CN115186714B - Feature Amplification Method of NIC Spectrum Fingerprint Based on Feature Correlation and Adaptive Decomposition - Google Patents
Feature Amplification Method of NIC Spectrum Fingerprint Based on Feature Correlation and Adaptive Decomposition Download PDFInfo
- Publication number
- CN115186714B CN115186714B CN202210854425.3A CN202210854425A CN115186714B CN 115186714 B CN115186714 B CN 115186714B CN 202210854425 A CN202210854425 A CN 202210854425A CN 115186714 B CN115186714 B CN 115186714B
- Authority
- CN
- China
- Prior art keywords
- spectral
- feature
- radiation source
- spectrum
- signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001228 spectrum Methods 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 39
- 238000003199 nucleic acid amplification method Methods 0.000 title claims abstract description 27
- 230000003321 amplification Effects 0.000 title claims abstract description 26
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 10
- 230000005855 radiation Effects 0.000 claims abstract description 110
- 230000003595 spectral effect Effects 0.000 claims abstract description 85
- 238000004364 calculation method Methods 0.000 claims abstract description 15
- 238000001514 detection method Methods 0.000 claims abstract description 5
- 238000004458 analytical method Methods 0.000 claims description 21
- 238000012545 processing Methods 0.000 claims description 16
- 238000000605 extraction Methods 0.000 claims description 11
- 230000009286 beneficial effect Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 2
- 230000004069 differentiation Effects 0.000 claims description 2
- 230000008030 elimination Effects 0.000 claims description 2
- 238000003379 elimination reaction Methods 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 8
- 238000012935 Averaging Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000012512 characterization method Methods 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 229920002430 Fibre-reinforced plastic Polymers 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000037237 body shape Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000011151 fibre-reinforced plastic Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Image Analysis (AREA)
Abstract
本发明属于辐射源指纹识别技术领域，具体涉及一种基于特征相关与自适应分解的网卡频谱指纹特征放大方法，具体指的是对无线网络信号检测过程中收到的多个网卡信号进行高精度频率估计，提取频谱指纹特征，并解决网卡设备之间辐射源个体特征差异较小的问题，实现细节放大。具体方法是提取了指纹信息更加丰富的局部频谱作为指纹特征，通过对多个辐射源的频谱特征进行加权求均值获得辐射源频谱特征的公共部分，基于自适应分解得到频谱特征的主体成分和杂散成分，同时在计算过程中基于相关运算实现频谱特征中的细节放大，同时消除了有意调制量，得到的经过放大的最终频谱特征，为后续提高网卡设备辨识成功率奠定基础。
The invention belongs to the technical field of radiation source fingerprint identification, and in particular relates to a network card spectrum fingerprint feature amplification method based on feature correlation and self-adaptive decomposition, specifically refers to performing high-precision frequency estimation on multiple network card signals received in the wireless network signal detection process, extracting spectrum fingerprint features, solving the problem of small differences in individual characteristics of radiation sources between network card devices, and realizing detail amplification. The specific method is to extract the local spectrum with richer fingerprint information as the fingerprint feature, obtain the common part of the spectral feature of the radiation source by weighting and averaging the spectral features of multiple radiation sources, and obtain the main components and spurious components of the spectral feature based on adaptive decomposition. At the same time, in the calculation process, the details of the spectral feature are amplified based on the correlation operation, and the intentional modulation amount is eliminated.
Description
Technical Field
The invention belongs to the technical field of radiation source fingerprint identification, and particularly relates to a network card frequency spectrum fingerprint feature amplification method based on feature correlation and self-adaptive decomposition.
Background
Radiation source fingerprint recognition technology, also called specific radiation source recognition (Specific Emitter Identification, SEI), refers to the technology of extracting features of electromagnetic signals, which can embody the hardware differences of radiation source transmitters, so as to identify specific radiation source equipment. The hardware difference characteristics are independent of the transmitted signal pattern parameters, are independent of the transmitted information, cannot be forged, and cannot be avoided. This information specific to the radiation source hardware is carried on the signal in an unintentionally modulated manner.
For radiation source fingerprinting, a core problem is the accurate characterization of radiation source hardware differences. The key information for identification is actually the unintentional modulation of the device contained in the signal. The unintentional modulation is not a major component and is subtle and imperceptible as compared to the intentional modulation. Moreover, the characteristics calculated for signals with the same intentional modulation mode are highly similar, and the attached fingerprint information which can represent the unintentional modulation of the emission source is very weak in practice.
There is thus an important problem in radiation source fingerprinting: the hardware difference of different radiation sources is very small in practice, most of information of extracted features represents the commonality of equipment, and only very fine individual differences exist, so that the subsequent classification and identification are greatly plagued.
Prior studies have been made to solve this problem, and there is a demand for recognition of fine differences by using a higher-precision classifier, for example, by using a neural network (CHEN P B, GUO Y L, LI g.discrete adversarial networks for specific emitter identification, electronics Letters,2020,56 (1)); some estimate and cancel the intentional modulation of the signal (MERCHANT K, REVAY S, STANTCHEV G, NOUSAIN B, deep Learning for RF Device Fingerprinting in Cognitive Communication Networks, IEEE Journal of Selected Topics in Signal Processing,2018.12 (1): 160-167.Doi: 10.1109/jstsp.2018.2794646.) however the accuracy of the estimate affects the unintentional modulation information and introduces new errors; some suppress the principal components of the signal by wavelet transformation (WU L W, NIU J P, WANG Z, et al Primary Signal Suppression Based on Synchrosqueezed Wavelet Transform, journal of Electronics & Information Technology,2019,42 (8): 2045-2052.), however, these all deal directly with the signal waveform in the time domain.
Compared with other domain features, the frequency domain radiation source fingerprint features are relatively stable and less affected by noise environment. But the difference between the signal spectra is small, in particular the signal spectra of the same modulation pattern, modulation parameters are highly similar. The radiation source fingerprint features extracted on the basis are very fine, and further processing is needed.
Although the same modulated signal spectrum from different devices is similar in body, subtle differences are always present, the latter being of interest for radiation source fingerprinting. The main component and the differential component representing the information of the radiation source equipment can be decomposed by specific processing and signal self-adaptive decomposition means. The variational modal decomposition (Variational mode decomposition, VMD) is an adaptive, completely non-recursive method of modal variational and signal processing. The method has the advantages that the number of modal decomposition can be determined, the adaptivity is represented by determining the number of modal decomposition of a given sequence according to actual conditions, the optimal center frequency and the limited bandwidth of each modal can be adaptively matched in the subsequent searching and solving process, the effective separation of inherent modal components (Intrinsic Mode Function, IMF) and the frequency domain division of signals can be realized, the effective decomposition components of the given signals are further obtained, and the optimal solution of the variation problem is finally obtained.
Disclosure of Invention
The invention aims to solve the technical problems that: among the radiation source individual identification problems, it is most critical to extract the unintentional modulation information contained in the signal. The unintentional modulation information tends to be lower energy than the intentional modulation, and is "submerged" in the intentional modulation information. If the modulation modes of signals transmitted by different radiation source devices are the same, the modulation parameters are the same, even the information is the same, the intentional modulation is the same under the condition, the extracted characteristics are highly similar due to the influence of the intentional modulation, the differences among individuals are only reflected at the fine positions, and the difficulty of fingerprint identification is greatly improved. In order to better exert the main effect of unintentional modulation in SEI and fully utilize the advantages of spectral features in fingerprint feature characterization, the invention provides a network card spectral feature amplification method based on feature correlation and self-adaptive decomposition.
The existing radiation source fingerprint identification method mainly comprises two steps, wherein the first step is to extract features based on a large amount of data of known tags; the second step is based on the existing characteristics as a database, the received signal without the tag is extracted and characterized for analysis, and the extracted characteristics are compared with the existing characteristic data in the database to finish the identification of the target of the unknown tag.
Compared with the processing flow of the existing method, the invention needs to add a step to analyze the characteristic components and remove the part which is not beneficial to the fingerprint identification of the radiation source, therefore, the invention is mainly divided into three steps, namely:
(1) Data processing and spectral feature extraction: the first step is consistent with the prior method, and the difference is that the pretreatment precision and efficiency are improved, and a new frequency spectrum characteristic is extracted.
(2) Spectral reference feature extraction and feature component analysis: this is a newly added step of the present invention. In order to achieve feature amplification, fingerprint features need to be analyzed based on the data of known tags. The spectral reference characteristics and the amount of intentional modulation independent of the SEI are obtained in this step. This step needs to be performed only once.
(3) Extracting label-free data amplification characteristics: extracting features from the newly received data without tags and achieving target identification corresponds to the second step of the existing method, namely completing the identification of the radiation source for the newly received data.
The invention adopts the technical scheme that the network card spectrum fingerprint feature amplifying method based on feature correlation and self-adaptive decomposition comprises the following steps:
s1: data processing and spectrum feature extraction;
this step is mainly for processing and extracting spectral features to be amplified from the signal received at the receiver. The final output of this step is the spectral feature to be amplified, denoted by the symbol f (k). The method comprises the following steps:
s1.1 data reception and preprocessing
After the receiver receives the signal, the original signal is subjected to pretreatment such as detection, filtering, noise reduction and the like to obtain a signal sample x to be processed 0 (t)；
According to the IEEE802.11 series standard, a specific frame head structure (Preamble) exists in a wireless network signal, and the invention mainly extracts the fingerprint characteristics of a radiation source based on the frame head part of the signal.
Signal sample x to be processed 0 (t) performing an analysis and obtaining a frame header portion x of the signal sample to be processed by detection 1 (t) (pretreatment of signal frame header is described in L.Sun, X.Wang, A.Yang and Z.Huang, "Radio Frequency Fingerprint Extraction Based on Multi-Dimension Approximate Entropy," in IEEE Signal Processing Letters, vol.27, pp.471-475,2020.).
S1.2, signal frequency offset estimation and elimination
The frequency band positions of the signals actually transmitted are different, the actual receiving also can cause certain frequency deviation, and the existence of the frequency deviation value can cause certain deviation in the range determination of the frequency spectrum characteristics extracted subsequently, so that the frame head part x of the signal sample to be processed needs to be estimated accurately 1 Frequency offset f of (t) 0 . Due to the frame header part x of the signal samples to be processed 1 (t) is the signal sample x to be processed 0 (t) a truncated portion, thus x 0 (t) and x 1 The frequency offset of (t) is the same, based on x longer in length in the frequency offset estimation 0 (t) performing frequency offset estimation to improve estimation accuracy; the method comprises the following steps:
s1.2.1 pair x based on Fourier interpolation algorithm 0 Frequency offset value f of (t) 0 Estimation is performed (specific estimation methods see Aboutanios E, mulgiew B.Iteractive frequency estimation by interpolation on Fourier coeffients.IEEE Transactions on signal processing 2005;53 (4): 1237-42.);
s1.2.2 according to f 0 Will x 1 (t) moving to zero frequency to obtain a baseband signal, and eliminating the influence of frequency offset;
s1.2.3 removing the mean value of the baseband signal, and normalizing the energy to obtain a signal x to be extracted 2 (t)；
S1.3, calculating the frequency spectrum characteristics
Fingerprint information is not uniformly distributed over the spectral features, and the distribution differences are more pronounced near the center frequency (limiting Sun, xiang Wang, zhitao Huang, and Baoguo Li, radio Frequency Fingerprint Extraction based on Feature Inhomogeneity. IEEE Internet of Things Journal,2022.DOI: 10.1109/JIOT.2022.3154595).
Therefore, the invention only reserves the frequency spectrum with a certain width near the main lobe by extracting the local frequency spectrum with the most abundant fingerprint information in the frequency spectrum as the fingerprint characteristic of the radiation source, and does not depend on the whole frequency spectrum information. The calculation method is simple and convenient, and compared with the method using all frequency spectrum information, the signal characteristic dimension is greatly reduced.
The method comprises the following steps:
s1.3.1 calculating the signal x to be feature extracted at 0 frequency 2 Spectrum f of (t) 0 (k)：
Wherein,,representing a fourier transform; k represents the frequency index value of the fourier transform and is also the dimension index value of the spectral feature; n (N) FFT Points representing fourier transforms; |·| represents the calculated amplitude value.
S1.3.2 the signal x to be extracted 2 Spectrum f of (t) 0 (k) The width of the main lobe 3dB is W B Taking out the bandwidth around peak value as lambda 1 W B As a preliminary spectral feature value f 1 (k),k＝1,...,K 2 Wherein lambda is 1 Weighting parameters, K, representing spectral ranges 2 Representing the spectral length after the cut-out, i.e. the preliminary spectral feature value f 1 (k) Is a length of (c). Experiments show that lambda 1 =2 can satisfy the recognitionRequirements.
S1.3.3 pair f 1 (k) Calculating the second derivativeSelect to satisfy->Is the minimum value k of (2) min Wherein lambda is 2 Represents the weight coefficient, lambda 2 ＞0。
Select f 1 (k) Upper [ k ] min ,K 2 ]The segment serves as the spectral feature f (k) to be amplified.
S2: extracting spectrum reference characteristics and analyzing characteristic components;
the method mainly comprises the steps of comprehensively utilizing signal characteristics of all known tags of all radiation sources extracted in the step S1, calculating common components of the signal characteristics as spectrum reference characteristics, amplifying details based on correlation operation, and finally completing component analysis based on self-adaptive decomposition to obtain intentional modulation quantity which is not beneficial to fingerprint identification of the radiation sources.
The output of this step is a spectral reference featureAnd an intentional modulation amount h (j). In practical applications, the main purpose of this stage is to obtain a spectral reference feature +.>And intentional modulation amount h (j), which is obtained based on all training data of all radiation sources, only once. The method comprises the following steps:
s2.1, calculating the spectrum reference characteristics
S1 shows a method for calculating the spectral characteristic f (k) of a signal sample to be amplified. Multiple signal samples from multiple sources are processed in the SEI problem. The baseline characteristic of the present invention is a common characteristic of all training samples of multiple radiation sources, and therefore multiple signal samples of multiple radiation sources need to be processed.
Assume that there are M wireless network cardsFor individual identification analysis of radiation sources, the mth radiation source has N m The spectral characteristics to be amplified of the ith sample of the signal samples can be expressed as f i m (k),i＝1,...,N m ,m＝1，…,M。
Here, f i m (k) I.e. the output f (k) of S1, the label of the corresponding radiation source is indicated by the upper corner mark m, and the lower corner mark i indicates the i-th signal sample of the current radiation source.
S2.1.1 the spectral characteristics to be amplified of all samples of all radiation sources are calculated and combined into a set { f } in the same way as S1 i m (k)},i＝1,...,N m ,m＝1，…,M。
S2.1.2 calculating the characteristic mean value of each radiation source
S2.1.3 weighted average to obtain spectrum reference characteristics
Wherein alpha is m Is the characteristic weight of the mth radiation source, meets the following conditionsThe characteristic weight of the radiation source is determined according to the data scale and the characteristic distribution condition of each radiation source. Normally +.>
Spectral reference featuresIs the common part obtained by computing all training samples for all radiation sources,what is considered to be a common part of the radiation source, i.e. the intentional modulation that is not beneficial for individual identification of the radiation source, is suppressed below.
S2.2, spectrum reference feature detail amplification
For the spectrum reference features obtained in S2.1Performing autocorrelation operation to amplify details and obtain standard spectral feature +.>
Wherein,,representing an autocorrelation operation. Here, the feature dimensions change, j=2k_1.
In this stepIs based on the weighted average of multiple radiation source signals-spectrum reference feature +.>The obtained product is symmetric in distribution and can represent the commonality of the radiation source. For the purpose of matching the spectral reference features in S2.1->Differentiation (I) of>Known as standard spectral features.
S2.3, adaptive decomposition of standard spectral features
Standard spectral features obtained in S2.2VMD decomposition is performed to obtain different components, which are called IMFs. VMD decomposition concrete implementation process reference (DRAGOMIRESSIYK, ZOSSO D.Variation Mode Decomposition, IEEE Trans.on Signal Processing,2014,62 (3): 531-544.DOI: 10.1109/tsp.2013.2288675).
Assuming that the number of decomposition layers is L, the obtained decomposition result can be expressed as:
v, V and omega are respectively the time domain combination of the decomposed IMF, the frequency domain combination and the center frequency of the IMF.
The time domain combination V of IMFs is expressed as:
V＝[v 1 (j),v 2 (j),…,v L (j)],j＝1,2,...,J
wherein v is l (j) L=1, 2,..l is a single IMF component
S2.4, analysis of spectral feature components
And (3) carrying out spectrum component analysis on the L IMF components obtained by VMD decomposition in S2.3, wherein the spectrum component analysis is specifically as follows:
s2.4.1 calculation of different IMF Components v l (j) Energy value of (2):
s2.4.2 according to { E l Sorting IMFs, wherein the highest energy IMF component is the principal componentThe low energy is used as stray component +.>
S2.4.3 the selected principal component componentsAnd stray component->Is expressed as +.>Obtaining the intentional modulation feature quantity h (j):
s2, calculating frequency spectrum reference characteristicsStandard spectral features->Is based on a cumulative weighting of a plurality of samples of a plurality of radiation sources and is therefore considered to be a common component common to all radiation sources.
The intentional modulation feature h (j) is calculated on the basis of these common components, considered to be common to the signals of the different radiation sources, and is an extraneous component that is not of interest to the SEI, which needs to be rejected. Obtaining h (j) means that analysis of the spectral feature components is completed.
S3: extracting the label-free data amplification characteristics;
the real purpose of the SEI is to realize the identification of the radiation source for the newly received no-tag signal. Therefore, the stage mainly performs analysis processing on the newly received signals, and extracts amplified final frequency spectrum fingerprint characteristics so as to improve the performance of individual identification of the radiation source.
Note that each newly received signal requires first repeating the data processing of S1 and spectral feature extraction, which is the basis for feature amplification.
Then, the spectrum characteristic of the new signal is combined with the spectrum reference characteristicAnd performing correlation operation, and finally, eliminating the intentional modulation quantity h (j) to obtain the amplified final frequency spectrum characteristic.
The output of this stage is the amplified final spectral signatureBecause the radiation source tag to which the signal corresponds is not known here, denoted by the symbol n. The method comprises the following steps:
s3.1, correlation transformation of the sample to be identified
S3.1.1 calculating the spectral characteristics f of the newly received signal according to S1 i n (k) A. The invention relates to a method for producing a fibre-reinforced plastic composite In the identification phase, the radiation source label is denoted by n, since the corresponding network card label is unknown. To match the spectral feature f calculated in S2 with the spectral feature f to be amplified with explicit tag information i m (k) To distinguish, here, the spectral features to be identified for the ith sample of the nth radiation source are denoted by f i n (k) And (3) representing. f (k), f i m (k)、f i n (k) The method is calculated based on the same method, and the footmarks represent the radiation source labels and sample number information corresponding to the signals.
S3.1.2 f i n (k) With spectral reference featuresPerforming cross-correlation operation to obtain new characteristic +.>Differential amplification is realized:
it should be noted that the features hereinAlthough formally matching the standard spectral features obtained in S2/>Similarly, but standard spectral features->Is a common component of all known samples, does not refer to the characteristic of a certain radiation source, only needs to be calculated once and only needs to be carried out once in the analysis of the spectral characteristic component, and therefore, the method has no upper and lower marks. And the novel features obtained here +>Is the current signal sample f i n (k) Features of (2), thus taking over feature f i n (k) N represents the label of the radiation source and i represents the sample number.
S3.2, final spectrum characteristic calculation after amplification
New features in S3.1Subtracting S2.4.3 to obtain the final amplified spectrum feature:
is to eliminate the intentional modulation common to all radiation sources and to obtain an amplified unintentional modulation characteristic.
The invention relates to an amplification technology of a wireless network card spectrum fingerprint characteristic. The method specifically refers to high-precision frequency estimation of a plurality of network card signals received in the wireless network signal detection process, extraction of spectrum fingerprint characteristics, and realization of detail amplification, wherein the problem of small individual characteristic difference of radiation sources among network card devices is solved. The specific method is that local frequency spectrums with more abundant fingerprint information are extracted as fingerprint characteristics, public parts of the frequency spectrum characteristics of the radiation sources are obtained by weighting and averaging the frequency spectrum characteristics of the radiation sources, main components and spurious components of the frequency spectrum characteristics are obtained based on self-adaptive decomposition, meanwhile, detail amplification in the frequency spectrum characteristics is realized based on related operation in the calculation process, the intentional modulation quantity is eliminated, and the obtained amplified final frequency spectrum characteristics lay a foundation for subsequently improving the identification success rate of network card equipment.
The invention has the following technical effects: the invention realizes the amplification of the characteristic of slight difference of the network card signal in frequency spectrum, and is mainly characterized in that:
(1) The local frequency spectrum with the most abundant fingerprint information in the signal frequency spectrum is extracted as the fingerprint characteristic of the radiation source, rather than relying on all information, particularly, the main lobe is removed, and only the frequency spectrum with a certain width near the main lobe is reserved; the calculation method is simple and convenient, and compared with the full length, the feature dimension is greatly reduced.
(2) Common components of fingerprint features of different radiation sources are obtained based on weighted average values of a large number of samples of the plurality of radiation sources, spectrum reference features and standard spectrum features are defined, and corresponding calculation methods are provided.
(3) The newly extracted fingerprint features of the radiation source are decomposed through a self-adaptive decomposition VMD algorithm, and components which are not beneficial to SEI (solid-state imaging device) are separated, including main components and spurious components which are intentionally modulated, and eliminated.
(4) And amplifying the fine difference in the fingerprint characteristics of the radiation source by using correlation operation.
(5) The calculation complexity is low, and the real-time performance is high.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention;
FIG. 2 is a signal spectral feature image;
FIG. 3 is a raw spectral feature;
FIG. 4 is a spectral reference feature autocorrelation;
FIG. 5 is a standard spectral feature decomposition result;
FIG. 6 is a cross-correlation of spectral features of different radiation sources with reference features;
FIG. 7 is a final feature map of the radiation source fingerprint feature after differential amplification;
fig. 8 is a partial view of the features of the radiation source with the fingerprint differences enlarged.
Detailed Description
The following describes the embodiments of the present invention further with reference to the drawings.
Fig. 1 is a flowchart of the implementation of the present invention, and the present invention provides a network card spectrum fingerprint feature amplifying method based on feature correlation and adaptive decomposition, which includes the following three major steps:
s1: data processing and spectrum feature extraction;
s2: extracting spectrum reference characteristics and analyzing characteristic components;
s3: extracting the label-free data amplification characteristics;
the method comprises the following steps:
firstly, processing the received data by S1.1-S1.3 to obtain the frequency spectrum characteristics to be amplified;
then, judging whether the spectrum reference feature calculation is needed, if so (namely, if no reference feature result exists), performing S2.1-S2.4 to obtain the reference featureAnd an intentional modulation amount h (j). Because the calculations of S2.1-S2.4 are based on the multiple sample cumulative weighting data of the multiple radiation sources and are thus common components common to all radiation sources.
Finally, the steps S1.1-S1.3 are repeated for the new signal received later, and S3.1 and S3.2 are performed to obtain the amplified final spectral characteristics.
Therefore, in the practical application process, S2.1-S2.4 only need to be carried out once, and corresponding reference characteristics are obtainedAnd the steps S2.1-S2.4 are not needed to be repeated after the intentional modulation quantity h (j), so that the calculation efficiency is improved.
FIG. 2 is a calculated (S1.1-S1.3) signal spectrumReference featuresThe horizontal axis represents the feature dimension and the vertical axis represents the corresponding feature value. The features here come from the right region of the main peak of the spectrum, with dimensions 4000. In the feature calculation process, lambda in S1.3.2 1 W B Lambda of (2) 1 λ in=2, s1.3.3 2 =10, the following chart calculates the parameters consistently. The result is a common part calculated based on 300 signal samples of a 3-part wireless network card, wherein each radiation source contains 100 signal samples, i.e. m=3, n m ＝100,m＝1,2,3。
Fig. 3 shows the original spectral characteristics obtained by calculating the plurality of radiation source signals, namely f (k) obtained by steps S1.1-S1.3, wherein different colors represent different network card individuals, and the emitted signals of different individuals are the same. The left panels represent the original spectral signature of the two radiation sources, the dark set of curves representing the signature of radiation source 1 and the light set representing the signature of radiation source 2; the right hand graph shows the addition of the features of the radiation source 3 to the original radiation sources 1 and 2, which are represented in the graph as the outermost second set of dark curves.
It can be seen that the signals of the different radiation sources differ in the original features extracted by the present invention and can be used for identification of the device. However, the main body shapes of these characteristic curves are uniform, and differ only in fine points. These subtle differences are of real interest for radiation source fingerprinting. The feature curves have high similarity due to the subtlity of the difference along with the increase of the number of the radiation source individuals, the feature curves are seriously overlapped with each other, and the recognition difficulty is greatly increased.
FIG. 4 is a reference featureStandard spectral characteristics amplified by detail information after autocorrelation +.>Is shown (S2.2). The same horizontal axis as in FIG. 2 and FIG. 3 represents feature dimensions, and the vertical axis represents featuresAnd (5) a sign value. The calculation is based on the spectral reference feature +.>Features at this time->And the characteristic dimension is 7999 with 4000 as the center and is bilaterally symmetrical. />Is a common part from which individual differences of the radiation sources are eliminated, and the shape of the main body can be used to represent the intended modulation component, which is the input of the component analysis of VMD decomposition in S2.3.
FIG. 5 is a standard spectral featureAfter VMD decomposition with l=3, graph of IMF in time domain (S2.3). It can be seen that after decomposition, different components IMF 1-IMF 3 are decomposed, wherein IMF1 has the strongest energy, which is +.>Whereas IMF3 has the lowest energy and highest frequency, is a spurious component.
FIG. 6 is a feature f calculated by steps S1.1-S1.3 for a newly received sample i n (k) With reference featuresResults of performing cross-correlation->The two sets of differently colored curves in the left panel of fig. 6 represent the characteristics of radiation source 1 (dark) and radiation source 2 (light), respectively, as defined by the curves in fig. 3; the newly added set of dark curves in the right panel represents the radiation source 3 compared to the left panel.
From the symmetry of the waveform on the left and right sides, it can be seen that compared with the standard spectral features(FIG. 4),>and not left-right symmetric. The difference is clearly seen to be exaggerated compared to the original spectral features in fig. 3. And after the target number is increased, the characteristic difference between different network cards is still obvious.
FIG. 7 is a final spectral feature of fingerprint differential amplification obtained after the inventive methodIs shown (S3.2). The characteristic obtained here is that the main component and the stray component h (j) which are not beneficial to the individual identification of the radiation source and are obtained after the VMD decomposition are eliminated, and only unintentional modulation information which can represent the difference between individuals is reserved, so that the amplification of fingerprint information is realized.
The meaning of the curves in the small pictures on the left and right sides in fig. 7 is identical to that in fig. 3 and 6. The curve sub-tables of different colours represent the amplified final spectra of the different radiation sourcesThe curves in the left graph represent the radiation source 1 (dark color) and the radiation source 2 (light color), respectively, and the right graph represents the features of the radiation source 3 (uppermost dark color curve) additionally plotted on the left basis.
The amplification effect is very pronounced compared to the original spectral features in fig. 3.
FIG. 8 is a feature of fingerprint difference amplificationIs a partial enlarged view of (c). The meaning of the curves in the figure corresponds to that of figure 7. The characteristic curves of different targets have different value ranges and different fluctuation rules, so that the distinguishing property is greatly enhanced.
The invention has the effect of performing verification analysis based on the signals transmitted by three wireless network cards of the same model of the IEEE802.11b standard which are actually acquired. The signal working frequency point is 2.4GHz, and the sampling rate is 10GHz. The header portion of the resulting signal frame is detected for analysis at 2500 points in data length. Fig. 2 to 8 are each a practical effect diagram. The human eyes can clearly see that the characteristic curves which are originally distributed together show obvious differences after being processed by the invention, and the original fine differences are obviously amplified.
Claims (5)
1. A network card frequency spectrum fingerprint feature amplifying method based on feature correlation and self-adaptive decomposition is characterized by comprising the following steps:
s1: data processing and spectral feature extraction
The method comprises the following steps:
s1.1 data reception and preprocessing
After the receiver receives the signal, the original signal is detected, filtered and noise-reduced to obtain a signal sample x to be processed 0 (t)；
Signal sample x to be processed 0 (t) performing an analysis and obtaining a frame header portion x of the signal sample to be processed by detection 1 (t)；
S1.2, signal frequency offset estimation and elimination
The method comprises the following steps:
s1.2.1 pair x based on Fourier interpolation algorithm 0 Frequency offset value f of (t) 0 Estimating;
s1.2.2 according to f 0 Will x 1 (t) moving to zero frequency to obtain a baseband signal, and eliminating the influence of frequency offset;
s1.2.3 removing the mean value of the baseband signal, and normalizing the energy to obtain a signal x to be extracted 2 (t)；
S1.3, calculating the frequency spectrum characteristics
The method comprises the following steps:
s1.3.1 calculating the signal x to be feature extracted at 0 frequency 2 Spectrum f of (t) 0 (k)：
Wherein,,representing a fourier transform; k represents the frequency index value of the fourier transform and is also the dimension index value of the spectral feature; n (N) FFT Points representing fourier transforms; |·| represents the calculated amplitude value;
s1.3.2 the signal x to be extracted 2 Spectrum f of (t) 0 (k) The width of the main lobe 3dB is W B Taking out the bandwidth around peak value as lambda 1 W B As a preliminary spectral feature value f 1 (k),k＝1,...,K 2 Wherein lambda is 1 Weighting parameters, K, representing spectral ranges 2 Representing the spectral length after the cut-out, i.e. the preliminary spectral feature value f 1 (k) Is a length of (2);
s1.3.3 pair f 1 (k) Calculating the second derivativeSelect to satisfy->Is the minimum value k of (2) min Wherein lambda is 2 Represents the weight coefficient, lambda 2 ＞0；
Select f 1 (k) Upper [ k ] min ,K 2 ]The segment serves as a spectral feature f (k) to be amplified;
s2: spectral reference feature extraction and feature component analysis
The method comprises the following steps:
s2.1, calculating the spectrum reference characteristics
Assuming that M wireless network cards are required to perform individual identification analysis of radiation sources, the mth radiation source has N m The spectral characteristics to be amplified of the ith sample of the signal samples can be expressed as f i m (k),i＝1,...,N m ,m＝1，...,M；
Here, f i m (k) I.e. the output f (k) of S1, is information representing the source and the signal sampleThe angle mark m represents the label of the corresponding radiation source, and the lower angle mark i represents the ith signal sample of the current radiation source;
s2.1.1 the spectral characteristics to be amplified of all samples of all radiation sources are calculated and combined into a set { f } in the same way as S1 i m (k)},i＝1,...,N m ,m＝1，...,M；
S2.1.2 calculating the characteristic mean value of each radiation source
S2.1.3 weighted average to obtain spectrum reference characteristics
Wherein alpha is m Is the characteristic weight of the mth radiation source, meets the following conditions
Spectral reference featuresThe common part obtained by calculating all training samples of all radiation sources is regarded as a common part of the radiation sources, namely the intentional modulation which is not beneficial to the individual identification of the radiation sources, and the part is restrained;
s2.2, spectrum reference feature detail amplification
For the spectrum reference features obtained in S2.1Performing autocorrelation operation to amplify details and obtain standard frequency spectrum characteristics
Wherein,,representing an autocorrelation operation; here, the feature dimensions change, j=2k—1;
in this stepIs based on the weighted average of multiple radiation source signals-spectrum reference feature +.>The obtained product is symmetric in distribution and can represent the commonality of the radiation source; for the purpose of matching the spectral reference features in S2.1->Differentiation (I) of>Known as standard spectral features;
s2.3, adaptive decomposition of standard spectral features
Standard spectral features obtained in S2.2Performing VMD decomposition to obtain different components, which are called IMFs;
assuming that the number of decomposition layers is L, the obtained decomposition result can be expressed as:
v, V and omega are respectively the time domain combination of the decomposed IMF, the frequency domain combination and the center frequency of the IMF;
the time domain combination V of IMFs is expressed as:
wherein v is l (j) L=1, 2,., L is a single IMF component;
s2.4, analysis of spectral feature components
And (3) carrying out spectrum component analysis on the L IMF components obtained by VMD decomposition in S2.3, wherein the spectrum component analysis is specifically as follows:
s2.4.1 calculation of different IMF Components v l (j) Energy value of (2):
s2.4.2 according to { E l Sorting IMFs, wherein the highest energy IMF component is the principal componentThe low energy is used as stray component +.>
S2.4.3 the selected principal component componentsAnd stray component->Is expressed as +.>Obtaining the intentional modulation feature quantity h (j):
h(j)＝∑ vl∈l v l (j)
s2, calculating frequency spectrum reference characteristicsStandard spectral features->Is based on a plurality of samples of a plurality of radiation sources, and is therefore considered as a common component common to all radiation sources;
the intentional modulation feature h (j) is calculated on the basis of the common components, is considered to be common to signals of different radiation sources, is an irrelevant component which is not concerned by SEI, and needs to be removed; obtaining h (j) means that analysis of the spectral feature components is completed;
s3: label-free data amplification feature extraction
The method comprises the following steps:
s3.1, correlation transformation of the sample to be identified
S3.1.1 calculating the spectral characteristics f of the newly received signal according to S1 i n (k) The method comprises the steps of carrying out a first treatment on the surface of the In the identification stage, as the corresponding network card label is unknown, the radiation source label is denoted by n; to match the spectral feature f calculated in S2 with the spectral feature f to be amplified with explicit tag information i m (k) To distinguish, here, the spectral features to be identified for the ith sample of the nth radiation source are denoted by f i n (k) A representation; f (k), f i m (k)、f i n (k) The method is calculated based on the same method, and the footmarks represent the radiation source labels and sample number information corresponding to the signals;
s3.1.2 f i n (k) With spectral reference featuresPerforming cross-correlation operation to obtain new characteristic +.>Differential amplification is realized:
s3.2, final spectrum characteristic calculation after amplification
New features in S3.1Subtracting S2.4.3 to obtain the final amplified spectrum feature:
is to eliminate the intentional modulation common to all radiation sources and to obtain an amplified unintentional modulation characteristic.
2. A method for amplifying spectral fingerprint characteristics of a network card based on characteristic correlation and adaptive decomposition according to claim 1, wherein: s1.2 based on x longer in length in frequency offset estimation 0 (t) frequency offset estimation can improve estimation accuracy.
3. A method for amplifying spectral fingerprint characteristics of a network card based on characteristic correlation and adaptive decomposition according to claim 1, wherein: s1.3.2 weighting parameters lambda of spectral ranges 1 ＝2。
4. A method for amplifying spectral fingerprint characteristics of a network card based on characteristic correlation and adaptive decomposition according to claim 1, wherein: s2.1.3 the mth radiation source characteristic weight alpha m According to the data size, eachAnd determining the characteristic distribution condition of the radiation source.
5. A method for amplifying spectral fingerprint characteristics of a network card based on characteristic correlation and adaptive decomposition according to claim 4, wherein: mth radiation source characteristic weight
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210854425.3A CN115186714B (en) | 2022-07-15 | 2022-07-15 | Feature Amplification Method of NIC Spectrum Fingerprint Based on Feature Correlation and Adaptive Decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210854425.3A CN115186714B (en) | 2022-07-15 | 2022-07-15 | Feature Amplification Method of NIC Spectrum Fingerprint Based on Feature Correlation and Adaptive Decomposition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115186714A CN115186714A (en) | 2022-10-14 |
| CN115186714B true CN115186714B (en) | 2023-07-21 |
Family
ID=83520227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210854425.3A Active CN115186714B (en) | 2022-07-15 | 2022-07-15 | Feature Amplification Method of NIC Spectrum Fingerprint Based on Feature Correlation and Adaptive Decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115186714B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115510924B (en) * | 2022-11-17 | 2023-04-07 | 中铁第一勘察设计院集团有限公司 | Radio frequency fingerprint identification method based on improved variational modal decomposition |
| CN117765575B (en) * | 2024-01-12 | 2024-06-21 | 中国人民解放军国防科技大学 | A method for fingerprint identification of radiation sources across receiving systems based on multi-receiver collaboration |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109307862A (en) * | 2018-07-05 | 2019-02-05 | 西安电子科技大学 | A method for individual identification of target radiation source |
| CN110991378A (en) * | 2019-12-11 | 2020-04-10 | 中国人民解放军空军工程大学 | Power amplifier individual identification method and device |
| CN111814703A (en) * | 2020-07-13 | 2020-10-23 | 四川大学 | A joint feature extraction method for signals based on HB under non-reconstruction conditions |
| CN114255487A (en) * | 2021-12-22 | 2022-03-29 | 安徽农业大学 | Internet of things equipment identity authentication method for semi-supervised radio frequency fingerprint extraction |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8160163B1 (en) * | 2007-08-06 | 2012-04-17 | University Of South Florida | Method for OFDM signal identification and parameter estimation |
| CN106792706A (en) * | 2016-12-09 | 2017-05-31 | 中国人民解放军国防科学技术大学 | A kind of illegal invasion detection method based on wireless network signal spectrum non-stationary property |
| CN109150340B (en) * | 2018-07-28 | 2020-11-24 | 中国人民解放军战略支援部队信息工程大学 | Fingerprint feature extraction method, identity detection method and radiation source identification correction method |
| CN113219431A (en) * | 2021-04-30 | 2021-08-06 | 北京理工大学 | Interference source individual identification method based on waveform fingerprints |
| CN113219415A (en) * | 2021-04-30 | 2021-08-06 | 北京理工大学 | Interference source individual identification method based on envelope fingerprint characteristics |
-
2022
- 2022-07-15 CN CN202210854425.3A patent/CN115186714B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109307862A (en) * | 2018-07-05 | 2019-02-05 | 西安电子科技大学 | A method for individual identification of target radiation source |
| CN110991378A (en) * | 2019-12-11 | 2020-04-10 | 中国人民解放军空军工程大学 | Power amplifier individual identification method and device |
| CN111814703A (en) * | 2020-07-13 | 2020-10-23 | 四川大学 | A joint feature extraction method for signals based on HB under non-reconstruction conditions |
| CN114255487A (en) * | 2021-12-22 | 2022-03-29 | 安徽农业大学 | Internet of things equipment identity authentication method for semi-supervised radio frequency fingerprint extraction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115186714A (en) | 2022-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115186714B (en) | Feature Amplification Method of NIC Spectrum Fingerprint Based on Feature Correlation and Adaptive Decomposition | |
| Tang et al. | FingerNet: An unified deep network for fingerprint minutiae extraction | |
| CN114580476B (en) | A method for constructing a recognition model of drone signals and a corresponding recognition method and system | |
| Liu | Multi-feature fusion for specific emitter identification via deep ensemble learning | |
| CN114118131A (en) | Multi-feature fusion wireless device RF fingerprint extraction method based on attention mechanism | |
| CN101847210A (en) | Multi-group image classification method based on two-dimensional empirical modal decomposition and wavelet denoising | |
| CN108364277A (en) | A kind of infrared small target detection method of two-hand infrared image fusion | |
| CN114896887B (en) | Frequency-using equipment radio frequency fingerprint identification method based on deep learning | |
| CN111050315B (en) | A wireless transmitter identification method based on multi-core dual-channel network | |
| CN112749633A (en) | Separate and reconstructed individual radiation source identification method | |
| CN104008386A (en) | Method and system for identifying type of tumor | |
| Huang et al. | Radar waveform recognition based on multiple autocorrelation images | |
| CN114255487A (en) | Internet of things equipment identity authentication method for semi-supervised radio frequency fingerprint extraction | |
| CN107610118A (en) | One kind is based on dMImage segmentation quality evaluating method | |
| Kusban et al. | Excellent performance of palmprint recognition by using wavelet filter | |
| CN114863189B (en) | Intelligent image identification method based on big data | |
| CN113780521A (en) | Radiation source individual identification method based on deep learning | |
| CN115169406B (en) | Instantaneous phase fingerprint feature enhancement method based on empirical mode decomposition | |
| CN109271902B (en) | Infrared weak and small target detection method based on time domain empirical mode decomposition under complex background | |
| CN117574210A (en) | A radio frequency fingerprint identification method based on domain adversarial network | |
| CN106326827A (en) | Palm vein recognition system | |
| CN115278518B (en) | Indoor positioning method for channel state information based on subcarrier selection | |
| CN117593261A (en) | An image hidden target detection method | |
| CN111404852B (en) | Modulation mode identification method based on amplitude and spectral amplitude characteristics | |
| CN115809426A (en) | Radiation source individual identification method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |