A two-level WBAN authentication method based on iris features and ECG features

Abstract

The present application belongs to the field of information security, and particularly relates to a two-stage WBAN authentication method based on iris features and ECG features, which comprises the following steps: collecting an iris image through an iris collector, and performing first-stage identification and authentication on the identity of a user through four stages of iris image preprocessing, iris image feature extraction and quantization, iris template making and iris template matching; after the first-stage authentication, collecting an electrocardiogram signal of the user using a sensor, and performing second-stage authentication on the identity of the user through four stages of ECG signal preprocessing, ECG feature extraction, ECG feature quantization and a fuzzy commitment-based key agreement mechanism. Only the user who passes the two-stage authentication is considered to have passed the authentication. The method adopts two-stage identity authentication, improves the accuracy of identity identification, and is beneficial to protecting the security of the WBAN and protecting the privacy of the user.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of information security, and particularly relates to a two-level WBAN authentication method based on iris features and ECG features. BACKGROUND

[0002] With the development of sensor and wireless network technology, wireless body area network (WBAN) as a typical application scenario of Internet of Things has attracted more and more attention from the academic and industrial circles. Mobile medical care is an emerging technology and one of the important application fields of wireless body area network, which allows patients to collect physiological information such as heart rate, blood pressure, etc. through various medical device sensors worn or implanted on the body. These medical devices often have wireless communication functions and can remotely send the collected user information to medical personnel or monitoring sites. In this way, doctors can analyze the physiological parameter indicators received by patients to judge the development of the patient's condition and timely propose treatment suggestions without the patient leaving home, which is a time-saving and effective strategy for patients, especially those who are not convenient to travel due to physical reasons. Under the background of repeated epidemics, it is also very important for ordinary people to minimize contact with others. This technology has been widely used in the monitoring of chronic diseases in practical scenarios.

[0003] Since the wireless body area network stores a lot of physiological information related to patient privacy, once it is obtained and analyzed by hackers, it will pose a serious threat to the privacy and safety of patients. Therefore, only authorized users can access these private data, and how to protect the security of these data has become the focus of industry discussion. The current IEEE802.15.6 standard describes the security requirements and security levels required in WBAN, and the standard also suggests several security schemes for protecting WBAN, but these schemes have been proven to have insufficient security in practical application scenarios. Therefore, there is an urgent need for a more powerful security scheme to protect the collection and transmission of WBAN data. The utility model with the authorization announcement number CN205721741U discloses a random number generation device suitable for RFID reader, the core idea of which is to improve the cycle jitter of the oscillator to increase randomness, which is suitable for RFID reader. The utility model with the authorization announcement number CN205620988U discloses a hardware random number generator that can generate a one-bit random number. The invention with the application publication number CN103870757A provides a digital password generation method based on rectangular coordinate system, which can generate a two-digit password. However, the security performance of these methods is insufficient.

[0004] Since the sensors in WBAN are used to collect physiological data of patients, the WBAN protection scheme based on physiological characteristics gradually begins to be welcomed by researchers. In fact, physiological characteristic signals such as fingerprints, ECG (electrocardiogram), PPG, etc. have been proved to be suitable for privacy protection work of WBAN due to their uniqueness, randomness, and have been widely studied. But most of the protection schemes have the problems of weak identification performance and long authentication time. SUMMARY

[0005] The present application aims at the deficiencies of the prior art, and provides a two-level authentication mechanism fusing iris and ECG signals for protecting the security of WBAN.

[0006] To achieve the above object, the technical scheme adopted by the present application is as follows.

[0007] A two-level WBAN authentication method based on iris features and ECG features, comprising the following steps:

[0008] Step S1, an iris image is collected by an iris collector, and the identity of a user is identified and authenticated in the first level through four stages of iris image preprocessing, iris image feature extraction and quantization, iris template making and iris template matching.

[0009] Step S11, the specific steps of the iris image preprocessing are as follows:

[0010] First, the collected iris image is positioned by using the integral differential operator proposed by Daugmann, which is defined as:

[0011]

[0012] In formula (1), (x, y) refers to the pixel point coordinates of the collected image containing eyes, G σ (r) is a Gaussian smoothing function, the symbol * represents convolution operation, and ds refers to the curve integral on the circle with r as the radius and (x0, y0) as the center. The operator behaves as a circular detector, and with the increase of the radius, the partial derivative is used to search for the local maximum value in the region.

[0013] Then, the straight line and curve parts in the iris image are found by using the Hough transform and the Canny edge detection method. The Canny edge detection uses a filter based on the Gaussian derivative to calculate the intensity of the gradient, which can reduce the influence of noise in the image. The boundary detection by Hough transform is a standardized machine recognition technology for fitting the simple contour model of the image.

[0014] Step S12, the specific steps of iris image feature extraction and quantization are as follows:

[0015] The iris image is extracted and quantized using a Log Gabor filter. The frequency response formula of the Log Gabor filter is as follows:

[0016]

[0017] In equation (2), f0 represents the center frequency of the filter, and σ represents the bandwidth of the filter.

[0018] The specific process of extraction and quantization is as follows: Each row of the standardized iris image in step S11 is treated as a one-dimensional vector, and convolved with a Log Gabor filter. The convolution result has real and imaginary parts. We record the positive and negative values ​​of the real and imaginary parts, using 1 to represent positive numbers and 0 to represent negative numbers. In this way, each pixel of the iris image is quantized into 2 bits. Finally, the calculation results are concatenated row by row to quantize the iris image into a binary iris code of length X bits.

[0019] Step S13, the specific steps for making the iris template are as follows:

[0020] First, select the clearest iris image with the naked eye and name its corresponding binary iris code as reference code D1. Then, record the indices where reference code D1 matches the binary iris code values ​​of other iris images from the same eye. Finally, take the intersection of these indices, which is called the pre-template D. f Pre-template D f The definition is as follows:

[0021] D f =D 12 ∩D 13 ∩…∩D 1k (3)

[0022] In equation (3), D ij It refers to D i and D j An index sequence with the same position.

[0023] Secondly, the length of the pre-template obtained varies for different eyes, therefore it is necessary to adjust D. f Perform interval mapping; in order to fully utilize D f All the information, we take the D obtained by each eye f The sequence is mapped in the interval [0, N), and the N indices closest to the integer (0, 1, 2, ..., N-1) are selected as the new pre-template D. fnew ; that is, D of different lengths f Convert to D of the same lengthfnew .

[0024] Then, the D fnew is further simplified. Specifically, the binary sequence of D fnew is divided into bit blocks each having a length of b (b is an odd number), and then the middle element of each bit block is XORed with all other elements in the same block in turn. After the XOR operation is completed, a new and shorter bit block is obtained. Finally, all the bit blocks are concatenated to obtain a new binary sequence, i.e., a new index sequence D irisindex , and in the reference encoding D1, the corresponding binary number is extracted according to the value of D irisindex as the index to obtain the final iris encoding template D iris .

[0025] Step S14, the specific steps of the iris template matching are as follows:

[0026] In the iris template matching stage, the Hamming distance is introduced to judge the similarity between the to-be-tested sample and the iris template. The Hamming distance can be simply described as the number of different characters at the same position of two equal-length sequences. The smaller the Hamming distance is, the more similar the two sequences are. We take the iris threshold TH iris , and when the Hamming distance of the to-be-tested sample is less than the threshold TH iris , the user passes the first-level iris authentication, otherwise, the authentication fails.

[0027] Step S2, after passing the first-level iris authentication, the ECG signal of the user is collected by using the sensor, and the identity of the user is authenticated in four stages of ECG signal preprocessing, ECG feature extraction, ECG feature quantization and fuzzy commitment-based key agreement mechanism. Only the user who passes the two-level authentication is considered to pass the authentication.

[0028] Step S21, the ECG signal preprocessing is to preprocess the obtained ECG signal by using the wavelet transform method to remove baseline drift, power frequency interference, electromyographic interference and other noises. The specific steps are as follows:

[0029] First, L 2 (R) represents the totality of energy-limited signals, and when f(t) ∈ L 2 (R) is satisfied, the continuous wavelet transform formula of the signal f(t) can be defined as:

[0030]

[0031] In formula (4), a is a scale factor, b is a translation factor, and φ (t) represents a mother wavelet function. ​complex conjugate of the mother wavelet function, but the continuous wavelet transform has the defects of large amount of calculation and long calculation time, and in practical application, especially in computer implementation, the continuous wavelet transform often needs to be discretized in binary. j j k (j, k∈Z), wherein j represents a decomposition scale, and Z represents all integer sets.

[0032] Secondly, the Sym4 wavelet is selected as the base function; the wavelet threshold function is a rule for modifying the wavelet coefficients, and different wavelet threshold functions represent different strategies for processing the wavelet coefficients, and here, a soft and hard threshold function is selected to filter the wavelet coefficients containing noise and remove the Gaussian noise coefficients, and the expression of the wavelet threshold function is as follows:

[0033]

[0034] In formula (5), ω j,k represents the wavelet coefficient after decomposition, represents the wavelet coefficient after processing by the threshold function, λ represents a threshold value, and the value of T is in the interval (0, 1).

[0035] The selection of the threshold value λ is also a key step for wavelet denoising, and in order to maintain the characteristics of the ECG signal, the improved fixed threshold selection method is adopted in the application.

[0036]

[0037] In formula (6), σ n represents a variance, N represents a signal length, and j represents a decomposition scale.

[0038] Step S22: performing feature extraction on the ECG signal processed in step S21, and IPI is widely used to protect the wireless body area network device due to its unique randomness.

[0039] IPI i = R i -R i-1 (7)

[0040] In formula (7), R i ​R peak of the ith heartbeat cycle. In this method, IPI is regarded as a feature of ECG signal, therefore, we are only interested in R peak. We find the R peak of the preprocessed ECG signal by using biopeaks peak detection method, and calculate the IPI sequence corresponding to R peak, denoted as {IPI i}.

[0041] Step S23, ECG feature quantization, after obtaining {IPI i}, we use a fusion method to quantize it, and generate a binary sequence. The fusion method is called IPI Fusion Coding, IPIFSC for short, in this invention. The specific process of quantization is as follows:

[0042] First, input the collected IPI sequence {IPI i}, select n consecutive IPIs in {IPI i}, denoted as {IPI in};

[0043] Secondly, calculate the average value IPI mean of {IPI in}; then calculate the difference between each IPI in {IPI in} and IPI mean , store the difference value in {array}, calculate the mathematical expectation μ and variance σ 2 of {array}, according to table 1, calculate 16 domains, and convert the elements in {IPI in} falling into the corresponding domain into the corresponding gray code; at this time, generate binary sequence BS1;

[0044] Compare the IPI trend of {IPI in}; if the value of the previous IPI is less than the value of the next IPI (IPI i <IPI i+1 ), then quantize the ith element as 1, otherwise, complement 0; for the last element in {IPI in}, complement a random 0 or 1 at the end; at this time, generate binary sequence BS2;

[0045] Quantize each element in {IPI in} into 8-bit binary, and only use 2-4 bits at this time, generate binary sequence BS3;

[0046] Finally, concatenate BS1, BS2 and BS3 in the same index to obtain the final binary sequence BS.

[0047] Table 1 16 equal parts of normal distribution and corresponding gray code

[0048]

[0049] In step S24, the key is distributed by a fuzzy commitment key agreement mechanism, and the BSs calculated by the ECG signals detected at different positions at the same time by the same person are slightly different due to measurement errors of the devices. The fuzzy commitment mechanism introduces error correction code, which can be used to correct the difference. The method uses the BS generated by the IPI to protect the key distribution in the fuzzy commitment framework. The specific steps are as follows:

[0050] The sender and the receiver detect the ECG signals of the subject at the same time, and generate binary sequences BS1 and BS2 respectively. For the sender, the key key needs to be encoded by error correction code to generate a sequence BS key , and then a commitment is created:

[0051]

[0052] In formula (8), hash(key) represents one-way hash operation on key, and the symbol represents bitwise XOR operation.

[0053] The commitment will be sent to the receiver, which first performs operation to generate BS key′ , and then performs error correction code decoding operation on BS key′ to generate key', if hash(key) = hash(key'), the authentication is successful, and the user passes the second-level authentication; otherwise, the authentication fails, and the two sensor devices (the sender and the receiver) for collecting the ECG signals will detect the ECG signals of the subject again.

[0054] The beneficial effects of the present application are:

[0055] The two-level WBAN authentication method of the present application can be used for identity authentication of users in a wireless body area network environment, and can be used in the fields of medical information or sports health. The fusion method adopted by the present application can shorten the time required for quantization without losing the randomness of the binary sequence, and has the effects of saving WBAN resource space and shortening the encoding time. The two-level identity authentication adopted by the present application improves the accuracy of identity recognition, is beneficial to protecting the security of the WBAN, and protects the privacy of the user. DETAILED DESCRIPTION

[0056] Figure 1 The flowchart of the two-level WBAN authentication method of the present application.

[0057] Figure 2 The ECG key distribution process based on fuzzy commitment.

[0058] Figure 3Hamming distance distribution of iris encoding templates generated for different subjects Figure 1 .

[0059] Figure 4 Hamming distance distribution of iris encoding templates generated for different subjects Figure 2 .

[0060] Figure 5 Hamming distance distribution of iris encoding templates generated for the same subject Figure 1 .

[0061] Figure 6 Hamming distance distribution of iris encoding templates generated for the same subject Figure 2 .

[0062] Figure 7 Figure of the test results of the individual ECG feature recognition performance in the authentication method of the present application.

[0063] Figure 8 Figure of the test results of the recognition performance of the two-level authentication method of the present application after combining the iris and ECG features together. DETAILED DESCRIPTION

[0064] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the present application will be further described in detail below in combination with the drawings and specific embodiments. In the following description, a large number of specific details are set forth in order to facilitate a full understanding of the present application, however, the present application can also be implemented in other manners different from those described herein, and therefore, the present application is not limited to the specific embodiments disclosed below.

[0065] As shown in the two-level WBAN authentication method based on iris features and ECG features in the present application, the method comprises the following steps: Figures 1-2

[0066] Step S1, through an iris collector, an iris image is collected, and through four stages of iris image preprocessing, iris image feature extraction and quantization, iris template making and iris template matching, the identity of a user is identified and authenticated in the first level.

[0067] The specific steps of the iris image preprocessing are as follows:

[0068] First, the collected iris image is positioned by using the integral differential operator proposed by Daugmann, which is defined as:

[0069]

[0070] In formula (1), (x, y) is the collected iris image containing eyes, G(x, y) is the gray value of the pixel at the position (x, y), and G(x, y) is the gray value of the pixel at the position (x, y) in the iris image after the integral differential operator is applied.​σ (r) is a Gaussian smoothing function, the symbol * represents a convolution operation, and ds denotes a curve integral over a circle with radius r and center (x0, y0). This operator behaves as a circular detector that searches for local maxima in the region using partial derivatives as the radius increases.

[0071] Then, the Hough transform and Canny edge detection method are used to find the straight line and curve part in the iris image, so as to normalize the iris region to remove the error such as scaling distortion of the same iris, and obtain the standardized iris image. In the Canny edge detection, a filter based on the Gaussian derivative is used to calculate the gradient intensity, which can reduce the influence of noise in the image. The boundary detection by the Hough transform is a standardized machine recognition technology for fitting a simple contour model of the image.

[0072] Step S12, the specific steps of the iris image feature extraction and quantization are:

[0073] The iris image is extracted and quantized by using the Log Gabor filter to extract the texture information such as crypts, spots and filaments of the iris. The Log Gabor filter inherits the advantages of the Gabor filter and has good local analysis ability; in addition, it will not produce DC component. The frequency response formula of the Log Gabor filter is as follows:

[0074]

[0075] In formula (2), f0 represents the center frequency of the filter, and σ represents the bandwidth of the filter.

[0076] The specific process of the iris image feature extraction and quantization is that each row of the iris image after normalization in step S11 is regarded as a one-dimensional vector, and is respectively convolved with the Log Gabor filter. The convolution result exists in the real part and the imaginary part, we record the positive and negative of the real part and the imaginary part, and respectively use 1 to represent the positive number and 0 to represent the negative number, so that each pixel point will be quantized into 2 bits. Finally, the calculation results are concatenated by rows, so that the iris picture can be quantized into a binary iris code with a length of X bits.

[0077] Step S13, the specific steps of the iris template making are:

[0078] For CASIA1 database, each iris image can be quantized to 9600 bits finally. For the wireless body area network sensor with very limited resource, the quantization length needs to be reduced. First, a clearest iris image is selected by naked eyes, and the corresponding binary iris code is called reference code D1. Then, the indexes of the reference code D1 and other iris codes of the same eye are recorded in turn. Finally, the intersection of the indexes is taken as the pre-template D f . The pre-template D f is defined as follows:

[0079] D f =D 12 ∩D 13 ∩…∩D 1k (3)

[0080] In formula (3), D ij means the index sequence of the same bit of D i and D j , i.e. D 12 means the index sequence of the same bit of D1 and D2, D 1k means the index sequence of the same bit of D1 and D k .

[0081] However, the length of the pre-template obtained is inconsistent for different eyes, so interval mapping is needed for D f ; in order to make full use of the information of D f , the D f sequence obtained for each eye is mapped in the interval [0, N), and N indexes closest to the integer (0, 1, 2, …, N-1) are selected as the new pre-template D fnew ; that is, the D f of different lengths is converted into the D fnew of the same length.

[0082] Then, D fnew is further simplified. The specific process is that the binary sequence of D fnew is divided into bit blocks each with a length of b (b is an odd number), then the middle element of each bit block is XORed with all other elements in the same block in turn, when the XOR operation is finished, a new and shorter bit block is obtained, finally all the bit blocks are concatenated to obtain a new binary sequence, i.e. a new index sequence D irisindex , and the corresponding binary number in the reference code D1 is extracted according to the value in D irisindex , to obtain the final iris code template D iris .

[0083] In the authentication system of the present application, Dirisindex With D iris Known as a set of iris templates, need to be stored in advance inside the sensor.

[0084] Step S14, the specific steps of the iris template matching are:

[0085] In the iris template matching stage, the Hamming distance is introduced to judge the similarity between the test sample and the iris template. The Hamming distance can be simply described as the number of different characters of two equal-length sequences at the same position. The smaller the Hamming distance, the more similar the two sequences. We take the iris threshold TH iris When the Hamming distance of the test sample is less than the threshold TH iris , the user passes the first level of iris authentication, otherwise, the authentication fails.

[0086] Step S2, after passing the first level of authentication, the sensor is used to collect the user's ECG signal, which goes through four stages of ECG signal preprocessing, ECG feature extraction, ECG feature quantization and fuzzy commitment-based key agreement mechanism to perform the second level of authentication on the user's identity. Only the user who passes the two levels of authentication is considered to have passed the authentication.

[0087] Step S21, due to the complex and variable environment in daily life, the ECG signal collected by the sensor will inevitably be disturbed, so it is necessary to preprocess the ECG signal to remove baseline drift, power frequency interference, electromyographic interference and other noises. Wavelet transform can localize the analysis of time domain and frequency domain by changing the scale factor and the translation factor. Many researchers have proved that wavelet transform has good performance in the field of ECG signal processing. This is the method used in this invention. The wavelet transform method is used to preprocess the obtained ECG signal to remove baseline drift, power frequency interference, electromyographic interference and other noises; the specific steps of the ECG signal preprocessing are:

[0088] First, L 2 (R) represents the totality of energy-limited signals, when f(t)∈L 2 (R), the continuous wavelet transform formula of the signal f(t) can be defined as:

[0089]

[0090] In formula (4), a is the scale factor, b is the translation factor, is the complex conjugate of . However, continuous wavelet transform has the disadvantages of large calculation amount and long calculation time, and in practical application, especially in computer implementation, it is often necessary to discretize the continuous wavelet transform in binary. Discretizing the scale factor a and the translation factor b of the continuous wavelet transform can obtain a discrete wavelet sequence. The commonly used discretization is a=2j b = 2 j k (j, k∈Z), wherein j represents a decomposition scale, and Z represents a set of all integers. The discrete wavelet transform can be represented as:

[0091] Secondly, since the Sym4 wavelet is similar to the ECG peak value and has good performance in ECG signal pretreatment, the Sym4 wavelet is selected as the base function. The wavelet threshold function is a rule for modifying wavelet coefficients, and different wavelet threshold functions represent different strategies for processing wavelet coefficients.

[0092]

[0093] In formula (5), represents the wavelet coefficient after threshold function processing, ω represents the wavelet coefficient after decomposition, and λ represents a threshold value. j,k The value of T is in the interval (0, 1).

[0094] The selection of the threshold value λ is also a key step for wavelet denoising.

[0095]

[0096] In formula (6), σ n represents a variance, N represents a signal length, and j represents different scales. With the increase of the decomposition scale j, the improved threshold value λ gradually decreases, and is consistent with the propagation characteristics of noise under different scales of wavelet transform, thereby ensuring the effectiveness of denoising.

[0097] Step S22, ECG feature extraction, IPI is widely used to protect wireless body area network devices due to its unique randomness.

[0098] IPI i = R i -R i-1 (7)

[0099] In formula (7), R iR-peak of the i-th heartbeat cycle. In this method, IPI is regarded as a feature of ECG signal, so we are only interested in R-peak. We find the R-peak of ECG signal by biopeaks peak detection method (i.e. biopeaks quote "Brammer, J.. (2020). Biopeaks: a graphical user interface for feature extraction from heart-and breathing biosignals. The Journal of Open Source Software, 5(54), 2621." search method "Lastly, the ECG R-peak detector is a custom algorithm"), and calculate the IPI sequence corresponding to R-peak, denoted as {IPI i}.

[0100] Step S23, ECG feature quantization, after obtaining {IPI i}, a fusion method is used to quantize it to generate a binary sequence. The fusion method is called IPI Fusion Coding, abbreviated as IPIFSC, and the specific process of quantization is as follows:

[0101] First, input the collected IPI sequence {IPI i}, select n consecutive IPIs in {IPI i}, denoted as {IPI in};

[0102] Secondly, calculate the average value IPI in of {IPI mean}; then calculate the difference between each IPI in {IPI in} and IPI mean , store each difference value in {array}, calculate the mathematical expectation μ and variance σ 2 of {array}, according to Table 1, calculate 16 domains, and convert the elements in {IPI in} falling into the corresponding domain into the corresponding Gray code; at this time, the binary sequence BS1 is generated.

[0103] Compare the IPI trend of {IPI in}; if the value of the previous IPI is less than the value of the next IPI (IPI i <IPI i+1 ), then the i-th element is quantized to 1, otherwise, it is complemented by 0; for the last element in {IPI in}, a random 0 or 1 is appended at the end; at this time, the binary sequence BS2 is generated.

[0104] Each element in the set {IPI in} is quantized into 8-bit binary, and only 2-4 bits are used, and a binary sequence BS3 is generated.

[0105] Finally, BS1, BS2, BS3 are concatenated in the same index manner to obtain the final binary sequence BS.

[0106] Table 1 Normal distribution of 16 equal parts and corresponding Gray codes

[0107]

[0108] Step S24, distribute the key through the fuzzy commitment key agreement mechanism:

[0109] Due to the measurement error of the device and other reasons, even the same person, the BS calculated by detecting the ECG signal at different parts at the same time is slightly different. The fuzzy commitment mechanism introduces error correction code, which can be used to correct this difference. In this method, the BS generated by IPI is used to protect the key distribution within the fuzzy commitment framework.

[0110] The sender and the receiver detect the ECG signal of the subject at the same time, and generate binary sequences BS1 and BS2 respectively. For the sender, the key key needs to be error correction code encoded to generate a sequence BS key , and then create a commitment:

[0111]

[0112] In formula (8), hash(key) means one-way hash operation on key, The symbol represents the bitwise XOR operation.

[0113] The commitment will be sent to the receiver, and the receiver first performs operation to generate BS key′ , and then performs error correction code decoding operation on BS key′ to generate key', if hash(key) = hash(key'), the authentication is successful, and the user passes the second level authentication; otherwise, the authentication fails, and the two sensor devices (sender and receiver) that collect the ECG signal will detect the ECG signal of the subject again.

[0114] Experimental analysis

[0115] We use public datasets: CASIA1 iris dataset (108 subjects, each has 7 iris images); MIT-BIH Normal Sinus Rhythm Database (18 subjects, each has 2 synchronized ECG signals), hereinafter referred to as NSRDB; MIT-BIH Arrhythmia Database (48 subjects, each has 2 synchronized ECG signals), hereinafter referred to as MITDB; MIT-BIH Supraventricular Arrhythmia Database (78 subjects, each has 2 synchronized ECG signals), hereinafter referred to as SVDB; to simulate and analyze our system.

[0116] It should be noted that not all samples and all sampling points in the electrocardiogram signal database can be used, for example, the sampling points of sample No. 19830 in NSRDB for about 28 seconds are not in working condition, and we filter and delete similar cases.

[0117] A. Performance of binary sequence

[0118] For the binary sequence generated by the system for the identity authentication system, in addition to the Hamming distance mentioned above, we also introduce entropy analysis and NIST test suite to verify its performance.

[0119] Entropy analysis: the randomness is judged by calculating the entropy of the feature sequence. For a random variable x = {0, 1} n , the entropy value can be calculated by the following formula:

[0120] E(x) = -pmf0log2pmf0-pmf1log2pmf1 (9)

[0121] Where pmf0 is the probability mass function of 0, pmf1 is the probability mass function of 1, and the maximum value of E(x) is 1 when the data obeys uniform distribution.

[0122] NIST test suite: NIST test suite is a famous randomness test tool, which is widely used by researchers to test random numbers and pseudo-random numbers in cryptography. NIST suite has multiple test indicators, and each indicator will output a P value. If the P value is less than the threshold (0.01), it can be assumed that the data is not random.

[0123] 1) Iris encoding test

[0124] We use CASIA1 iris dataset to test the performance of iris encoding. In this database, each subject has 7 iris images from the same eye, and we select 4 iris images to extract a set of iris templates Diris and D irisindex , the remaining 3 pictures as test set.

[0125] Assume that the parameter N is 900 when interval mapping is performed on D f The parameter b is 3 when the bit block method is used on D fnew After iris encoding, each subject will generate a set of iris templates and 3 test encodings.

[0126] After obtaining the iris encoding templates generated by 108 different subjects, we compared the Hamming distances of the encodings two by two, and the test results are shown in Figures 3-4 ; Figure 3 is the Hamming distance distribution chart of the iris encodings of 108 different subjects before using the bit block method (the length of the binary sequence is 900 bits), and the test results show that the average Hamming distance is 0.491973; Figure 4 is the Hamming distance distribution chart of the iris encodings of 108 different subjects after using the bit block method (the length of the binary sequence is 600 bits), and the test results show that the average Hamming distance is 0.470357. They all conform to the normal distribution, and the Hamming distance is close to 0.5, which can indicate that the iris template encodings generated by different individuals are also different.

[0127] For the 3 test encodings of the same subject, we also compared their Hamming distances with the iris encoding templates, and the test results are shown in Figures 5-6 ; Figure 5 is the Hamming distance distribution chart of the iris encodings of the same individual before using the bit block method (the length of the binary sequence is 900 bits), and the test results show that the average Hamming distance is 0.274304; Figure 6 is the Hamming distance distribution chart of the iris encodings of the same individual after using the bit block method (the length of the binary sequence is 600 bits), and the test results show that the average Hamming distance is 0.164763.

[0128] It can be seen that after using the bit block method, although the average Hamming distance of the iris encodings generated by different individuals only decreases slightly, for the same individual, the average Hamming distance decreases faster, that is, the iris encodings generated by each iris image collection are more similar, which is conducive to identity authentication. From this perspective, the bit block method can improve the performance of iris encoding.

[0129] 2) ECG feature encoding test

[0130] In order to compare the performance of different encoding methods, we use NSRDB, MITDB, and SVDB to compare the popular six IPI encoding methods with IPIFSC method. We perform two experiments, "Hamming distance and entropy experiment" and "NIST test". In "Hamming distance and entropy experiment" (Table 2), "average Hamming distance of the whole database" means that 100 times of random sampling are performed on each sample in the database, and a total of 3600 codes are generated, and the Hamming distances are compared; "average Hamming distance of the same sample" means that the Hamming distance of the codes generated by simultaneously sampling two different ECG signals of the same subject is compared; "average entropy value" means the average entropy value of the codes generated by 5000 times of random sampling of the database. The results of "NIST test" are shown in Table 3, and the numbers in the table represent P-value.

[0131] Table 2 Hamming distance and entropy test

[0132]

[0133] Table 3 NIST suite randomness test

[0134]

[0135]

[0136] From the above experimental results, it can be seen that in the Hamming distance and entropy test, the performance of IPIFSC method does not decrease significantly, while in the NIST suite randomness test, the overall performance of IPIDTC method improves slightly. This shows that the IPIFSC method generates codes that are not random even if less R-peak information is sampled.

[0137] B. Performance of two-level authentication system

[0138] False acceptance rate (FAR) and false rejection rate (FRR) are two key indicators for evaluating the performance of an authentication system. We randomly select 108 samples from NSRDB, MITDB, and SVDB, and one-to-one correspond to the 108 samples in CASIA1, to create a virtual database with ECG signals and iris images. In our test, FAR represents the probability that a subject passes through an authentication system that does not belong to him; FRR represents the probability that a subject cannot pass through his own authentication system. It is obvious that when the values of FAR and FRR of a system are lower, the performance of the system is better. In order to prove the effectiveness of the two-level authentication system, we first perform experiments on iris authentication and ECG feature authentication separately, and then combine the two to perform experiments.

[0139] As before, we selected four iris images from each sample in CASIA1 for template extraction, and used the remaining three images for testing. The individual iris recognition performance test results are shown in Table 4.

[0140] Table 4 Iris Recognition Performance Test

[0141]

[0142] As can be seen from Table 4, when the iris recognition threshold TH... iris A value of 0.43 is appropriate.

[0143] We will conduct experiments using 316 ECG signals from 108 randomly selected samples from NSRDB, MITDB, and SVDB. The individual ECG feature recognition performance test results are as follows: Figure 7 As shown, the codeword length of the BCH is 255 bits at this time; Figure 7 In the experiment, FRR decreased with the improvement of BCH error correction capability, and reached a minimum value of 0.032170 under the BCH(255, 9, 63) parameter; while FAR increased with the improvement of error correction capability, and reached a maximum value of 0.020215 under the BCH(255, 9, 63) parameter.

[0144] After combining iris and ECG features, the recognition performance of the secondary authentication system is as follows: Figure 8 As shown, in our virtual database, when the iris recognition threshold TH... iris When the value is 0.43, the FRR reaches its minimum value of 0.032287 under the BCH(255, 9, 63) parameter. Regardless of how the error correction capability of the BCH code changes, our secondary authentication system FAR is always close to 0, which means that the system only accepts legitimate visitors, which is very important for patients.

[0145] This invention provides a two-level authentication system combining iris information and ECG signals. Each patient's iris information is compiled into a binary template of a certain length and pre-stored in the sensor. Users must pass the first-level iris authentication before proceeding to the second-level ECG signal feature authentication. Since WBAN devices have limited resources, to avoid consuming excessive hardware resources when encoding iris information and ECG features, we applied the bit block method to iris encoding. Simultaneously, we compared the performance of six popular IPI encoding methods and fused them to create a new encoding method, IPIFSC. Simulation tests on a public database demonstrate that these two encoding methods can save WBAN resource space and shorten encoding time. Due to the two-level authentication, our system achieves a near-zero FAR (Free Anonymous) value, which is crucial for protecting user privacy.

Claims

1. A two-level WBAN authentication method based on iris features and ECG features, characterized in that, Includes the following steps: Step S1: After acquiring iris images using an iris scanner, the acquired iris images are preprocessed using the integral-differential operator proposed by Dougman. Then, the iris images are feature extracted and quantized using a Log Gabor filter. An iris template is then created, the quantization length is simplified, and Hamming distance is introduced to determine the similarity between the test sample and the iris template. Iris template matching is then performed to conduct the first level of identification and authentication of the user's identity. Step S2: After passing the first-level authentication, the user's ECG signal is collected using a sensor. First, wavelet transform is used to preprocess the ECG signal, then feature extraction is performed on the preprocessed ECG signal, and then the ECG feature is quantized using a fusion method. Finally, a key is distributed through a fuzzy commitment key negotiation mechanism to perform the second-level authentication of the user's identity. Only users who pass these two levels of authentication are considered to have passed the authentication. In step S1, the specific steps for creating the iris template are as follows: First, select the clearest iris image and call its corresponding binary iris code the reference code. Then, record the reference codes in sequence. The indices that have the same encoding values ​​as other iris codes in the same eye are used to form the intersection of these indices, which is called the pretemplate. Pre-template The definition is as follows: (3) In equation (3), It refers to and An index sequence with the same position; Secondly, the length of the pre-template obtained is inconsistent for different eyes. The sequence is mapped in the interval [0, N), and the N indices closest to the integer (0, 1, 2, ..., N-1) are selected as new pre-templates. ; that is, different lengths Convert to the same length ; Then, to To further simplify, the specific process is as follows: The binary sequence is divided into bit blocks of length b, where b is an odd number. Then, the middle element of each bit block is XORed with all other elements in the same block sequentially. After the XOR operations are complete, all bit blocks are concatenated to obtain a new binary sequence, i.e., a new index sequence. and in reference coding Chinese Press The value in the index is used to extract the corresponding binary number, resulting in the final iris coding template. ; In step S2, the specific steps for ECG feature quantization are as follows: obtain... Then, a fusion method is used to quantize it and generate a binary sequence. The specific process is as follows: First, input the acquired IPI sequence. ,exist Selecting n consecutive IPIs is called... ; Secondly, calculation average ; then calculate Each IPI within The difference is calculated, and the results of each difference are stored in... In the middle, calculation Mathematical expectation and variance Calculate 16 domains, and Elements falling into the corresponding field are converted into their corresponding Gray codes; at this point, the binary sequence BS1 is generated. Compare The trend of IPI changes; if the value of the previous IPI is less than the value of the next IPI, that is... If the i-th element is quantized to 1, then the i-th element is quantized to 1; otherwise, it is padded with 0. The last element in the sequence is padded with random 0s or 1s; this generates the binary sequence BS2. Will After each element is quantized into 8 bits, only 2-4 bits are used, thus generating the binary sequence BS3. Finally, BS1, BS2, and BS3 are concatenated using the same index to obtain the final binary sequence BS.

2. The two-level WBAN authentication method based on iris features and ECG features according to claim 1, characterized in that, In step S1, the specific steps of iris image preprocessing are as follows: First, the iris localization is performed on the acquired iris image using the integral-differential operator proposed by Dougman. This operator is defined as: (1) In equation (1), It refers to the pixel coordinates of the captured image containing the eye. It is a Gaussian smoothing function. The symbol represents the convolution operation. It refers to For radius, The line integral over a circle centered at the given point. Then, using Hough transform and Canny edge detection, straight and curved sections in the iris image are found, and the iris region is normalized to obtain a standardized iris image.

3. The two-level WBAN authentication method based on iris features and ECG features according to claim 2, characterized in that, In step S1, the specific steps for feature extraction and quantization of the iris image are as follows: Feature extraction and quantization of iris images are performed using a Log Gabor filter. The frequency response formula of the Log Gabor filter is as follows: (2) In equation (2), Represents the center frequency of the filter. Represents the bandwidth of the filter; The specific process of extraction and quantization is as follows: each row of the standardized iris image is regarded as a one-dimensional vector and convolved with a Log Gabor filter. The convolution result has real and imaginary parts. The positive and negative values ​​of the real and imaginary parts are recorded, with 1 representing a positive number and 0 representing a negative number. Each pixel is quantized into 2 bits. Finally, the calculation results are concatenated row by row to quantize the iris image into a binary iris code of length X bits.

4. The two-level WBAN authentication method based on iris features and ECG features according to claim 1, characterized in that, In step S1, the specific steps for iris template matching are as follows: Hamming distance is introduced to determine the similarity between the test sample and the iris template; an iris threshold is then used. When the Hamming distance between the sample to be tested and the iris template is less than the threshold At that time, the user passes the first level of iris authentication.

5. The two-level WBAN authentication method based on iris features and ECG features according to claim 1, characterized in that, In step S2, the ECG signal preprocessing is performed using wavelet transform. The specific steps are as follows: First of all, with Represents the set of all signals with finite energy, when satisfying When, define the signal The formula for continuous wavelet transform is: (4) In equation (4), It is a scale factor. It is the translation factor. express The complex conjugate of the continuous wavelet transform needs to be determined; then, the continuous wavelet transform needs to be discretized into binary form, and the scaling factor of the continuous wavelet transform needs to be determined. Translation factor Discretization yields the discrete wavelet sequence, which is then discretized into... Where j represents the decomposition scale and Z represents the set of all integers, the discrete wavelet transform is expressed as: ; Secondly, the Sym4 wavelet is selected as the basis function; the threshold function between the soft and hard compromise is selected, and the expression of the function is: (5) In equation (5), This represents the wavelet coefficients after thresholding. The decomposed wavelet coefficients are represented by λ, where λ represents the threshold. The value of λ is within the interval (0,1); the calculation formula for the threshold λ is as follows: (6) In equation (6), Represents variance. This represents the signal length.

6. The two-level WBAN authentication method based on iris features and ECG features according to claim 5, characterized in that, In step S2, the specific feature extraction of the preprocessed ECG signal involves searching for features in the ECG signal. Peak and calculation The IPI sequence corresponding to the peak is denoted as The definition of IPI is as follows: (7) In equation (7), Refers to the first One heartbeat cycle Peaks, or IPIs, are considered a characteristic of ECG signals.

7. The two-level WBAN authentication method based on iris features and ECG features according to claim 6, characterized in that, In step S2, the specific steps for distributing the key through the fuzzy commitment key negotiation mechanism are as follows: Within the fuzzy commitment framework, a key distribution block (BS) generated by IPI is used to protect key distribution; the sender and receiver simultaneously detect the subject's ECG signal and generate binary sequences respectively. and ; For the sender, the key is required. Perform error correction code encoding processing to generate and Equal-length sequences Then create a commitment: (8) In equation (8), Point to Perform a one-way hash operation. The symbol represents a bitwise XOR operation; The commitment It will be sent to the receiving end, and the receiving end will first perform... Operation generation Then to Generate error correction code by performing decoding operation ,like If the authentication is successful, the user passes the second-level authentication; otherwise, the authentication fails, and the sensor device will re-detect the subject's ECG signal.

Images