Search

CN-121580172-B - ASM-oriented radio frequency fingerprint identification method and system

CN121580172BCN 121580172 BCN121580172 BCN 121580172BCN-121580172-B

Abstract

The invention relates to an ASM-oriented radio frequency fingerprint identification method and system, and relates to the field of radio frequency fingerprint identification, wherein the method comprises the steps of acquiring ASM signals sent by a ship and extracting a multi-element time sequence of the ASM signals; extracting time domain local features and frequency domain structural features of the multi-element time sequence, splicing the time domain local features and the frequency domain structural features to obtain first spliced features, fusing the first spliced features to obtain fused features, fusing feature enhancement and space information of the fused features to obtain multi-scale fused features, extracting global statistical information and key detail information of the multi-scale fused features, splicing the global statistical information and the key detail information to obtain second spliced features, identifying the second spliced features, and outputting radio frequency fingerprint identification results. The invention can effectively identify the radio frequency fingerprint and further judge whether the ASM signal is tampered.

Inventors

  • JIANG QI

Assignees

  • 苏州大学

Dates

Publication Date
20260508
Application Date
20260123

Claims (8)

  1. 1. An ASM-oriented radio frequency fingerprint identification method, comprising: step S1, acquiring ASM signals sent by a ship, and extracting a multi-element time sequence of the ASM signals; S2, extracting time domain local features and frequency domain structural features of the multi-element time sequence, splicing the time domain local features and the frequency domain structural features to obtain first spliced features, and fusing the first spliced features to obtain fused features; S3, carrying out feature enhancement and spatial information fusion on the fusion features to obtain multi-scale fusion features; S4, extracting global statistical information and key detail information of the multi-scale fusion feature, and splicing the global statistical information and the key detail information to obtain a second spliced feature; S5, identifying the second splicing characteristics and outputting a radio frequency fingerprint identification result; the processes of the steps S2 to S5 are implemented through the SE-Net network, and the loss function adopted when training the SE-Net network is expressed as follows: ; Wherein, the As a function of the total loss, In order to be the main task loss, For the channel sparsity regularization term, For the channel diversity regularization term, 、 Regularized weight coefficients for the first and second adjustable; the main task loss Expressed as: ; Wherein, the In order to be of the size of the batch, For the total number of rf fingerprint categories, Is the first in the batch The index of the individual samples is set to be, Is the first An index of the individual categories is provided, One-hot encoding for a genuine tag, Is a predictive probability; the channel sparsity regularization term Expressed as: ; Wherein, the For the number of Inception modules in a SE-Net network, Is the first The channel attention weight vectors output by the Inception modules, Is an L1 norm, promoting sparsity; The KL divergence of weight distribution and uniform distribution is adopted to prevent excessive sparseness; the operation of normalizing the weight vector into probability distribution is realized through a Softmax function; is distributed in a discrete and uniform way; is a balance parameter; The channel diversity regularization term Expressed as: ; Wherein, the For the number of Inception modules in a SE-Net network, The channel attention weight vector output for the i Inception th module, Is the L2 norm of the vector.
  2. 2. The ASM-oriented radio frequency fingerprint identification method of claim 1, wherein the step S2 extracts the time domain local feature and the frequency domain structural feature of the multi-element time sequence, then splices the time domain local feature and the frequency domain structural feature, and then fuses the first spliced feature to obtain a fused feature, the method comprises the following steps: The first convolution layer and the second convolution layer are respectively used for extracting the time domain local feature and the frequency domain structural feature of the multi-element time sequence, and the time domain local feature and the frequency domain structural feature extracted by the first convolution layer and the second convolution layer are spliced through Concat layers to obtain a first spliced feature And then the first splicing characteristic Feature fusion is realized through the batch normalization layer and MISH functions in sequence, and fusion features are obtained 。
  3. 3. The ASM-oriented radio frequency fingerprint identification method of claim 1, wherein the step S3 performs feature enhancement and spatial information fusion on the fusion features, and the method for obtaining the multi-scale fusion features comprises the following steps: By passing through Individual cascaded Inception module pair fusion features Performing multi-scale feature enhancement and spatial information fusion processing to obtain multi-scale fusion features , wherein, ; The Inception module adopts a three-branch parallel structure for inputting characteristics Multi-scale extraction and fusion are performed by first inputting features Simultaneously inputting a first branch, a second branch and a third branch which are parallel, wherein, The first branch extracts input characteristics through a cascade structure of a third convolution layer, a batch normalization layer and a ReLU activation function Basic local features of (a) ; The second branch inputs the characteristic through the fourth convolution layer Channel compression Obtaining characteristics Features are then extracted simultaneously using three parallel fifth, sixth, and seventh convolution layers Is a multi-scale feature of (2) 、 、 The three parallel convolved outputs are then processed 、 、 Splicing through Concat layers, and finally realizing feature fusion through an eighth convolution layer to obtain features ; The third branch extracts global features of large receptive fields through a ninth convolution layer, a tenth convolution layer, a batch normalization layer and a ReLU activation function ; Output characteristics of three branches 、 、 Splicing along the channel dimension to form splice features Reuse of eleventh convolutional layer pair-splice feature Performing cross-channel information fusion and dimension adjustment to obtain initial multi-scale fusion characteristics And finally, fusing the initial multi-scale fusion features And carrying out normalization and nonlinear activation through the batch normalization layer and the ReLU activation function to obtain a channel attention weight vector output by a single Inception module.
  4. 4. The ASM-oriented radio frequency fingerprint identification method of claim 1, wherein the step S4 extracts global statistics information and key detail information of the multi-scale fusion feature, and the method for splicing the global statistics information and the key detail information to obtain a second spliced feature comprises the following steps: fusing multiple scale features A global average pooling layer and a global maximum pooling layer are input in parallel, wherein the global average pooling layer is used for extracting global statistical information of feature distribution The global maximum pooling layer is used for extracting key detail information of feature distribution Global statistics extracted by global average pooling layer and global maximum pooling layer And key detail information Splicing again through Concat layers to obtain a second splicing characteristic by fusing two pooling results 。
  5. 5. The ASM-oriented radio frequency fingerprint identification method as set forth in claim 1, wherein the step S5 identifies the second stitching feature, and the method for outputting the radio frequency fingerprint identification result comprises: Stitching the second stitching feature The method comprises the steps of inputting a first full-connection layer to achieve self-adaptive dimension reduction, carrying out nonlinear activation through a ReLU function, inputting a result obtained after nonlinear activation of the ReLU function into a second full-connection layer to achieve feature reconstruction, carrying out final activation on the reconstructed features through a Sigmoid function, and outputting a radio frequency fingerprint identification result.
  6. 6. An ASM-oriented radio frequency fingerprint identification system for implementing the ASM-oriented radio frequency fingerprint identification method according to any one of claims 1 to 5, comprising: The acquisition module is used for acquiring ASM signals sent by the ship and extracting a multi-element time sequence of the ASM signals; the splicing and fusing module is used for extracting the time domain local feature and the frequency domain structural feature of the multi-element time sequence, splicing the time domain local feature and the frequency domain structural feature to obtain a first splicing feature, and fusing the first splicing feature to obtain a fused feature; the feature enhancement module is used for carrying out feature enhancement and spatial information fusion on the fusion features to obtain multi-scale fusion features; The extraction and splicing module is used for extracting global statistical information and key detail information of the multi-scale fusion features and splicing the global statistical information and the key detail information to obtain second splicing features; and the identification module is used for identifying the second splicing characteristic and outputting a radio frequency fingerprint identification result.
  7. 7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the ASM-oriented radio frequency fingerprinting method of any one of claims 1 to 5 when the computer program is executed by the processor.
  8. 8. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the ASM-oriented radio frequency fingerprint identification method as claimed in any one of claims 1 to 5.

Description

ASM-oriented radio frequency fingerprint identification method and system Technical Field The invention relates to the field of radio frequency fingerprint identification, in particular to an ASM-oriented radio frequency fingerprint identification method and system. Background In recent years, the central position of an automatic ship identification system (AIS) in global marine navigation safety and management has been increasingly highlighted. However, with the increasing demand for maritime safety-related communications, AIS has long been overwhelmed. Accordingly, the international navigation mark association (IALA) and the International Telecommunications Union (ITU) have proposed the concept of very high frequency data exchange systems (VDES) on the basis of AIS. VDES integrates very high frequency data exchange (VDE), AIS and application specific information (ASM) functions in the Very High Frequency (VHF) band. As part of VDES, ASM supports efficient exchange of information between ship-to-ship, ship-to-shore, and ship-to-satellite. ASM occupies two frequency bands, the center frequency is 161.950MHz and 162.000MHz respectively, and the channel bandwidth is 50kHz. However, in practical applications, some important data carried by ASM signals may be tampered with by a fraudster. Currently, radio Frequency Fingerprinting (RFFI) has been used as a physical layer security solution in wireless communications to distinguish between authentic and spoofed signals. Radio frequency fingerprinting refers to a hardware defect of a device in the production and manufacturing process, and has uniqueness, stability, uniqueness and collectability. Currently, the advent of Deep Learning (DL) has led to high performance solutions for radio frequency fingerprinting. Some typical deep learning models for radio frequency fingerprint identification comprise a Convolutional Neural Network (CNN), a residual network (ResNet), a long-short-term memory (LSTM) and the like, but the conventional neural network has inherent limitations that when processing time-domain or frequency-domain signals, the conventional neural network performs indiscriminate convolution operation on all channels, and is unable to distinguish a key region from a redundant or noise region, which results in weak feature discrimination learned by the model, and meanwhile, when facing complex scenes with large variation of signal amplitude dynamic range and severe fluctuation of signal-to-noise ratio, the model is unstable due to lack of self-adaption capability, and in order to improve performance, network depth and parameter quantity can only be increased, so that the calculation cost is high and the deployment is difficult. In summary, the existing deep learning model and neural network model have large calculated amount, and key features cannot be really and effectively extracted when feature extraction is performed on ASM signals, so that radio frequency fingerprints cannot be effectively identified. Disclosure of Invention Therefore, the technical problem to be solved by the invention is to solve the problem that key features cannot be really and effectively extracted when feature extraction is carried out on ASM signals in the prior art, so that radio frequency fingerprints cannot be effectively identified. In order to solve the technical problems, the invention provides an ASM-oriented radio frequency fingerprint identification method, which comprises the following steps: step S1, acquiring ASM signals sent by a ship, and extracting a multi-element time sequence of the ASM signals; S2, extracting time domain local features and frequency domain structural features of the multi-element time sequence, splicing the time domain local features and the frequency domain structural features to obtain first spliced features, and fusing the first spliced features to obtain fused features; S3, carrying out feature enhancement and spatial information fusion on the fusion features to obtain multi-scale fusion features; S4, extracting global statistical information and key detail information of the multi-scale fusion feature, and splicing the global statistical information and the key detail information to obtain a second spliced feature; and S5, identifying the second splicing characteristic and outputting a radio frequency fingerprint identification result. In one embodiment of the present invention, the step S2 extracts a time domain local feature and a frequency domain structural feature of the multi-element time sequence, then splices the time domain local feature and the frequency domain structural feature, and fuses the first spliced feature to obtain a fused feature, where the method includes: The first convolution layer and the second convolution layer are respectively used for extracting the time domain local feature and the frequency domain structural feature of the multi-element time sequence, and the time domain local feature and the frequency domain st