An IoT device identity authentication method based on online learning

Abstract

The present application relates to the field of device security technology, and in particular to an Internet of Things device identity authentication method based on online learning. It includes: obtaining a number of target device signals of a target Internet of Things device within a target time window; performing feature extraction on a number of the target device signals to obtain an initial radio frequency fingerprint of the target Internet of Things device; reducing the dimension of the initial radio frequency fingerprint according to a semi-supervised dimensionality reduction algorithm to obtain a target radio frequency fingerprint; inputting the target radio frequency fingerprint into an online learning authentication model to obtain an identity authentication result of the target Internet of Things device. The present application improves the accuracy of Internet of Things device identity authentication and solves the problem of high-dimensional data compression and real-time and efficient authentication in a dynamic environment.

Description

Background Art

[0002] In the era of the Internet of Things, communication networks are changing with each passing day. The security of IoT devices faces a series of new challenges. In order to adapt to the complex application environment, in recent years, the identity authentication method based on Radio Frequency Fingerprint (RFF) has received widespread attention. RF fingerprint is a unique feature generated by hardware differences when IoT devices transmit wireless signals. It has the characteristics of uniqueness and difficulty in forging, and can be used for device identity identification. However, there are still some problems with the existing identity authentication methods based on RF fingerprints: on the one hand, in the process of extracting RF fingerprints, a large number of device signals usually need to be collected, which not only increases the burden of data processing, but also may be affected by environmental noise and interference signals, resulting in low accuracy of the extracted RF fingerprints. For example, in complex electromagnetic environments, such as factory workshops and large shopping malls, the interference signals generated by various electronic devices will seriously affect the extraction effect of RF fingerprints, making the subsequent identity authentication results unreliable. On the other hand, most of the existing RF fingerprint processing methods use traditional dimensionality reduction algorithms. When processing high-dimensional RF fingerprint data, these algorithms often cannot make full use of the information contained in limited labeled data and a large amount of unlabeled data, resulting in the loss of some important discriminant information in the RF fingerprint after dimensionality reduction, thereby affecting the accuracy of identity authentication. In addition, traditional identity authentication models are usually trained offline and cannot adapt to the state changes of IoT devices and emerging attack methods in real time, and lack flexibility and adaptability. Summary of the invention

[0003] The present application provides an Internet of Things device identity authentication method based on online learning to solve at least one problem in the prior art.

[0004] According to a first aspect of the present application, there is provided an Internet of Things device identity authentication method based on online learning, comprising the following steps:

[0005] S100, obtaining a plurality of target device signals of a target IoT device within a target time window; wherein the target time window has N consecutive target time periods, each target time period has a corresponding target working frequency band; the target working frequency band is one of a plurality of working frequency bands; the target working frequency band is determined according to the frequency band information of each working frequency band within a historical time window; the end time of the historical time window is the current time;

[0006] S200, extracting features from signals of a plurality of target devices to obtain an initial radio frequency fingerprint of the target IoT device;

[0007] S300, reducing the dimension of the initial radio frequency fingerprint according to a semi-supervised dimensionality reduction algorithm to obtain a target radio frequency fingerprint;

[0008] S400: Input the target radio frequency fingerprint into an online learning authentication model to obtain an identity authentication result of the target Internet of Things device.

[0009] This application has at least the following beneficial effects:

[0010] The present application provides an IoT device identity authentication method based on online learning. During signal acquisition, compressed sensing is used to utilize signal sparsity to acquire signals at a sampling rate far lower than the Nyquist sampling rate, thereby reducing the amount of data, improving acquisition efficiency, and reducing hardware sampling frequency requirements, saving hardware costs. In addition, the spectrum environment is monitored in real time, and the acquisition frequency band is intelligently switched according to the frequency band selection index to effectively avoid interference, improve signal acquisition quality and stability, and is particularly suitable for IoT device signal acquisition in complex electromagnetic environments. Subsequently, for accurately acquired signals, radio frequency fingerprint recognition technology is introduced to improve the security and authentication efficiency of IoT devices in wireless communication environments. In addition, key technologies such as supervised dimensionality reduction methods, online learning algorithms, and dynamic classification models are utilized to improve the accuracy of IoT device identity authentication, solving the problem of high-dimensional data compression and real-time and efficient authentication in dynamic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A flowchart of an Internet of Things device identity authentication method based on online learning is provided for one embodiment of the present application. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0014] like Figure 1 As shown, an embodiment of the present application provides an IoT device identity authentication method based on online learning, comprising:

[0015] S100, obtaining several target device signals of the target IoT device within a target time window; wherein the target time window has N consecutive target time periods, each target time period has a corresponding target working frequency band; the target working frequency band is one of several working frequency bands; the target working frequency band is determined according to the frequency band information of each working frequency band in a historical time window; the end time of the historical time window is the current time.

[0016] Wherein, the step S100 includes:

[0017] S110, obtaining frequency band information of each working frequency band corresponding to the target IoT device within a historical time window; wherein the frequency band information includes signal strength and interference signal strength; and the end time of the historical time window is the current time.

[0018] Specifically, the target IoT devices include smart home devices, industrial IoT devices, etc. The target IoT device in this embodiment is any IoT device. In order to meet different connection requirements or realize a wider range of functions, each IoT device will integrate multiple wireless communication technologies. Each technology may work in different frequency bands, that is, each IoT device will have multiple working frequency bands. As an example: the smart speaker in the smart home must connect to the home network through Wi-Fi (2.4GHz or 5GHz frequency band) to obtain online music, voice interaction and other services, and may use Bluetooth (2.4GHz frequency band) to connect to mobile phones, tablets and other devices at close range to realize functions such as audio transmission or device configuration. In this way, as an IoT device, the smart speaker will transmit signals on multiple frequency bands. Or some industrial IoT gateway devices may support ZigBee (2.4GHz frequency band) for connecting low-power, short-distance sensor networks, and also support LoRa (different frequency bands in different regions, such as 433MHz, 868MHz, 915MHz, etc.) for long-distance, low-rate data transmission to achieve monitoring and data collection of devices in a large area. This application obtains the signal strength and interference signal strength of each working frequency band within a historical time window. Here, the interference signal may come from other IoT devices, wireless communication terminals, wireless access points, or lightning in the natural environment.

[0019] S120, obtaining a frequency band indicator parameter corresponding to each working frequency band according to the frequency band information corresponding to each working frequency band.

[0020] Specifically, the frequency band index parameter corresponding to each working frequency band is obtained according to the signal strength and interference signal strength corresponding to each working frequency band. Here, the frequency band index parameter corresponding to each working frequency band is proportional to the corresponding signal strength and inversely proportional to the interference signal strength. That is, if the frequency band index parameter is larger, it means that the signal strength of the corresponding working frequency band is better and the interference is less. On the contrary, if the frequency band index parameter is smaller, it means that the signal strength of the corresponding working frequency band is relatively weak and the interference is more serious.

[0021] S130, obtaining a target working frequency band according to the frequency band indicator parameter; wherein the frequency band indicator parameter corresponding to the target working frequency band is the largest frequency band indicator parameter among the frequency band indicator parameters of all working frequency bands corresponding to the target IoT device.

[0022] Specifically, the working frequency band corresponding to the largest frequency band index parameter is selected as the target working frequency band, that is, the ratio of the signal strength corresponding to the target working frequency band to the interference signal strength is the largest. Then, the quality of the signal collected by the target working frequency band is the best.

[0023] S140, according to the target working frequency band, collecting the initial device signal of the target IoT device on the target working frequency band within the target time period; wherein the start time of the target time period is the current time; and the target time window includes N consecutive target time periods.

[0024] In this embodiment, the target time window to be collected is divided into a number of target time periods, and a working frequency band is selected each time to collect the corresponding initial device signal within the target time period.

[0025] S150, performing sparse processing, compressive sampling and reconstruction on the initial device signal to obtain a reconstructed original signal.

[0026] Specifically, the signals generated by IoT devices are sparse in certain transform domains (such as wavelet transform domain and Fourier transform domain). Through sparse transform, the signal is converted into a sparse representation in which most coefficients are zero or close to zero, which can remove a lot of redundant information. Then, the sparsely processed signal is compressed and sampled, which greatly reduces the amount of data while ensuring the key information of the signal, thus reducing the burden for subsequent processing. Finally, reconstruction is performed, and the reconstructed signal can be used for subsequent analysis, and also provides data support for evaluating the acquisition effect of the current frequency band.

[0027] S160, obtaining quality indicator parameters of the reconstructed original signal.

[0028] Specifically, the quality of the reconstructed signal is evaluated, for example, the signal bit error rate, signal-to-noise ratio and other indicators are calculated to determine the interference status of the signal in the current acquisition frequency band and the acquisition effect.

[0029] S170, if the quality indicator parameter is less than the preset quality indicator parameter threshold, jump to step S110; if the quality indicator parameter is equal to or greater than the preset quality indicator parameter threshold, jump to step S140, until all N target time periods are collected to obtain a number of target device signals.

[0030] Specifically, if the quality indicator parameter is less than the preset quality indicator parameter threshold, it means that the evaluation result shows that the signal quality of the current frequency band is not good. At this time, jump to step S110, that is, reselect the target working frequency band for signal acquisition in the next target time period. On the contrary, if the quality indicator parameter is equal to or greater than the preset quality indicator parameter threshold, it means that the signal quality meets the requirements, and the signal can continue to be collected in the current frequency band according to the above steps. Until all N target time periods are collected, a number of target device signals are obtained.

[0031] The signal acquisition method provided in this embodiment uses compressed sensing and signal sparsity to acquire signals at a sampling rate far lower than the Nyquist rate, thereby reducing the amount of data, improving acquisition efficiency, and reducing hardware sampling frequency requirements, saving hardware costs. In addition, the spectrum environment is monitored in real time, and the acquisition frequency band is intelligently switched according to the frequency band selection index, effectively avoiding interference, improving signal acquisition quality and stability, and is particularly suitable for IoT device signal acquisition in complex electromagnetic environments.

[0032] It should be noted that the above target device signal is extracted from a binary image obtained by short-time Fourier transform and threshold determination of the collected signal.

[0033] Since the distribution range of the collected signals in the time-spectrum diagram is limited, most of them appear in the form of a rectangular distribution with a certain time width and frequency width. According to the invariance of these parameters, a matching pattern window with a certain time width and frequency width is designed. The target signal is detected by moving the matching window on the entire binary image. Finally, the detected target signal is restored to a time series, that is, the target device signal is obtained. It is stored in a computer for subsequent signal feature extraction.

[0034] S200, extracting features from signals of a plurality of target devices to obtain an initial radio frequency fingerprint of the target IoT device.

[0035] Specifically, this embodiment further measures the nonlinear complexity of the extracted features based on the extraction of specific radio frequency fingerprints to obtain the initial radio frequency fingerprint of the target IoT device.

[0036] Furthermore, the feature extraction includes: feature extraction based on variational mode decomposition, feature extraction based on vertical synchronous compression transform, feature extraction based on bispectral Radon transform, feature extraction based on intrinsic time scale decomposition, feature extraction based on pulse unintentional modulation and feature extraction based on singular spectrum analysis.

[0037] Among them, feature extraction based on variational mode decomposition (VMD): the received signal can be decomposed into three modes. Then the output of VMD is the time mode, spectrum mode and center frequency mode.

[0038] The output function of VMD is: x(t)=Σ K k=1 h k (t); where h k (t) is the kth mode function; x(t) is the output function of VMD at time t.

[0039] Feature extraction based on Vertical Sync Compression Transform (VSST): SST is a post-time-frequency processing tool. The acquired signal contains a large number of complex modulated signals. The effect of first-order sync compression will be better only when the second-order derivative of the phase with respect to time is approximately zero or much smaller than its first-order derivative. VSST compresses the energy distributed in a frequency range into certain instantaneous frequency points, compresses the spectrum of the time-frequency distribution, and obtains a more accurate instantaneous frequency estimate.

[0040] Feature extraction based on Bispectrum-Radon Transform (BRT): Bispectrum-Radon Transform (BRT) is a method that combines bispectrum and Radon transform. In two-dimensional Euclidean space, "Radon transform" can project a matrix to the direction of a one-dimensional vector, so a one-dimensional feature vector will be obtained after the transformation.

[0041] Feature extraction based on intrinsic time-scale decomposition (ITD): Intrinsic time-scale decomposition (ITD) decomposes the original signal into a baseline signal component, i.e., a low-frequency component, and a rotation signal component, i.e., a high-frequency component. The baseline signal is then decomposed again as the original signal until the obtained baseline signal is a monotonic signal. The expression of intrinsic time-scale decomposition (ITD) is: ITD(t)=b(t)+r(t); where b(t) is the baseline signal component; r(t) is the rotation signal component.

[0042] Feature extraction based on Unintentional Modulation On Pulse (UMOP): The changes in the instantaneous amplitude, instantaneous phase, and instantaneous frequency within the signal pulse envelope are subtle features of a type of unintentional pulse modulation (UMOP). First, the original signal is Hilbert transformed to obtain a complex signal, and its transformation expression is: z(t)=c(t)+jy(t); where c(t) is the original signal; jy(t) is the analytical signal after Hilbert transform. Then, the amplitude response, phase response, and frequency response of the signal are calculated. The variance, skewness, and kurtosis of the amplitude, phase, and frequency are calculated respectively, and then combined into a feature vector of a single sub-region signal. Then, the feature vectors of the signals in each sub-region are combined to obtain the signal unintentional pulse modulation feature.

[0043] Feature extraction based on singular spectrum analysis (SSA): Singular spectrum analysis (SSA) is mainly used to study nonlinear time series data. The trajectory matrix of the sequence is constructed according to the observed time series, and the trajectory matrix is ​​embedded, singular value decomposed, grouped and diagonally averaged to extract different component signal components in the original time series.

[0044] S300, reducing the dimension of the initial radio frequency fingerprint according to a semi-supervised dimensionality reduction algorithm to obtain a target radio frequency fingerprint.

[0045] Specifically, since the extracted feature dimension is relatively high, a semi-supervised dimensionality reduction method is needed to reduce the dimension. The semi-supervised dimensionality reduction method corresponds to two cost functions, and its purpose is to make samples of different categories have a larger inter-class distance in the projection space, while allowing samples of the same category to have a smaller intra-class distance. This application intends to use similar order-preserving discriminant analysis based on the graph embedding framework.

[0046] For any two given samples d x and d y , the standardized similarity calculation formula between them is:

[0047] S xy =exp(-γ||d x -d y ||2 2 );where exp is an exponential function with base e;||()||2 2 is the calculation formula of the preset Euclidean distance square; γ>0, γ is the control sample d x and d y The similarity parameter between them, 0<S xy ≤1; x, y = 1, 2, ..., n; when d x =dy When S xy =1. Assume that for any sample d x , whose projection in low-dimensional space is represented by g x =P T d x , for any given sample d x d y and d n , their projections in low-dimensional space are g x , g y and g n If S xy ≤S xn , then the distance e xy =||g x -g n || 2 2=e xn , then the projection matrix P can keep the sample similarities in order.

[0048] In the projection space, if the similar order of samples belonging to the same class is maintained, a small intra-class distance is also required to improve the compactness of the samples. Based on the graph embedding framework, using the idea of ​​MMC, a method is used to balance the intra-class similarity order divergence and inter-class similarity divergence, thereby obtaining the cost function of SOPDA. The optimal solution of the cost function can be obtained through eigenvalue decomposition.

[0049] S400: Input the target radio frequency fingerprint into an online learning authentication model to obtain an identity authentication result of the target Internet of Things device.

[0050] Specifically, the online learning certification model is an online learning certification model based on Markov sampling. Furthermore, it is an online learning certification model based on Markov sampling one-to-many SVM (OAA-SVM-MS). If the state space of the Markov chain is finite, all irreducible non-periodic Markov chains are uniformly ergodic. For a finite data set, if the transition probability between any two states is always positive, the Markov chain is also uniformly ergodic. In order to meet the real-time update of the model structure and achieve fast and accurate recognition, an online learning algorithm will be added to the classifier. Adopting the one-to-one idea, the voting method is used to realize the decision-making of the model.

[0051] When a new IoT device arrives, if the category of the IoT device is known, it is determined whether it meets the Karush-Kuhn-Tucker (KKT) condition. If the KKT condition is met, it means that the IoT device maintains the classification boundary of the existing model, so there is no need to update the model. If the KKT condition is not met, it means that new knowledge can be learned from the IoT device, and the classification boundary of the existing model can be adjusted to update the model. If the category of the IoT device is unknown, a new category needs to be added to the model and its classification boundary needs to be readjusted. Finally, illegal IoT devices can be quickly and accurately identified based on the online learning authentication model.

[0052] The IoT device identity authentication method based on online learning provided in this embodiment uses compressed sensing during signal acquisition, and utilizes signal sparsity to acquire signals at a sampling rate far lower than the Nyquist sampling rate, thereby reducing the amount of data, improving acquisition efficiency, and reducing hardware sampling frequency requirements, saving hardware costs. In addition, the spectrum environment is monitored in real time, and the acquisition frequency band is intelligently switched according to the frequency band selection index to effectively avoid interference, improve signal acquisition quality and stability, and is particularly suitable for IoT device signal acquisition in complex electromagnetic environments. Subsequently, for accurately acquired signals, radio frequency fingerprint recognition technology is introduced to improve the security and authentication efficiency of IoT devices in wireless communication environments. In addition, key technologies such as supervised dimensionality reduction methods, online learning algorithms, and dynamic classification models are utilized to improve the accuracy of IoT device identity authentication, solving the problem of high-dimensional data compression and real-time and efficient authentication in dynamic environments.

[0053] In an exemplary embodiment of the present application, the i-th working frequency band f i The corresponding frequency band index parameter R(f i ) meets the following conditions:

[0054]

[0055] Among them, S(f i ) is f i The average signal strength in the historical time window; S stable (f i ) is f i The corresponding stable value of signal strength in the historical time window; I(f i ) is f i The average interference signal strength in the historical time window; ε is the frequency band adjustment parameter; α is the fading weight; β is the historical weight; F fade (f i ) is f i The corresponding fading factor; H used (f i ) is f i The corresponding number of historical transmission completions; H total (fi ) is f i The corresponding total number of historical transmissions. stable (f i ) meets the following conditions:

[0056] S stable (f i )=1 / (σ(f i )+1)

[0057] Among them, σ(f i ) is f i The variance of the corresponding signal strength within the historical time window.

[0058] Specifically, R(f i ) is larger, the more suitable the corresponding working frequency band is to be selected. i ) value, the greater the positive impact on the frequency band selection. The variance of the signal strength corresponding to the historical time window σ(f i ) is smaller, which means that data transmission in this frequency band is more reliable and less likely to experience sudden signal deterioration or interruption. stable (f i ) is larger, that is, the signal strength is more stable. Therefore, in this embodiment, R(f i ) and S(f i ) and S stable (f i ), and inversely proportional to the average interference signal strength, that is, the stronger the interference signal, the greater the negative impact on the frequency band selection. And F fade (f i ) reflects the degree of signal fading, and the larger the value, the more serious the fading. In this embodiment, the number of successful data transmissions in the frequency band in the historical time is also considered.

[0059] In summary, the frequency band index parameters of this embodiment comprehensively consider multiple factors such as signal strength, frequency band stability, fading characteristics, and historical usage, and can more comprehensively evaluate the pros and cons of the frequency band, thereby selecting a more suitable frequency band for IoT device signal acquisition.

[0060] In an exemplary embodiment of the present application, step S150 includes:

[0061] S151, transform the initial device signal according to a sparse transformation matrix to obtain a sparse representation coefficient vector.

[0062] Specifically, the transformation can be performed through a wavelet transform domain or a Fourier transform domain, etc. For the initial device signal, the transformation is performed based on a determined sparse transform matrix to obtain a sparse representation coefficient vector. This step converts the collected signal into a sparse representation form, creating conditions for compressed sampling, because compressed sensing is only feasible if the signal has sparseness in a certain transform domain.

[0063] S152, performing linear projection measurement on the sparse representation coefficient vector according to a measurement matrix to obtain a measurement value vector; wherein the measurement matrix is ​​irrelevant to the sparse transformation.

[0064] Specifically, a measurement matrix that is unrelated to the sparse transformation is selected, such as a Gaussian random matrix or a Bernoulli random matrix.

[0065] The selected measurement matrix is ​​used to perform linear projection measurement on the sparse representation coefficient vector to obtain the measurement value vector. Through this step, the amount of data is greatly reduced while ensuring the key information of the signal, which reduces the burden for subsequent processing.

[0066] S153, obtaining a reconstructed original signal according to the measurement value vector and a preset reconstruction algorithm.

[0067] According to actual needs and signal characteristics, select a suitable reconstruction algorithm, such as the orthogonal matching pursuit (OMP) algorithm or the compressed sampling matching pursuit (CoSaMP) algorithm. Using the measurement value vector obtained by compressed sampling as input, the selected reconstruction algorithm is used to gradually restore the sparse representation of the signal. The reconstructed signal can be used for subsequent analysis and also provides data support for evaluating the acquisition effect of the current frequency band.

[0068] In an exemplary embodiment of the present application, the quality indicator parameter CQI meets the following conditions:

[0069] CQI=SNR 2 × (1-BER) / (ISI+δ)

[0070] Among them, SNR is the signal-to-noise ratio corresponding to the target working frequency band in the target time period; BER is the bit error rate corresponding to the target working frequency band in the target time period; ISI is the inter-symbol interference corresponding to the target working frequency band in the target time period; δ is the quality adjustment parameter.

[0071] Specifically, the quality indicator parameter CQI is used to evaluate the quality of the reconstructed signal. The higher the value, the better the signal quality and the better the acquisition effect. SNR reflects the ratio of the effective signal power to the noise power in the signal. A larger SNR means that the signal is stronger relative to the noise and less affected by interference. BER refers to the ratio of the number of error symbols to the total number of symbols during transmission. The lower the BER, the higher the accuracy of signal transmission. ISI measures the degree of mutual interference between previous and next symbols due to factors such as multipath propagation. The smaller the ISI, the smaller the interference between symbols. δ avoids the denominator being zero.

[0072] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0073] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by combining software with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0074] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

[0075] Those skilled in the art will appreciate that various aspects of the present application may be implemented as a system, method or program product. Therefore, various aspects of the present application may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to as "circuit", "module" or "system" herein.

[0076] The electronic device according to this embodiment of the present application is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0077] The electronic device is presented in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: the at least one processor mentioned above, the at least one storage device mentioned above, and a bus connecting different system components (including storage devices and processors).

[0078] The storage device stores program codes, which can be executed by the processor, so that the processor executes the steps described in the “exemplary method” section of the specification according to various exemplary embodiments of the present application.

[0079] The memory may include readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory, and may further include read only memory (ROM).

[0080] The storage may also include a program / utility having a set (at least one) of program modules, such program modules including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0081] The bus may represent one or more of several types of bus structures including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0082] The electronic device may also communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may communicate with one or more devices that enable a user to interact with the electronic device, and / or may communicate with any device that enables the electronic device to communicate with one or more other computing devices (e.g., routers, modems, etc.). This communication may be performed through an input / output (I / O) interface. In addition, the electronic device may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) through a network adapter. The network adapter communicates with other modules of the electronic device through a bus. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0083] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0084] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present application can also be implemented in the form of a program product, which includes a program code. When the above program product is run on a terminal device, the above program code is used to enable the above terminal device to execute the steps according to various exemplary implementations of the present application described in the above "Exemplary Method" section of the present specification.

[0085] The above-mentioned program product can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0086] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0087] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0088] Program code for performing the operations of the present application may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0089] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present application, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0090] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0091] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An Internet of Things device identity authentication method based on online learning, characterized in that: The method comprises: S100, obtaining a plurality of target device signals of a target IoT device within a target time window; wherein the target time window has N consecutive target time periods, each target time period has a corresponding target working frequency band; the target working frequency band is one of a plurality of working frequency bands; the target working frequency band is determined according to the frequency band information of each working frequency band within a historical time window; the end time of the historical time window is the current time; S200, extracting features from signals of a plurality of target devices to obtain an initial radio frequency fingerprint of the target IoT device; S300, reducing the dimension of the initial radio frequency fingerprint according to a semi-supervised dimensionality reduction algorithm to obtain a target radio frequency fingerprint; S400, inputting the target radio frequency fingerprint into an online learning authentication model to obtain an identity authentication result of the target IoT device; Step S100 includes: S110, obtaining frequency band information of each working frequency band corresponding to the target IoT device within a historical time window; wherein the frequency band information includes signal strength and interference signal strength; S120, obtaining a frequency band indicator parameter corresponding to each working frequency band according to the frequency band information corresponding to each working frequency band; S130, obtaining a target working frequency band according to the frequency band index parameter; wherein the frequency band index parameter corresponding to the target working frequency band is the largest frequency band index parameter among the frequency band index parameters of all working frequency bands corresponding to the target IoT device; S140, according to the target working frequency band, collecting the initial device signal of the target IoT device on the target working frequency band within the target time period; wherein the start time of the target time period is the current time; and the target time window includes N consecutive target time periods; S150, performing sparse processing, compressive sampling and reconstruction on the initial device signal to obtain a reconstructed original signal; S160, obtaining quality indicator parameters of the reconstructed original signal; S170, if the quality indicator parameter is less than the preset quality indicator parameter threshold, jump to step S110; if the quality indicator parameter is equal to or greater than the preset quality indicator parameter threshold, jump to step S140, until all N target time periods are collected to obtain a number of target device signals.

2. The method for authenticating an IoT device based on online learning according to claim 1, characterized in that: The i-th working frequency band f i The corresponding frequency band index parameter R(f i ) meets the following conditions: ； Among them, S(f i ) is f i The average signal strength in the historical time window; S stable (f i ) is f i The corresponding stable value of signal strength in the historical time window; I(f i ) is f i The average interference signal strength in the historical time window; ε is the frequency band adjustment parameter; α is the fading weight; β is the historical weight; F fade (f i ) is f i The corresponding fading factor; H used (f i ) is f i The corresponding number of historical transmission completions; H total (f i ) is f i The corresponding total number of historical transmissions.

3. The method for authenticating an IoT device based on online learning according to claim 2, characterized in that: S stable (f i ) meets the following conditions: S stable (f i )=1 / (σ(f i )+1); Among them, σ(f i ) is f i The variance of the corresponding signal strength within the historical time window.

4. The method for authenticating an IoT device based on online learning according to claim 1, characterized in that: Step S150 includes: S151, transforming the initial device signal according to a sparse transformation matrix to obtain a sparse representation coefficient vector; S152, performing linear projection measurement on the sparse representation coefficient vector according to the measurement matrix to obtain a measurement value vector; wherein the measurement matrix is ​​irrelevant to the sparse transformation; S153, obtaining a reconstructed original signal according to the measurement value vector and a preset reconstruction algorithm.

5. The method for authenticating an IoT device based on online learning according to claim 1, characterized in that: The quality indicator parameter CQI meets the following conditions: CQI = SNR 2 × (1-BER) / (ISI+δ); Among them, SNR is the signal-to-noise ratio corresponding to the target working frequency band in the target time period; BER is the bit error rate corresponding to the target working frequency band in the target time period; ISI is the inter-symbol interference corresponding to the target working frequency band in the target time period; δ is the quality adjustment parameter.

6. The method for Internet of Things device identity authentication based on online learning according to claim 1, characterized in that: The feature extraction includes: feature extraction based on variational mode decomposition, feature extraction based on vertical synchronous compression transformation, feature extraction based on bispectral Radon transformation, feature extraction based on intrinsic time scale decomposition, feature extraction based on pulse unintentional modulation and feature extraction based on singular spectrum analysis.

7. The method for Internet of Things device identity authentication based on online learning according to claim 1, characterized in that: The online learning authentication model is an online learning authentication model based on Markov sampling.

8. The method for Internet of Things device identity authentication based on online learning according to claim 1, characterized in that: The semi-supervised dimensionality reduction algorithm has two corresponding cost functions.

Images