A dynamic authentication method based on laser relative intensity noise fingerprint

Abstract

The present application relates to the field of optical communication physical layer security, and discloses a dynamic authentication method based on laser relative intensity noise fingerprint, which fully excavates the unique features of the full working area of laser relative intensity noise, constructs a dynamic authentication system of one-time password based on a true random bias current challenge sequence, the challenge sequence of each authentication is not repeated and unpredictable, and can resist replay attacks; multi-dimensional feature fusion of time domain autocorrelation + frequency domain power spectrum + high-order statistics is adopted, and high authentication accuracy is still maintained in a wide temperature and wide voltage fluctuation range; the traditional digital encryption system based on mathematical complexity is not relied on, the relative intensity noise in the photoelectric conversion process of the semiconductor laser is used as an entropy source to construct a physical layer security authentication system, there is no mathematical link that can be cracked by quantum computing, and quantum attack resistance is realized from the physical bottom layer; the standard semiconductor laser can be directly adapted, and any structural modification of the laser is not needed, thereby reducing the additional cost of hardware.

Description

Technical Field

[0001] This invention relates to the field of optical communication physical layer security technology, and in particular to a dynamic authentication method based on laser relative intensity noise fingerprint. Background Technology

[0002] Current authentication systems for optoelectronic terminal devices are primarily based on symmetric / asymmetric digital encryption methods such as RSA, ECC, and AES, whose security relies entirely on mathematical computational complexity. With the rapid development of quantum computing technology, these methods face the fundamental threat of being quickly cracked by Shor's method, and the keys are constantly at risk of being stolen, tampered with, or reverse engineered throughout their entire lifecycle of distribution, storage, and transmission.

[0003] Physically Unclonable Function (PUF) technology utilizes the nanoscale random differences naturally formed during hardware manufacturing to generate identity fingerprints deeply bound to the hardware. Among them, optical PUF has become a research hotspot due to its advantages such as high entropy and resistance to electromagnetic interference. However, existing technologies generally rely on the design of special micro-nano structures or scattering media, which requires special modification of the hardware. This results in technical problems such as complex fabrication processes, high hardware costs, susceptibility of static features to replay attacks, and insufficient environmental robustness. Summary of the Invention

[0004] The purpose of this invention is to address the technical problems of existing digital encryption authentication systems, which rely on mathematical computational complexity and thus face the risk of quantum cracking and key leakage, as well as existing physical layer authentication schemes, which rely on special structures, resulting in high hardware costs, susceptibility to replay attacks on static features, and insufficient environmental robustness. This invention provides a dynamic authentication method based on laser relative intensity noise fingerprints.

[0005] To achieve the above-mentioned objectives, the embodiments of the present invention provide the following technical solutions:

[0006] A dynamic authentication method based on laser relative intensity noise fingerprint, comprising a registration calibration stage and a dynamic authentication stage;

[0007] The registration and calibration phase includes the following sub-steps:

[0008] Obtain a valid bias current sampling point for the semiconductor laser to be tested. Use the bias current sampling point to collect the time-domain relative intensity noise signal of the semiconductor laser. Preprocess the time-domain relative intensity noise signal to obtain a clean time-domain relative intensity noise signal.

[0009] The time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset are obtained by using a clean time-domain relative intensity noise signal. The registration feature vector is obtained by integrating the time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset.

[0010] Sequence transformation is performed on each feature component of the registration feature vector to obtain the binary bit sequence of the registration calibration stage. The binary bit sequence of the registration calibration stage and the common auxiliary data are processed by the security sketch method to obtain the key bit sequence. The key bit sequence is hashed and compressed to obtain the registration key bound to the bias current.

[0011] The dynamic authentication phase includes the following sub-steps:

[0012] After the semiconductor laser under test initiates an authentication request, a dynamic feature vector and a dynamic key are obtained through a randomly generated bias current challenge sequence. The dynamic feature vector and the corresponding registered feature vector are verified, and the dynamic key and the corresponding key are verified. When all two-dimensional verifications pass, the semiconductor laser under test is a legitimate device and the authentication is successful; otherwise, the semiconductor laser under test is an illegitimate device and the authentication fails.

[0013] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprinting, comprising the steps of acquiring the bias current sampling points of a legitimate semiconductor laser to be tested, collecting the time-domain relative intensity noise signal of the semiconductor laser using the bias current sampling points, and preprocessing the time-domain relative intensity noise signal to obtain a clean time-domain relative intensity noise signal, includes the following sub-steps:

[0014] The threshold current and maximum rated operating current of a valid and testable semiconductor laser are calibrated. The dynamic modulation range of the bias current is obtained using the threshold current and maximum rated operating current. Multiple bias current sampling points are selected at equal intervals within the dynamic modulation range and integrated into a calibration sampling sequence.

[0015] The semiconductor laser is driven to operate stably by each bias current sampling point, and the time-domain relative intensity noise signal of the semiconductor laser is collected.

[0016] The time-domain relative intensity noise signal is preprocessed to obtain a clean time-domain relative intensity noise signal; the preprocessing includes removing DC components, filtering low-frequency drift and high-frequency noise, outlier removal, and sliding smoothing denoising.

[0017] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprinting, wherein obtaining a time-domain autocorrelation feature subset, a frequency-domain power spectrum feature subset, and a higher-order statistical feature subset from a clean time-domain relative intensity noise signal, and integrating the time-domain autocorrelation feature subset, the frequency-domain power spectrum feature subset, and the higher-order statistical feature subset to obtain a registration feature vector includes the following sub-steps:

[0018] The autocorrelation function is obtained from the time-domain relative intensity noise signal, and the time-domain autocorrelation feature subset is extracted using the autocorrelation function.

[0019] The relative intensity noise power spectral density is obtained from the time-domain relative intensity noise signal, and the frequency-domain power spectrum feature subset is extracted using the relative intensity noise power spectral density.

[0020] The third-order and fourth-order cumulants are obtained from the relative intensity noise signal in the time domain, and the higher-order statistical feature subsets are extracted using the third-order and fourth-order cumulants.

[0021] The time-domain autocorrelation feature subset, the frequency-domain power spectrum feature subset, and the higher-order statistics feature subset are integrated to obtain the registration feature vector.

[0022] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprinting, wherein obtaining the autocorrelation function from the time-domain relative intensity noise signal and extracting a subset of time-domain autocorrelation features using the autocorrelation function includes the following sub-steps:

[0023] The autocorrelation function is obtained based on the relative intensity noise signal in the time domain at the current moment and the relative intensity noise signal in the time domain under the time delay.

[0024] A subset of time-domain autocorrelation features is extracted from the numerical curve of the autocorrelation function; the subset of time-domain autocorrelation features includes the peak value of the autocorrelation function, the delay time corresponding to the peak value, the full width at half maximum (FWHM), the attenuation coefficient of the main peak, and the peak-to-peak ratio of the sidelobes.

[0025] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprinting, wherein obtaining the relative intensity noise power spectral density from the time-domain relative intensity noise signal and extracting a frequency-domain power spectral feature subset using the relative intensity noise power spectral density includes the following sub-steps:

[0026] Power spectrum estimation of time-domain relative intensity noise signal is performed based on Welch method to obtain relative intensity noise power spectral density;

[0027] A subset of frequency domain power spectrum features is extracted using the relative intensity noise power spectral density; the subset of frequency domain power spectrum features includes the power spectrum integral value of the characteristic frequency band, spectral slope, peak frequency, valley frequency, in-band power fluctuation variance, and harmonic distortion.

[0028] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprinting, wherein each feature component of the registration feature vector is sequence-transformed to obtain a binary bit sequence for the registration calibration stage, the binary bit sequence and common auxiliary data for the registration calibration stage are processed by a security sketch method to obtain a key bit sequence, and the key bit sequence is hash-compressed to obtain a registration key bound to the bias current, includes the following sub-steps:

[0029] Adaptive interval quantization is performed on each feature component of the registration feature vector to obtain the binary bit sequence of the registration calibration stage;

[0030] The binary bit sequence of the registration calibration stage is processed by the security sketch generation method to obtain common auxiliary data. The binary bit sequence of the registration calibration stage is then corrected using the common auxiliary data and the security sketch recovery method to obtain the key bit sequence.

[0031] The key bit sequence is hashed and compressed using the SHA-3 method to obtain the registration key bound to the bias current.

[0032] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprinting involves initiating an authentication request for the semiconductor laser under test, obtaining a dynamic feature vector and a dynamic key through a randomly generated bias current challenge sequence, verifying the dynamic feature vector against its corresponding registration feature vector, and verifying the dynamic key against its corresponding key. If all two-dimensional verifications pass, the semiconductor laser under test is considered a legitimate device and authentication is successful; otherwise, it is considered an illegitimate device and authentication fails. This includes the following sub-steps:

[0033] The semiconductor laser under test initiates an authentication request. The authentication end randomly generates multiple non-repeating bias current challenge points from the dynamic modulation range using a true random number generator, and integrates them into a bias current challenge sequence.

[0034] By driving the semiconductor laser to operate stably at each bias current challenge point, the time-domain relative intensity noise signal of the semiconductor laser is collected, and the time-domain relative intensity noise signal is preprocessed to obtain a clean time-domain relative intensity noise signal.

[0035] The time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset are obtained by using a clean time-domain relative intensity noise signal. The time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset are integrated to obtain a dynamic feature vector.

[0036] Sequence transformation is performed on each feature component of the dynamic feature vector to obtain the binary bit sequence of the dynamic authentication stage. The binary bit sequence of the dynamic authentication stage and the common auxiliary data are processed by the security sketch method to obtain the key bit sequence. The key bit sequence is hashed and compressed to obtain the dynamic key bound to the bias current.

[0037] The system calls the registered feature vector and registered key corresponding to each bias current challenge point and performs a two-dimensional comparison and verification of feature similarity and key consistency. If the two-dimensional comparison and verification pass, the authentication is successful; otherwise, the authentication fails.

[0038] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprinting involves calling the registered feature vector and registered key corresponding to each bias current challenge point, and performing a two-dimensional comparison and verification of feature similarity and key consistency. If the two-dimensional comparison and verification pass, the authentication is successful; otherwise, the authentication fails. This includes the following sub-steps:

[0039] The similarity of feature vectors is obtained by processing dynamic feature vectors and registered feature vectors using cosine similarity.

[0040] The bit error rate is obtained by processing the dynamic key and the registration key using the Hamming distance method;

[0041] If the similarity of all feature vectors exceeds the similarity threshold and the bit error rate is less than the bit error rate threshold, then the semiconductor laser under test is a legitimate device and the authentication is successful; otherwise, the semiconductor laser under test is an illegitimate device, the authentication fails, and a security alarm mechanism is triggered.

[0042] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprints also includes encrypting and binding the laser's unique ID, bias current, registration feature vector, and registration key to a relative intensity noise standard encrypted fingerprint database, wherein the relative intensity noise standard encrypted fingerprint database updates data through a self-learning update mechanism.

[0043] Furthermore, a dynamic authentication method based on laser relative intensity noise fingerprints includes a self-learning update mechanism comprising the following sub-steps:

[0044] Once authentication is successful, the dynamic feature vector and dynamic key extracted during authentication will be smoothly updated by using a sliding window weighted average with the corresponding registration feature vector and registration key.

[0045] Set a periodic recalibration cycle. At each periodic recalibration cycle, the registration phase steps of the semiconductor laser are repeated to update the relative intensity noise standard encrypted fingerprint database.

[0046] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention establishes for the first time a three-dimensional nonlinear mapping model of relative intensity noise power spectral density with frequency and bias current, breaking through the limitations of static extraction with fixed bias current, fully exploring the unique characteristics of the relative intensity noise of the laser across the entire working range, and significantly improving the entropy value compared with existing static schemes; it constructs a one-time encryption dynamic authentication system based on a true random bias current challenge sequence, with each authentication challenge sequence being unique and unpredictable, completely resisting replay attacks, and attackers cannot replicate the full-range nonlinear response characteristics through finite sampling; it adopts multi-dimensional feature fusion of time-domain autocorrelation + frequency-domain power spectrum + higher-order statistics, maintaining high authentication accuracy over a wide temperature and voltage fluctuation range; it does not rely on traditional digital encryption systems based on mathematical complexity, but directly uses the inherent relative intensity noise in the photoelectric conversion process of semiconductor lasers as the entropy source to construct a physical layer security authentication system, without any mathematical links that can be cracked by quantum computing, achieving resistance to quantum attacks from the physical layer; this invention can be directly adapted to standard semiconductor lasers without any structural modification or special fabrication of the laser, resulting in zero additional hardware cost. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a dynamic authentication method based on laser relative intensity noise fingerprint. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0050] It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, the terms "first," "second," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance, or suggesting any such actual relationship or order between these entities or operations. Additionally, the terms "connected," "linked," etc., can refer to a direct connection between elements or an indirect connection via other elements.

[0051] The theoretical principle of this invention lies in the fact that, based on the amplitude of random fluctuations in optical power, the average output optical power, and the frequency resolution bandwidth, the power spectral density of relative intensity noise (RIN) is theoretically defined, as follows:

[0052] ;

[0053] in, For frequency The bias current is The power spectral density of relative intensity noise at that time. For frequency The bias current is The amplitude of random fluctuations in optical power at that time. The bias current is Average output optical power at that time This refers to the frequency resolution bandwidth.

[0054] It is important to note that by theoretically defining the power spectral density of relative intensity noise, a three-dimensional nonlinear mapping relationship was established between the power spectral density, frequency, and bias current of relative intensity noise. The physical principle is as follows: During the manufacturing processes of semiconductor lasers, including active region material growth, waveguide structure fabrication, cavity surface coating, and electrode bonding, uncontrollable and unreplicable nanoscale random errors naturally form. These nanoscale random errors directly lead to individual differences in the active region band structure, carrier transport characteristics, and photon scattering mechanisms of different semiconductor lasers, thus affecting the power spectral density of relative intensity noise in different semiconductor lasers. With bias current and frequency The changes exhibit unique and unclonable nonlinear response characteristics, which are deeply bound to the hardware of semiconductor lasers and cannot be cloned through reverse engineering. This makes it a natural high-entropy physical layer fingerprint entropy source.

[0055] like Figure 1 As shown, a dynamic authentication method based on laser relative intensity noise fingerprints includes a registration and calibration phase and a dynamic authentication phase.

[0056] The registration and calibration phase includes:

[0057] S1: Obtain the bias current sampling point of the valid semiconductor laser to be tested, use the bias current sampling point to collect the time-domain relative intensity noise signal of the semiconductor laser, preprocess the time-domain relative intensity noise signal to obtain a clean time-domain relative intensity noise signal.

[0058] S11: Calibrate the threshold current and maximum rated operating current of the valid semiconductor laser to be tested. Use the threshold current and maximum rated operating current to obtain the dynamic modulation range of the bias current. Select multiple bias current sampling points at equal intervals within the dynamic modulation range and integrate them into a calibration sampling sequence. The formula is:

[0059] ;

[0060] ;

[0061] in, For the i-th bias current sampling point, This is the index number of the bias current sampling point. For the threshold current, This represents the total number of bias current sampling points. For the maximum rated operating current, To calibrate the sampling sequence, This refers to the dynamic modulation range.

[0062] It is important to note that the dynamic modulation range ensures coverage of both the linear and nonlinear operating regions of the semiconductor laser.

[0063] S12: Drive the semiconductor laser to work stably through each bias current sampling point and collect the time-domain relative intensity noise signal of the semiconductor laser.

[0064] In this embodiment, the stabilization time of the semiconductor laser is greater than its thermal relaxation time before the time-domain relative intensity noise signal of the semiconductor laser is acquired. This avoids transient effects affecting the measurement accuracy of the time-domain relative intensity noise signal. The time-domain relative intensity noise signal of the semiconductor laser is acquired using a photodetector and a high-precision ADC (≥16bit), with a sampling rate of... ≥10 times the upper limit of target analysis frequency, single sampling length This ensures the resolution and accuracy of power spectrum estimation.

[0065] S13: Preprocess the time-domain relative intensity noise signal to obtain a clean time-domain relative intensity noise signal; the preprocessing includes removing the DC component, filtering out low-frequency drift and high-frequency noise, outlier removal, and sliding smoothing noise reduction.

[0066] It is important to note that preprocessing eliminates the influence of circuit noise and environmental interference on the relative intensity noise signal in the time domain.

[0067] S2: Obtain the time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset through a clean time-domain relative intensity noise signal. Integrate the time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset to obtain the registration feature vector.

[0068] S21: Obtain the autocorrelation function through the time-domain relative intensity noise signal, and use the autocorrelation function to extract the time-domain autocorrelation feature subset.

[0069] S211: Based on the time-domain relative intensity noise signal at the current moment and the time-domain relative intensity noise signal under the time delay, obtain the autocorrelation function, the formula of which is:

[0070] ;

[0071] in, The time delay is The bias current is The autocorrelation function over time, where t is a continuous-time variable. The bias current at time t is Time-domain relative intensity noise signal, For a moment Bias current is Time-domain relative intensity noise signal, This is a mathematical expectation operation.

[0072] S212: Extract a subset of time-domain autocorrelation features from the numerical curve of the autocorrelation function; the subset of time-domain autocorrelation features includes the peak value of the autocorrelation function, the delay time corresponding to the peak value, the full width at half maximum (FWHM), the attenuation coefficient of the main peak, and the peak-to-peak ratio of the sidelobes.

[0073] It is important to note that, based on the Wiener-Khinchin theorem, the autocorrelation function (time domain) and the power spectral density of the relative intensity noise (frequency domain) are Fourier transform pairs, as shown in the formula:

[0074] ;

[0075] ;

[0076] in, It is the imaginary unit.

[0077] In the embodiments, .

[0078] Therefore, the features extracted from the time domain and frequency domain are not redundant and can be used together.

[0079] S22: Obtain the relative intensity noise power spectral density from the time-domain relative intensity noise signal, and use the relative intensity noise power spectral density to extract the frequency domain power spectrum feature subset.

[0080] S221: Power spectrum estimation of time-domain relative intensity noise signals is performed based on the Welch method to obtain the relative intensity noise power spectral density, as shown in the formula:

[0081] ;

[0082] in, Number of signal segments For signal segmentation indexing, For discrete-time indexing, , The sampling interval is... For signal segment length, Let n be a window function with discrete time complexity n. For the first The discrete time in the signal segment is n, and the bias current is... The relative intensity noise signal in the time domain.

[0083] S222: Extract a feature subset of the frequency domain power spectrum by the relative intensity noise power spectral density; the feature subset of the frequency domain power spectrum includes the power spectrum integral value of the characteristic frequency band, the spectral slope, the peak frequency, the valley frequency, the variance of the in-band power fluctuation and the harmonic distortion ratio.

[0084] S23: Obtain third-order and fourth-order cumulants from the relative intensity noise signal in the time domain, and use the third-order and fourth-order cumulants to extract a subset of high-order statistical features.

[0085] S231: Obtain the third-order and fourth-order accumulators by the expectation of the relative intensity noise signal in the time domain, using the following formulas:

[0086] ;

[0087] ;

[0088] in, This is the third-order accumulation of the relative intensity noise signal in the time domain. It is the fourth-order accumulation of the relative intensity noise signal in the time domain.

[0089] It should be noted that the third-order and fourth-order accumulators are introduced to suppress measurement interference caused by Gaussian white noise, ambient temperature drift, and driving voltage fluctuations.

[0090] S232: Extract a subset of higher-order statistical features using third-order and fourth-order cumulants; the subset of higher-order statistical features includes the mean of third-order cumulants, the variance of third-order cumulants, the mean of fourth-order cumulants, the variance of fourth-order cumulants, skewness features, and kurtosis features.

[0091] S24: Integrate the time-domain autocorrelation feature subset, the frequency-domain power spectrum feature subset, and the higher-order statistics feature subset to obtain the registration feature vector, as shown in the formula:

[0092] ;

[0093] in, For bias current The registered feature vector at that time, This is a subset of time-domain autocorrelation features. A subset of the power spectrum in the frequency domain. A subset of features of higher-order statistics. For matrix transpose, For bias current When registering the j-th feature component of the feature vector (e.g., the power spectrum integral value of the feature band), j is the index of the feature component. This represents the total dimension of the features.

[0094] In the embodiments, .

[0095] S3: Perform sequence transformation on each feature component of the registration feature vector to obtain the binary bit sequence of the registration calibration stage. Process the binary bit sequence of the registration calibration stage and the common auxiliary data through the security sketch method to obtain the key bit sequence. Perform hash compression on the key bit sequence to obtain the registration key bound to the bias current.

[0096] S31: Perform adaptive interval quantization on each feature component of the registered feature vector to obtain the binary bit sequence of the registration calibration stage.

[0097] S311: Quantize each feature component of the registered feature vector using an adaptive quantization step size to obtain the interval number of each quantized feature component, as shown in the formula:

[0098] ;

[0099] ;

[0100] in, For characteristic components Quantized interval numbering, Let j be the minimum value obtained statistically during the registration and calibration phase for the j-th feature component. The adaptive quantization step size for the j-th feature component. Let j be the maximum value obtained statistically during registration calibration. The preset quantization bit depth, To quantify the total number of intervals, This is for floor function.

[0101] S312: Convert the interval number after quantization of each feature component into a b-bit binary number, and concatenate all b-bit binary numbers in sequence to obtain the binary bit sequence for the registration calibration stage. The formula is:

[0102] ;

[0103] in, For bias current The binary bit sequence during the registration and calibration phase. This is a conversion operation for b-bit binary numbers. This refers to the concatenation operation of b-bit binary numbers.

[0104] It is important to note that converting to a binary bit sequence eliminates inconsistencies caused by minute fluctuations in the characteristic components during the operation of the semiconductor laser.

[0105] S32: Process the binary bit sequence of the registration calibration stage using the security sketch generation method to obtain common auxiliary data. Use the common auxiliary data and the security sketch recovery method to correct the errors in the binary bit sequence of the registration calibration stage, and obtain the key bit sequence, as shown in the formula:

[0106] ;

[0107] ;

[0108] in, For bias current Public auxiliary data at that time The method for generating safety sketches, For bias current The key bit sequence at that time, Methods for restoring safe sketches.

[0109] It is important to note that the security sketch method is used to perform error correction on the binary bit sequence during the registration and calibration stage to ensure that the key bit sequence generated by the same semiconductor laser under the same bias current modulation has a consistency of ≥99.99%.

[0110] S33: The key bit sequence is hashed and compressed using the SHA-3 method to obtain the registration key bound to the bias current, as shown in the formula:

[0111] ;

[0112] in, For bias current The registration key at that time This is a SHA-3 hash operation.

[0113] In this embodiment, the SHA-3 method selected is SHA3-256, with an output length of 256 bits.

[0114] It is important to note that the laser's unique ID and bias current should be specified. Registration feature vector Registration key Encrypted binding is performed and stored in a relative intensity noise standard encrypted fingerprint database.

[0115] The dynamic authentication phase includes:

[0116] S4: After initiating an authentication request for the semiconductor laser under test, obtain the dynamic feature vector and dynamic key through a randomly generated bias current challenge sequence. Verify the dynamic feature vector and the corresponding registered feature vector, and verify the dynamic key and the corresponding key. When all two-dimensional verifications pass, the semiconductor laser under test is a legitimate device and authentication is successful; otherwise, the semiconductor laser under test is an illegitimate device and authentication fails.

[0117] S41: The semiconductor laser under test initiates an authentication request. The authentication end randomly generates multiple non-repeating bias current challenge points from the dynamic modulation range using a true random number generator, and integrates them into a bias current challenge sequence.

[0118] S42: Drive the semiconductor laser to operate stably through each bias current challenge point, collect the time-domain relative intensity noise signal of the semiconductor laser, preprocess the time-domain relative intensity noise signal to obtain a clean time-domain relative intensity noise signal.

[0119] S43: Obtain the time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset through a clean time-domain relative intensity noise signal, and integrate the time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset to obtain a dynamic feature vector.

[0120] S44: Perform sequence transformation on each feature component of the dynamic feature vector to obtain the binary bit sequence of the dynamic authentication stage. Process the binary bit sequence of the dynamic authentication stage and the common auxiliary data through the security sketch method to obtain the key bit sequence. Perform hash compression on the key bit sequence to obtain the dynamic key bound to the bias current.

[0121] It should be noted that because the bias current of the dynamic key is randomly generated, the unpredictability of the dynamic key is guaranteed.

[0122] It should be noted that the subsequent dynamic authentication phase follows the same process for obtaining the time-domain relative intensity noise signal, feature vector, and key, so it will not be repeated here.

[0123] S45: Call the registered feature vector and registered key corresponding to each bias current challenge point, and perform a two-dimensional comparison and verification of feature similarity and key consistency. If the two-dimensional comparison and verification pass, the authentication is successful; otherwise, the authentication fails.

[0124] It should be noted that the fingerprint database is accessed through a standard of relative intensity noise encryption.

[0125] S451: The similarity between dynamic and registered feature vectors is obtained by processing the cosine similarity. The formula is as follows:

[0126] ;

[0127] in, For feature vector similarity, For bias current The dynamic feature vector at time, where c is the index of the bias current challenge point.

[0128] S452: The bit error rate is obtained by processing the dynamic key and the registration key using the Hamming distance method, with the following formula:

[0129] ;

[0130] in, Bit error rate, For bias current Dynamic key at time, This represents the key length.

[0131] S453: If the similarity of all feature vectors exceeds the similarity threshold And the bit error rate is less than the bit error rate threshold. If the test is successful, the semiconductor laser under test is a legitimate device and the authentication is passed; otherwise, the semiconductor laser under test is an illegitimate device, the authentication fails, and a security alarm mechanism is triggered.

[0132] In the embodiments, , .

[0133] It is important to note that the two-dimensional comparison and verification process balances accuracy and security.

[0134] It also includes a self-learning and updating mechanism for the encrypted fingerprint database of relative intensity noise standards;

[0135] The steps of the self-learning and updating mechanism for the relative intensity noise standard encrypted fingerprint database are as follows:

[0136] S51: After successful authentication, the dynamic feature vector and dynamic key extracted during authentication are smoothly updated by applying a sliding window weighted average with the corresponding registration feature vector and registration key. The formula is as follows:

[0137] ;

[0138] ;

[0139] in, For the updated bias current The registered feature vector at that time, These are weighting coefficients. For the updated bias current The registration key at that time.

[0140] In the embodiments, It can adaptively adjust according to the aging characteristics of semiconductor lasers, ensuring smooth updates of the relative intensity noise standard encrypted fingerprint database, and taking into account both long-term robustness and environmental adaptability.

[0141] S52: Set a periodic recalibration cycle. At each periodic recalibration cycle, the registration stage steps of the semiconductor laser are re-performed, and the relative intensity noise standard encrypted fingerprint database is updated to avoid the influence of feature drift caused by long-term aging of semiconductor laser devices.

[0142] The beneficial effects of this invention are as follows:

[0143] 1. This invention utilizes the inherent uniqueness and non-cloning property of relative intensity noise in semiconductor lasers to achieve physical layer authentication, completely eliminating the reliance on mathematical complexity in traditional digital encryption. It resists the fundamental cracking risks brought by quantum computing and solves the technical problem of traditional digital keys being easily stolen and tampered with. Randomly generated bias current challenge points can resist replay attacks, and the success rate of cloning attacks by attackers is lower than [a certain percentage]. Its security level far exceeds that of existing encryption schemes.

[0144] 2. This invention employs multi-dimensional fusion feature extraction, resulting in a lower false recognition rate (FAR) compared to existing technologies. The false negative rate (FRR) is lower than With an accuracy rate of ≥99.99%, the introduction of high-order statistical feature subsets effectively suppresses the influence of environmental interference and circuit noise. It can still maintain stable certification performance within an industrial-grade wide temperature range of -40℃ to 85℃ and a driving voltage fluctuation range of ±10%, thus solving the technical problem of poor environmental adaptability of existing technologies.

[0145] 3. This invention can be directly implemented using existing semiconductor lasers without additional hardware costs. Compared with existing optical PUF solutions based on micro-nano structures, the hardware cost is reduced by more than 90%. The process of this invention is simple and can be dynamically implemented in general-purpose DSPs, MCUs, and FPGAs. The single authentication time is ≤10ms. It can be adapted to the high-speed authentication needs of various scenarios such as fiber optic communication equipment, laser sensing terminals, industrial IoT nodes, vehicle-mounted LiDAR, and optical modules, and has strong versatility and applicability.

[0146] 4. The relative intensity noise standard encrypted fingerprint database has an autonomous learning and update mechanism that can adaptively compensate for the slow changes in characteristics caused by long-term operation aging and temperature drift of semiconductor lasers. The certification stability is maintained at over 99.9% throughout the entire life cycle of the equipment, which solves the problem of performance degradation after long-term use of existing technical solutions and significantly reduces equipment operation and maintenance costs.

Claims

1. A dynamic authentication method based on laser relative intensity noise fingerprint, characterized in that, This includes the registration and calibration phase and the dynamic certification phase; The registration and calibration phase includes the following sub-steps: Obtain a valid bias current sampling point for the semiconductor laser to be tested. Use the bias current sampling point to collect the time-domain relative intensity noise signal of the semiconductor laser. Preprocess the time-domain relative intensity noise signal to obtain a clean time-domain relative intensity noise signal. The time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset are obtained by using a clean time-domain relative intensity noise signal. The registration feature vector is obtained by integrating the time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset. Sequence transformation is performed on each feature component of the registration feature vector to obtain the binary bit sequence of the registration calibration stage. The binary bit sequence of the registration calibration stage and the common auxiliary data are processed by the security sketch method to obtain the key bit sequence. The key bit sequence is hashed and compressed to obtain the registration key bound to the bias current. The dynamic authentication phase includes the following sub-steps: After the semiconductor laser under test initiates an authentication request, a dynamic feature vector and a dynamic key are obtained through a randomly generated bias current challenge sequence. The dynamic feature vector and the corresponding registered feature vector are verified, and the dynamic key and the corresponding key are verified. When all two-dimensional verifications pass, the semiconductor laser under test is a legitimate device and the authentication is successful; otherwise, the semiconductor laser under test is an illegitimate device and the authentication fails.

2. The dynamic authentication method based on laser relative intensity noise fingerprint according to claim 1, characterized in that, The process of obtaining a valid bias current sampling point for the semiconductor laser under test, acquiring the time-domain relative intensity noise signal of the semiconductor laser using the bias current sampling point, and preprocessing the time-domain relative intensity noise signal to obtain a clean time-domain relative intensity noise signal includes the following sub-steps: The threshold current and maximum rated operating current of a valid and testable semiconductor laser are calibrated. The dynamic modulation range of the bias current is obtained using the threshold current and maximum rated operating current. Multiple bias current sampling points are selected at equal intervals within the dynamic modulation range and integrated into a calibration sampling sequence. The semiconductor laser is driven to operate stably by each bias current sampling point, and the time-domain relative intensity noise signal of the semiconductor laser is collected. The time-domain relative intensity noise signal is preprocessed to obtain a clean time-domain relative intensity noise signal; the preprocessing includes removing DC components, filtering low-frequency drift and high-frequency noise, outlier removal, and sliding smoothing denoising.

3. The dynamic authentication method based on laser relative intensity noise fingerprint according to claim 1, characterized in that, The process of obtaining time-domain autocorrelation feature subsets, frequency-domain power spectrum feature subsets, and higher-order statistical feature subsets from a clean time-domain relative intensity noise signal, and integrating these subsets to obtain the registration feature vector, includes the following sub-steps: The autocorrelation function is obtained from the time-domain relative intensity noise signal, and the time-domain autocorrelation feature subset is extracted using the autocorrelation function. The relative intensity noise power spectral density is obtained from the time-domain relative intensity noise signal, and the frequency-domain power spectrum feature subset is extracted using the relative intensity noise power spectral density. The third-order and fourth-order cumulants are obtained from the relative intensity noise signal in the time domain, and the higher-order statistical feature subsets are extracted using the third-order and fourth-order cumulants. The time-domain autocorrelation feature subset, the frequency-domain power spectrum feature subset, and the higher-order statistics feature subset are integrated to obtain the registration feature vector.

4. The dynamic authentication method based on laser relative intensity noise fingerprint according to claim 3, characterized in that, The step of obtaining the autocorrelation function from the time-domain relative intensity noise signal and extracting a subset of time-domain autocorrelation features using the autocorrelation function includes the following sub-steps: The autocorrelation function is obtained based on the relative intensity noise signal in the time domain at the current moment and the relative intensity noise signal in the time domain under the time delay. A subset of time-domain autocorrelation features is extracted from the numerical curve of the autocorrelation function; the subset of time-domain autocorrelation features includes the peak value of the autocorrelation function, the delay time corresponding to the peak value, the full width at half maximum (FWHM), the attenuation coefficient of the main peak, and the peak-to-peak ratio of the sidelobes.

5. The dynamic authentication method based on laser relative intensity noise fingerprint as described in claim 3, characterized in that, The step of obtaining the relative intensity noise power spectral density from the time-domain relative intensity noise signal and extracting a frequency-domain power spectral feature subset using the relative intensity noise power spectral density includes the following sub-steps: Power spectrum estimation of time-domain relative intensity noise signal is performed based on Welch method to obtain relative intensity noise power spectral density; A subset of frequency domain power spectrum features is extracted using the relative intensity noise power spectral density; the subset of frequency domain power spectrum features includes the power spectrum integral value of the characteristic frequency band, spectral slope, peak frequency, valley frequency, in-band power fluctuation variance, and harmonic distortion.

6. The dynamic authentication method based on laser relative intensity noise fingerprint according to claim 1, characterized in that, The process of performing sequence transformation on each feature component of the registration feature vector to obtain the binary bit sequence of the registration calibration stage, processing the binary bit sequence of the registration calibration stage and the common auxiliary data through the security sketch method to obtain the key bit sequence, and performing hash compression on the key bit sequence to obtain the registration key bound to the bias current includes the following sub-steps: Adaptive interval quantization is performed on each feature component of the registration feature vector to obtain the binary bit sequence of the registration calibration stage; The binary bit sequence of the registration calibration stage is processed by the security sketch generation method to obtain common auxiliary data. The binary bit sequence of the registration calibration stage is then corrected using the common auxiliary data and the security sketch recovery method to obtain the key bit sequence. The key bit sequence is hashed and compressed using the SHA-3 method to obtain the registration key bound to the bias current.

7. The dynamic authentication method based on laser relative intensity noise fingerprint according to claim 1, characterized in that, After initiating an authentication request for the semiconductor laser under test, a dynamic feature vector and a dynamic key are obtained through a randomly generated bias current challenge sequence. The dynamic feature vector and its corresponding registration feature vector are verified, and the dynamic key and its corresponding key are verified. When all two-dimensional verifications pass, the semiconductor laser under test is a legitimate device and authentication is successful; otherwise, the semiconductor laser under test is an illegitimate device and authentication fails. This includes the following sub-steps: The semiconductor laser under test initiates an authentication request. The authentication end randomly generates multiple non-repeating bias current challenge points from the dynamic modulation range using a true random number generator, and integrates them into a bias current challenge sequence. By driving the semiconductor laser to operate stably at each bias current challenge point, the time-domain relative intensity noise signal of the semiconductor laser is collected, and the time-domain relative intensity noise signal is preprocessed to obtain a clean time-domain relative intensity noise signal. The time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset are obtained by using a clean time-domain relative intensity noise signal. The time-domain autocorrelation feature subset, frequency-domain power spectrum feature subset, and higher-order statistical feature subset are integrated to obtain a dynamic feature vector. Sequence transformation is performed on each feature component of the dynamic feature vector to obtain the binary bit sequence of the dynamic authentication stage. The binary bit sequence of the dynamic authentication stage and the common auxiliary data are processed by the security sketch method to obtain the key bit sequence. The key bit sequence is hashed and compressed to obtain the dynamic key bound to the bias current. The system calls the registered feature vector and registered key corresponding to each bias current challenge point and performs a two-dimensional comparison and verification of feature similarity and key consistency. If the two-dimensional comparison and verification pass, the authentication is successful; otherwise, the authentication fails.

8. The dynamic authentication method based on laser relative intensity noise fingerprint according to claim 6, characterized in that, The call and the registration feature vector and registration key corresponding to each bias current challenge point are compared and verified in two dimensions: feature similarity and key consistency. If the two-dimensional comparison and verification pass, the authentication is successful; otherwise, the authentication fails, including the following sub-steps: The similarity of feature vectors is obtained by processing dynamic feature vectors and registered feature vectors using cosine similarity. The bit error rate is obtained by processing the dynamic key and the registration key using the Hamming distance method; If the similarity of all feature vectors exceeds the similarity threshold and the bit error rate is less than the bit error rate threshold, then the semiconductor laser under test is a legitimate device and the authentication is successful; otherwise, the semiconductor laser under test is an illegitimate device, the authentication fails, and a security alarm mechanism is triggered.

9. The dynamic authentication method based on laser relative intensity noise fingerprint according to claim 1, characterized in that, It also includes encrypting and binding the laser's unique ID, bias current, registration feature vector, and registration key and storing them in a relative intensity noise standard encrypted fingerprint database, which updates the data through a self-learning update mechanism.

10. The dynamic authentication method based on laser relative intensity noise fingerprint according to claim 9, characterized in that, The autonomous learning and updating mechanism includes the following sub-steps: Once authentication is successful, the dynamic feature vector and dynamic key extracted during authentication will be smoothly updated by using a sliding window weighted average with the corresponding registration feature vector and registration key. Set a periodic recalibration cycle. At each periodic recalibration cycle, the registration phase steps of the semiconductor laser are repeated to update the relative intensity noise standard encrypted fingerprint database.

Images