the verification device is emitted at maximum power, the high current density of the bent part of the PCB antenna interacts with the microscopic rough edge to produce weak passive intermodulation products. At this time, the MobileViT network of the feature fusion processing module extracts the phase noise features of the intermodulation components in the received signal spectrum, which is a microscopic feature that smooth edge antennas do not have. This scheme combines the special structure of the PCB antenna with the deep learning algorithm of its microscopic features by utilizing the microscopic manufacturing process deviation of the PCB antenna, solving the problem of key extraction and cloning in traditional authentication systems, and improving the security and reliability of the authentication system.

[0025] (3) The serrated conductive edge of the PCB antenna is arranged at the bent part of the PCB antenna, which utilizes the dual physical mechanisms of current crowding and complex fluid flow channel to significantly amplify the random effects of the microscopic serrated structure, thereby significantly improving the strength of the radio frequency fingerprint compared to ordinary straight-line antennas, greatly improving the practicality and robustness of the system.

[0026] (4) The scheme has simple structure, and is especially suitable for application in products such as automobile keys, and can fully embody the advantage of the microstructure of the PCB antenna as an entropy source of the PUF authentication system under extremely limited space. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 Structure diagram of a PCB antenna containing tooth-shaped conductive edges. DETAILED DESCRIPTION

[0028] The application will be further explained in conjunction with the drawings and specific embodiments.

[0029] A radio frequency fingerprint PUF authentication system based on a PCB antenna, comprising:

[0030] A device to be authenticated, comprising a wireless communication chip and a PCB antenna, the PCB antenna being electrically connected with the wireless communication chip, the PCB antenna containing tooth-shaped conductive edges, and containing a plurality of tooth structures, the tooth structures being microstructures, the shape and size of each tooth structure having randomness within a critical tolerance; the micro tooth structure can utilize the randomness of fluid turbulence in a wet etching process to generate a micro-morphology entropy source of a physically unclonable (PUF).

[0031] A verification device, comprising a master control unit and a challenge sequence generation module, a radio frequency transceiver module, a feature fusion processing module, a security database, and a matching comparison unit connected with the master control unit, the output end of the challenge sequence generation module being connected with the sending input end of the radio frequency transceiver module, the receiving output end of the radio frequency transceiver module being connected with the input end of the feature fusion processing module, the feature fusion processing module and the matching comparison unit both being connected with the security database, and there being wireless communication, such as standard Bluetooth communication, between the radio frequency transceiver module and the wireless communication chip. The verification device can stimulate the radio frequency response of the micro tooth-shaped structure by sending a multi-dimensional challenge sequence, and can perform identity authentication by using the feature fusion processing module.

[0032] The PCB antenna can be of any shape, such as a snake-shaped PIFA antenna structure that can fully utilize space, and can realize 2.4GHz-2.48GHz band resonance in a limited size, and the snake-shaped structure can also lengthen the effective electrical length. An S-shaped continuous bend is made on the top layer of the PCB to form multiple parallel radiation arms. The starting end of the snake-shaped structure is provided with a short-circuit pin connected with the main ground, and the feeding point is located at the short-circuit pin.

[0033] The tooth structure can be one or more of a triangular tooth, a trapezoidal tooth, a rectangular tooth, a curved tooth, and an irregular tooth, such as Figure 1The image shows a triangular tooth. Preferably, in this embodiment, the micro-tooth structure is not provided on all antenna edges, but only on the outer corners of the bends 1 and the outer sides of the long sides 2 of the serpentine trace, where the micro-tooth structure 3 meets the critical tolerance. Figure 1 As shown. The physical basis is that when radio frequency current flows through the serpentine bend, a current congestion effect occurs, and the current tends to concentrate at the inner and outer corners of the bend, where the current density can be more than three times that at the straight section. Therefore, placing the micro-tooth structure here can generate the largest radio frequency fingerprint signal with minimal physical modification, resulting in better performance.

[0034] Parameter settings: PCB material selected as FR4, dielectric constant... The copper thickness is 1 oz. Microscopic tooth structure: tooth width 80 μm, tooth height 100 μm, period 160 μm. Due to the serpentine routing, the gaps between the copper foils present a complex labyrinthine structure. During wet etching, the flow rate of the etching solution is extremely uneven at the serpentine bends. In straight sections: the fluid is relatively stable, and the microscopic tooth structure forms a relatively regular sinusoidal residual structure. In corner sections: the fluid generates violent micro-vortices. This results in extreme randomness in the microscopic teeth at the corners; some tooth tips are completely flattened, while adjacent tooth valleys form deep pits due to vortex stagnation. As a result, the edge impedance at each bend is... It becomes an independent random variable.

[0035] Microscopic tooth structures generate RF fingerprints by altering the distribution parameters at the antenna edges: Based on the high-frequency skin effect, high-frequency signal currents mainly concentrate at the antenna edges for propagation. The random morphology of the microscopic tooth structure causes the antenna to generate RF fingerprints along the signal propagation path. Surface impedance distribution Random perturbations occur: ,in, To represent the imaginary part of the impedance, after removing the parentheses... The coefficient represents the reactance (capacitive reactance or inductive reactance). Angular frequency, Permeability, For electrical conductivity, The edge roughness function defined by random lateral etching is mathematically modeled as the superposition of deterministic design profile and random process noise. ,in This is the nominal width of the antenna radiating element. For a pre-defined periodic function of the tooth-like microstructure, such as a triangular waveform or a trapezoidal wave, The term is a Gaussian random process, representing the random deviation caused by etching turbulence; this impedance perturbation causes nonlinear phase noise and amplitude distortion in the transmitted signal in both the time and frequency domains.

[0036] Entropy Source: In PCB manufacturing, the flow behavior of the etching solution within tiny gaps follows the laws of fluid dynamics. The characteristic dimensions of the tooth structure are designed as follows: Satisfying the constraints based on PCB etching dynamics: ,in, For PCB copper foil thickness, This is the process etching factor (the ratio of vertical etching rate to lateral etching rate). , This is a hydrodynamic correction factor. According to the Reynolds number formula, the Reynolds number... The Reynolds number It can determine whether a fluid is in laminar or turbulent flow, among which The density of the etching solution, For microscopic local flow velocity, Dynamic viscosity. When At extremely low speeds, the fluid is in the critical region between laminar and turbulent flow. At this point, the flow velocity within the microscopic region... and reactant concentration The state is chaotic, resulting in lateral erosion of each tooth. It possesses true randomness. This randomness is solidified at the edge of the copper foil, becoming an unrepeatable physical fingerprint.

[0037] Fingerprint excitation: Transmission line impedance perturbation; according to the skin effect, high-frequency current exists only in a thin layer on the conductor surface. Transmission. Microscopic random roughness at the antenna edge. This directly changes the effective inductance of the microstrip line. and capacitor According to transmission line theory, local characteristic impedance Fluctuations occur along the line. When signals of different power and frequency pass through, these distributed impedance discontinuities generate complex superposition of micro-reflections, resulting in unique frequency responses and nonlinear phase distortion.

[0038] The challenge sequence generation module is used to generate challenge sequences with multi-dimensional parameter adjustments, including transmit power, communication channel, and communication rate dimensions. Each dimension contains at least two parameters, and two dimensions can be selected or other dimensions can be added as needed. For example, the transmit power dimension includes at least four power levels that allow the RF amplifier circuit to operate in different linearity regions, such as +4dBm to -20dBm; the communication channel dimension includes at least three broadcast or data channels with dispersed frequency bands, such as channels 0, 19, and 39; and the communication rate dimension includes at least two different physical layer transmission rates, such as 1Mbps and 2Mbps. The above parameter combinations form... Each independent RF challenge point forms a high-dimensional feature space. The multi-level power and multi-channel challenge sequence can excite the nonlinear response of the antenna, facilitating noise immunity certification using MobileViT neural networks and fuzzy extractors.

[0039] Frequency-dimensional excitation: The serpentine antenna exhibits multiple higher-order mode resonant points. The verification equipment not only communicates at the center frequency but also probes sideband frequencies. Due to the alteration of the equivalent inductance by the micro-tooth structure, the antenna's input impedance circle plot trajectory undergoes a unique distortion. Specifically, at the instant of frequency switching, the zero-crossing frequency at which the antenna transitions from inductive to capacitive reactance shifts.

[0040] Excitation in the power dimension: When the verification device requires the key to transmit at maximum power, the high current density at the bend of the serpentine antenna interacts with the microscopically rough edges, generating weak passive intermodulation products. At this time, the feature extraction module MobileViT network focuses on the phase noise characteristics of the intermodulation components in the received signal spectrum. This is a characteristic unique to the PCB antenna and not found in antennas with smooth edges.

[0041] The feature fusion processing module includes convolutional units, a Transformer self-attention mechanism unit, and a MobileViT lightweight neural network model. It is used to extract high-dimensional feature vectors, which contain local micro-features of the PCB antenna, including phase noise features of intermodulation components. Specifically, the feature fusion processing module is configured to receive a three-dimensional feature matrix containing Received Signal Strength Indicator (RSSI), Carrier Phase Offset (CPO), and Instantaneous Frequency Offset (FO), or two of these can be selected as needed. The module uses convolutional units to extract local micro-features, including phase noise features of intermodulation components, generating a local feature tensor. It then uses the Transformer self-attention mechanism unit to extract global fingerprint invariance features across different power levels and channels, generating a global feature tensor. Finally, the MobileViT lightweight neural network model fuses the local and global feature tensors to output a high-dimensional feature vector.

[0042] This embodiment of the PCB antenna-based radio frequency fingerprint PUF authentication system can be applied to vehicle authentication. The radio frequency transceiver module adopts a distributed multi-antenna system layout, including multiple distributed antennas located at least two positions: the left front door, the right front door, the center console, and the trunk. The feature fusion processing module combines the signals received by the antennas at different locations and uses spatial diversity technology to eliminate fingerprint instability caused by multipath effects, ensuring that only the physical features of the antenna of the device to be authenticated are extracted, further improving accuracy and security. Due to the complex multipath effects in the automotive environment, traditional single-parameter radio frequency fingerprint technology is difficult to achieve automotive-grade stability. This embodiment combines the special hardware structure of the PCB antenna with its unique phase noise characteristics of intermodulation components and a multi-dimensional feature fusion method, resulting in higher security and stability, making it more suitable for application in the automotive field.

[0043] The authentication method of the PCB antenna-based radio frequency fingerprint PUF authentication system includes the following steps:

[0044] Registration phase: The verification device sends a full traversal challenge sequence in a shielded environment and receives the radio frequency response signal of the device to be certified; a high-dimensional feature vector R is extracted from the radio frequency response signal, auxiliary data W and the original key K are generated through a dual error correction fuzz extractor, and the hash value of K and W are stored in the security database;

[0045] Authentication Phase: The verification device sends a challenge sequence, receives the real-time radio frequency response signal, and extracts the high-dimensional feature vector R' to be tested from the real-time radio frequency response signal; it then performs error correction operations on R' using W stored in the security database. The key to be tested, K', is reconstructed; the hash value of K' is calculated and matched with the hash value of K stored in the security database to determine whether the authentication was successful.

[0046] Methods for extracting high-dimensional feature vectors from radio frequency response signals include:

[0047] Step 1: Constructing the Feature Matrix. The feature fusion processing module first preprocesses the received real-time RF response signal, extracting the physical features of the RF response signal, such as received signal strength, carrier phase offset, and instantaneous frequency offset, and integrates them into an initial feature matrix. The dimensions of this matrix are determined according to the parameter combination of the challenge sequence and the signal sampling window. For example, a three-dimensional tensor of Batch, Time, and Channel can be constructed, where Batch is the number of RF challenge states, Time is the number of transient response time-domain sampling points, and Channel is the selected physical feature dimension.

[0048] Step 2: Extract local micro-features. The initial feature matrix is ​​input into the convolutional unit of the feature fusion processing module. The convolutional unit performs local feature extraction on the feature matrix, focusing on capturing the local micro-features unique to the PCB antenna, including the phase noise features of the intermodulation components, to obtain the local feature tensor. This process can uncover the details of the nonlinear response in the signal caused by the tooth-like microstructure at the antenna edge.

[0049] Step 3. Extract global invariant features. Input the local feature tensor into the Transformer self-attention mechanism unit of the module. This unit will perform correlation analysis on the local features under different challenge dimensions such as transmit power, communication channel, and communication rate based on the self-attention algorithm, extract global fingerprint invariant features across various challenge parameters, filter out feature information that is not related to the antenna itself, such as environmental interference, and form a global feature tensor.

[0050] Step four: Output high-dimensional feature vectors. The local and global feature tensors are fused. The fused features are then subjected to dimensionality reduction and feature refinement through the fully connected layers of the MobileViT lightweight neural network model, ultimately outputting a high-dimensional feature vector with uniqueness and stability, thus completing the extraction of the high-dimensional feature vector.

[0051] The specific processing steps of the MobileViT network include:

[0052] Input tensor: Construct a data volume of Batch(24) * Time(100) * Channel(3). Batch(24) represents the state space of the RF challenge, i.e., the transmit power is selected in 4 levels, such as +4dBm, 0dBm, -8dBm, and -20dBm. Time(100) represents the time domain capture window of the transient response. According to the Nyquist sampling theorem, the minimum sampling rate of 1M and 2M Bluetooth rates is 4MHz. Due to the impedance discontinuity caused by the microstructure of the PCB antenna, the most significant fingerprint feature occurs in the transient interval of the signal, i.e., the moment of RF activation or frequency switching. The window length is selected as 100 sampling points, and the time length corresponding to the time is 100*(1 / 4MHz) = 25us, which can basically cover the transient interval of the captured signal. Channel(3) represents the fusion of multi-dimensional physical features. 3 corresponds to three orthogonal physical features decoupled from the original RF signal. The first is the received signal strength RSSI, which reflects the random attenuation of the radiation efficiency of the serpentine structure; the second is the carrier phase offset CPO-Phase, which reflects the random attenuation of the radiation efficiency of the serpentine structure. The three factors are: group delay jitter caused by the antenna, and instantaneous frequency offset (FO-Frequency), which reflects the resonance offset of the serpentine antenna under different channels. The tensor dimensions (24, 100, 3) mentioned above are the preferred parameters in this embodiment. In practical applications, the batch dimension can be adjusted between 10 and 100 according to the design of the challenge sequence, and the timer dimension can be adjusted between 50 and 200 according to the sampling rate and transient length to adapt to different hardware resources and security level requirements. MobileViT's Transformer module automatically learns that the phase abrupt changes caused by certain fixed bends always exist in all challenge data. This structural invariance is extracted as the final fingerprint vector.

Claims

1. A PCB antenna based radio frequency fingerprint PUF authentication system, characterized in that, Comprise: A device to be authenticated, comprising a wireless communication chip and a PCB antenna, the PCB antenna being electrically connected with the wireless communication chip, the PCB antenna comprising a tooth-shaped conductive edge, the tooth-shaped conductive edge comprising a plurality of tooth structures, each tooth structure having randomness in shape and size within a critical tolerance, the PCB antenna comprising a long side and a bending part, and the outer side of the bending part and the outer side of the long side are both provided with the tooth-shaped conductive edge; A verification device, comprising a host unit and a challenge sequence generation module, a radio frequency transceiver module, a feature fusion processing module, a security database, and a matching comparison unit connected with the host unit, the output end of the challenge sequence generation module being connected with the sending input end of the radio frequency transceiver module, the receiving output end of the radio frequency transceiver module being connected with the input end of the feature fusion processing module, the feature fusion processing module and the matching comparison unit both being connected with the security database, and there being wireless communication between the radio frequency transceiver module and the wireless communication chip; the feature fusion processing module is used for extracting a feature vector, the feature vector comprising local microscopic features of the PCB antenna, and the local microscopic features including phase noise features of intermodulation components; Feature The feature fusion processing module extracts a global feature tensor using a Transformer self-attention mechanism and outputs a high-dimensional feature vector using a MobileViT lightweight neural network model.

2. The PCB antenna based radio frequency fingerprint PUF authentication system of claim 1, wherein, The tooth structure includes one or more of triangular teeth, trapezoidal teeth, rectangular teeth, curved teeth, and irregular teeth.

3. The PCB antenna based radio frequency fingerprint PUF authentication system according to claim 1 or 2, characterized in that, The PCB antenna comprises a snake-shaped structure.

4. The PCB antenna based radio frequency fingerprint PUF authentication system of claim 1 or 2, wherein, The challenge sequence generation module is used for generating a challenge sequence with multi-dimensional parameter adjustment, including at least two of transmission power dimension, communication channel dimension, and communication rate dimension, and each dimension comprises at least two parameters.

5. The PCB antenna based radio frequency fingerprint PUF authentication system of claim 1 or 2, wherein, The feature fusion processing module is configured to receive a feature matrix comprising at least two of received signal strength, carrier phase offset, and instantaneous frequency offset; the feature fusion processing module extracts a local feature tensor using a convolution unit, and the local microscopic features include phase noise features of intermodulation components.

6. The PCB antenna based radio frequency fingerprint PUF authentication system of claim 1 or 2, wherein, For vehicle authentication, the radio frequency transceiver module adopts a distributed multi-antenna system layout and comprises a plurality of distributed antennas located at least two positions of the left front door, the right front door, the center console, and the trunk of the vehicle.

7. An authentication method of the PCB antenna-based radio frequency fingerprint PUF authentication system according to any one of claims 1-6, characterized in that, Comprise the following steps: Registration phase: the verification device sends a full traversal challenge sequence in a shielding environment, and receives a radio frequency response signal of the device to be authenticated; Extract a high-dimensional feature vector R from the radio frequency response signal, generate auxiliary data W and an original key K through a double error correction fuzzy extractor, and store the hash value of K and W in the security database; Authentication phase: the verification device sends a challenge sequence, receives a real-time radio frequency response signal, and extracts a to-be-tested high-dimensional feature vector R' from the real-time radio frequency response signal; performs error correction operation on R' using W stored in the security database to reconstruct a to-be-tested key K'; calculates the hash value of K', and matches it with the hash value of K stored in the security database to determine whether the authentication is successful. 8.The authentication method of the PCB antenna-based radio frequency fingerprint PUF authentication system according to claim 7, wherein, The method for extracting a high-dimensional feature vector from a radio frequency response signal comprises: Step one: the feature fusion processing module first preprocesses the radio frequency response signal, extracts multi-dimensional physical features, and integrates the multi-dimensional physical features into an initial feature matrix; Step two: input the initial feature matrix into the convolution unit of the feature fusion processing module, the convolution unit extracts the local features of the PCB antenna from the feature matrix, the local features include the phase noise features of the intermodulation components, and obtains a local feature tensor; Step three: input the local feature tensor into the Transformer self-attention mechanism unit of the feature fusion processing module, which is used for correlation analysis of the local features under different challenge dimensions, extracts global fingerprint invariance features across various challenge parameters, filters out feature information that is not the antenna ontology, and forms a global feature tensor; Step four: fuse the local feature tensor and the global feature tensor, and reduce the dimension and feature of the fused features through the full connection layer of the MobileViT lightweight neural network model, and finally output a high-dimensional feature vector.

Images