CN114121305B - Sensor safety detection method and system based on sweep frequency technology - Google Patents
Sensor safety detection method and system based on sweep frequency technology Download PDFInfo
- Publication number
- CN114121305B CN114121305B CN202111394732.XA CN202111394732A CN114121305B CN 114121305 B CN114121305 B CN 114121305B CN 202111394732 A CN202111394732 A CN 202111394732A CN 114121305 B CN114121305 B CN 114121305B
- Authority
- CN
- China
- Prior art keywords
- frequency
- sensor
- sweep
- data
- detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 89
- 238000005516 engineering process Methods 0.000 title claims abstract description 23
- 238000012360 testing method Methods 0.000 claims abstract description 29
- 230000005284 excitation Effects 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000010408 sweeping Methods 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 238000000034 method Methods 0.000 claims description 33
- 230000002159 abnormal effect Effects 0.000 claims description 18
- 238000003062 neural network model Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 11
- 230000000737 periodic effect Effects 0.000 claims description 8
- 238000013480 data collection Methods 0.000 claims description 7
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 230000005856 abnormality Effects 0.000 claims description 2
- 230000003287 optical effect Effects 0.000 claims description 2
- 230000011218 segmentation Effects 0.000 claims description 2
- 239000000126 substance Substances 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 abstract description 5
- 238000010801 machine learning Methods 0.000 abstract description 5
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000001228 spectrum Methods 0.000 abstract description 2
- 230000001133 acceleration Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012806 monitoring device Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 230000002441 reversible effect Effects 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 240000007651 Rubus glaucus Species 0.000 description 1
- 235000011034 Rubus glaucus Nutrition 0.000 description 1
- 235000009122 Rubus idaeus Nutrition 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000005389 magnetism Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
- G06F2218/16—Classification; Matching by matching signal segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
- G16Y30/00—IoT infrastructure
- G16Y30/10—Security thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Signal Processing (AREA)
- Computer Security & Cryptography (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a sensor vulnerability detection method and system based on a frequency sweeping technology, comprising the following steps: step one, generating a sweep frequency excitation signal; step two, generating out-of-band signals; step three, testing out-of-band signals; step four, collecting sensor data; step five, preprocessing sensor data; step six, model training stage; step seven, detecting out-of-band vulnerability; and step eight, generating a test report. The system can detect out-of-band vulnerability of the sensor, has good generalization capability, can conveniently detect out-of-band vulnerability of any type of sensor, and can cover sensor vulnerabilities in full frequency spectrum ranges. The system adopts a digital sweep frequency technology and combines a machine learning algorithm to realize automatic detection and analysis of sensor vulnerability, thereby providing a good reference for safety protection of a sensing system in the Internet of things.
Description
Technical Field
The invention belongs to the field of vulnerability detection, and particularly relates to a sensor safety detection method and system based on a frequency sweeping technology.
Background
Along with the continuous development of the Internet of things and the continuous intellectualization of the terminal equipment, the sensor is used as the terminal equipment to sense the 'eyes' of the physical world, is a tie for the interactive connection of the Internet of things and the physical world, and is widely used in an Internet of things system. However, the security risk of the sensor is not paid enough attention, and in general, the upper layer defaults that all data acquired by the sensor are trusted, and once the sensor is attacked, the measured data of the sensor are tampered, so that the blind trust of hardware causes a huge threat to the security operation of the whole internet of things. If the automatic driving system makes driving decisions based on the sensing information of the sensor, when the sensing signal of the sensor is wrong, the wrong driving decisions are led out, so that serious consequences are caused. Therefore, ensuring the testing accuracy of the sensor is important to the safe operation of the Internet of things system.
Traditional sensor tests focus on performances such as sensitivity, precision and the like, and test is performed by using in-band signals, and the fact that the sensor is possibly affected by out-of-band signals is not considered, so that safety problems occur. The in-band (in-band) signal refers to a detection signal within the design range of the sensor, otherwise referred to as an out-of-band (out-of-band) signal, for example, the speed detection of the accelerometer is in-band signal detection, and the acoustic detection of the accelerometer is out-of-band signal detection. In a real application scenario of the internet of things based on sensor perception, most emergencies or malicious attacks are caused by out-of-band vulnerabilities of sensors. In addition, except for a hardware factory test, the existing multi-sensor fusion technology can reduce risks caused by abnormal operation of a single sensor to a certain extent, but does not solve the problem of out-of-band vulnerability of the sensor from the source.
In conclusion, how to realize the out-of-band vulnerability detection system of the sensor, which is convenient to operate, strong in portability, accurate and reliable in measurement, has great significance for safety research and protection of the Internet of things system based on sensor perception.
Disclosure of Invention
The invention aims to provide a sensor safety detection method and system based on a frequency sweeping technology, and aims to solve the problem that the existing sensor performance detection only detects in-band signals and extreme bearing environments of a sensor, but lacks out-of-band signal vulnerability detection and has limitation in practical application scenes.
The invention adopts the following technical scheme for solving the problems:
a sensor safety detection method based on sweep frequency technology comprises the following steps:
step one, an excitation signal generation stage: setting excitation signal types, sweep initial frequency, sweep cut-off frequency, equal interval sweep frequency point number, excitation signal amplitude and same frequency signal detection duration parameters by adopting a digital sweep frequency technology, and outputting a periodic excitation electric signal by an arbitrary waveform generator;
step two, an out-of-band signal generation stage: converting the periodic excitation electric signal into an out-of-band signal to be tested by using a transducer;
step three, an out-of-band signal testing stage: according to the contact type of the out-of-band signal to be tested and the sensor, the out-of-band signal to be tested is acted on the sensor and equipment for installing the sensor;
step four, test data collection stage: acquiring output data of a sensor and characteristic data of equipment provided with the sensor, and preprocessing the output data to obtain normal sample data;
step five, adding interference in the step three, and repeating the step three and the step four to obtain abnormal sample data;
step six, constructing a neural network model, and training the model by using normal sample data and abnormal sample data;
and seventhly, aiming at the sensor to be detected, performing safety detection by using a trained model, and in the safety detection process, firstly, executing frequency sweep in a wide frequency band range, and searching the optimal vulnerability frequency point according to the frequency sweep result and a dichotomy.
Further, the type of the out-of-band signal includes any one of acoustic, optical, electrical, magnetic, thermal, and chemical signals.
Further, the characteristic data of the sensor-mounted device comprises the working temperature and the vibration amplitude of the device.
Further, the preprocessing process of the data comprises the following steps:
4.1 Taking the sensor output data and the characteristic data of the equipment provided with the sensor as raw data, and carrying out filtering and normalization processing on the raw data;
4.2 Dividing the normalized data into a plurality of samples according to the number of the frequency sweeping frequency points at equal intervals.
Further, the length of the abnormal sample is the same as the length of the normal sample.
Further, in the training of the model in the step six, the probability value in the range of (0, 1) is outputted from the model, and the probability value is smaller, the probability of abnormality of the sensor is greater.
Further, the step seven specifically includes:
7.1 Firstly, setting the number of m equally-spaced frequency sweeping frequency points in a broadband range [ L, R ], wherein L is the frequency sweeping initial frequency, R is the frequency sweeping cut-off frequency, and m is an even number; taking the output value of the to-be-detected sensor, the vibration amplitude of equipment installed by the to-be-detected sensor and the working temperature data which are obtained in the frequency band range as raw data, and carrying out filtering and normalization preprocessing on the raw data;
7.2 Dividing the preprocessed data into m sections according to the number of the equidistant frequency sweeping points, wherein the length of a sample to be tested of each section is consistent with the length of a training sample when training a neural network model, and m sample data to be tested are obtained;
7.3 Taking m pieces of sample data to be detected as input of a trained neural network model according to the time sequence of segmentation to obtain m output results;
7.4 And (3) taking the average value of the output of the normal sample in the training process as a standard value, respectively calculating the average value of the output of the first half part and the average value of the output of the second half part of the model, comparing the average value with the standard value, resetting the frequency band range in the first half part or the second half part by a bisection method if the average value of the first half part or the second half part is smaller than the standard value and the difference value of the average value and the average value exceeds the detection threshold value, setting the frequency sweep range of the first half part as [ L, (L+R)/2 ], or setting the frequency sweep range of the second half part as [ (L+R)/2, R ], and repeating the steps 7.1) to 7.4) until the frequency sweep range is lower than the frequency sweep threshold value, and outputting the corresponding fragile frequency point.
Further, the method also comprises a step eight of generating a detection report according to a detection result; the detection report comprises a sweep frequency detection object, a sweep frequency range, test precision, a vulnerability frequency point and a sensor output value under the vulnerability frequency point.
A sensor safety detection system based on a sweep frequency technology is used for realizing the sensor safety detection method.
The sensor out-of-band vulnerability detection system based on the sweep frequency technology is reasonable in design, simple in structure, convenient to operate, strong in generalization capability and reliable in detection result. The main beneficial effects include:
(1) Aiming at the out-of-band vulnerability detection system of the sensor, the invention breaks through the defect that the traditional performance detection is only carried out on an in-band signal domain, perfects the performance detection system of the sensor, and has important significance for guaranteeing the reliability of perceived data in the Internet of things system.
(2) The invention adopts a digital sweep frequency technology to provide stable excitation signals for the transducer element, the transducer type in the prior art can cover the out-of-band leak detection of the sensor in the category 104 of the sound, light, electricity, magnetism, heat and chemistry, and the detection can be divided into contact detection and non-contact detection according to whether the detection signals directly contact the sensor to be detected, wherein the contact detection is to directly apply the out-of-band signals to the sensor and equipment for installing the sensor, and the non-contact detection is to apply the out-of-band signals to the sensor and equipment for installing the sensor in a non-contact manner, such as through air propagation and the like. Therefore, the invention establishes a complete out-of-band vulnerability detection system which is irrelevant to the sensor type and has good generalization performance.
(3) In the out-of-band vulnerability detection system of the sensor, a high-efficiency digital sweep frequency strategy is adopted to provide stable excitation signals, so that the out-of-band signal detection efficiency of the sensor is effectively improved; an abnormal detection model based on a machine learning algorithm is introduced, detection tasks are converted into classification problems, and detection accuracy and detection speed are further improved.
Drawings
FIG. 1 is a schematic overall flow diagram of a sensor safety detection system based on frequency sweep technology of the present invention;
FIG. 2 is a schematic diagram of out-of-band signal generation based on a swept frequency technique in accordance with the present invention;
FIG. 3 is a schematic diagram of the data acquisition and preprocessing process of the present invention;
FIG. 4 is a schematic illustration of vulnerability detection based on machine learning algorithm of the present invention;
FIG. 5 is a flow chart of the present invention employing a modified dichotomy to find fragile points;
FIG. 6 is a diagram illustrating the content structure of a test report in accordance with an embodiment of the present invention.
Detailed Description
The invention provides a sensor safety detection method and a sensor safety detection system based on a sweep frequency technology, which utilize the characteristic that a sensor outputs abnormal data at an out-of-band weak point, realize automatic vulnerability mining and detection by combining a machine learning classification algorithm, can effectively detect the vulnerability of the sensor in out-of-band physical signals, and overcome the defect that the traditional sensor detection system only aims at in-band signal detection. The detection method and the detection system complement the defect of out-of-band signal loopholes of the sensor in the whole frequency spectrum range of acoustic wave resonance loopholes, filter loopholes, saturation loopholes, frequency mixing loopholes, photoelectric coupling loopholes, electromagnetic coupling loopholes, nonlinear intermodulation distortion loopholes, envelope extraction loopholes and the like.
The following describes embodiments and technical schemes of the present invention in further detail with reference to the accompanying drawings:
as shown in fig. 1, the following describes an embodiment of the present invention for safety detection of a sensor based on a frequency sweep technology, taking a triaxial acceleration sensor for out-of-acoustic-band vulnerability detection as an example. In this embodiment, the sweep signal source module uses an arbitrary waveform generator (DG 4012, supporting maximum frequency of 100MHz, sampling rate of 500 MSa/s) and a high-voltage amplifier (high-power amplifier (HAS 4051, supporting maximum signal frequency of 500 KHz) of NF corporation, a high-voltage radio-frequency amplifier (ZHL-100W-gan+, supporting amplification frequency of 20-500 MHz) of coaxal corporation. The out-of-band signal transmitting module is a high-power loudspeaker, the sensor to be detected is a triaxial acceleration sensor, the sensor data acquisition platform and the microcontroller module are integrated on the raspberry group 4B, and the display module adopts a high-definition liquid crystal display screen.
Step 1: and generating an excitation signal. The sound wave frequency sweep test system on the control host is turned on, sound wave frequency sweep test parameters are set in a system interface, and the sound wave frequency sweep test parameters comprise excitation signal types, frequency sweep starting frequency, frequency sweep cut-off frequency, the number of frequency sweep frequency points at equal intervals, excitation signal amplitude and same-frequency signal detection duration, as shown in figure 2. Clicking a start button, transmitting a control instruction to a sweep frequency signal source module, namely an arbitrary waveform generator, and enabling the arbitrary waveform generator to enter a sweep frequency mode to generate an original excitation signal periodically changing at equal intervals. The original excitation signal is amplified by a high-voltage amplifier, and the high-voltage amplifier transmits the amplified excitation signal to an out-of-band signal transmitting module, namely a high-power loudspeaker.
Step 2: out-of-band acoustic wave signal generation. The out-of-band sound wave signal generation stage mainly refers to that the signal output end of the high-voltage amplifier is connected with the input end of the loudspeaker, and the loudspeaker converts a periodic sweep frequency excitation signal into a periodic variable-frequency out-of-band sound wave signal.
Step 3: the out-of-band signal test, because the acoustic wave signal mainly propagates through the air, a non-contact test mode is adopted in the test stage, the loudspeaker is close to the sensor module to be tested, namely the anti-shake module of the camera, the out-of-band acoustic wave signal propagates to the receiving end of the sensor to be tested through the air and acts on the MEMS vibration module of the triaxial accelerometer, so that abnormal sensor output and platform shake and heat are caused.
Step 4: sensor data collection. As shown in fig. 3, the data collection stage includes accessing the triaxial acceleration sensor to be tested to a sensor data collection platform and fixing the sensor data collection platform on a mechanical device, wherein the data collection platform integrates an analog-to-digital conversion module supporting high input frequency and having high linearity and ultra-low power consumption. Measuring the vibration amplitude of the sensor by adopting a vibration monitoring device, and connecting the output of the vibration monitoring device with a data acquisition platform; the infrared temperature measuring device is used for collecting the real-time working temperature of the sensor, and the output of the infrared temperature measuring device is connected with the data acquisition platform.
After the sensor to be tested is electrified to work, the data acquisition platform adopts a high-speed analog-to-digital converter to convert the analog output signals of the sensor triaxial, the vibration monitoring device and the infrared temperature measuring device into stable digital signals in real time, and stores the stable digital signals in time sequence. The collected data are divided into normal sample data and abnormal sample data, the normal sample data refers to data collected in a state that the sensor is stable and normal in operation, and the normal sample data are obtained by converting analog output signals of the sensor into stable digital signals in real time by adopting a high-speed analog-to-digital converter and storing the stable digital signals; abnormal sample data refers to data acquired when the sensor output is abnormal due to the artificial external interference, such as abnormal output of the sensor caused by the short-time vibration accelerometer.
Step 5: sensor data preprocessing, as shown in fig. 3, since the triaxial accelerometer is affected by ambient noise in a static state, a filtering operation is performed on raw data in a preprocessing stage. Considering the difference of the magnitude and magnitude of the output signals of different sensors, vibration monitoring and temperature monitoring platforms, the original data x is needed to improve the generalization of the detection algorithm 1 ,x 2 ,x 3 ,x 4 ,x 5 Normalization processing is carried out to obtain a preprocessed data sample x' 1 ,x′ 2 ,x′ 3 ,x′ 4 ,x′ 5 In this example, a normalization method is used, as shown in the following formula.
The preprocessed 50% of the data are used as training sets, and the rest are used as test sets.
Wherein x is 1 ,x 2 ,x 3 ,x 4 ,x 5 Respectively representing the x-axis output, the y-axis output, the z-axis output, the vibration amplitude and the working temperature of the acceleration sensor; x ' represents the normalized data set x ' = [ x ] ' 1 ,x′ 2 ,x′ 3 ,x′ 4 ,x′ 5 ]X represents the original data set x= [ x ] 1 ,x 2 ,x 3 ,x 4 ,x 5 ]Mu represents the mean set mu= [ mu ] of the raw data 1 ,μ 2 ,μ 3 ,μ 4 ,μ 5 ]σ represents the variance set σ= [ σ ] of the original data 1 ,σ 2 ,σ 3 ,σ 4 ,σ 5 ]。
Step 6: and in the model training stage, the preprocessed sensor training set data under a certain test frequency is segmented into a plurality of samples at equal intervals, the samples are input into the input end of the machine learning model, the weight is initialized randomly, and the gradient descent algorithm is adopted to optimize the model parameters. The method comprises the following specific steps:
step 6.1: as shown in fig. 4, a neural network model consisting of an input layer, a hidden layer and an output layer is built, the vulnerability detection problem is converted into a classification problem, and as each group of data has 5 inputs, the input layer consists of 5 neural units;
step 6.2: initializing parameters of a neural network model, wherein the weight w of each neuron is generated randomly by generating initialization weights n And offset value b n Initialized to a random number, where w n Representing the weight matrix at the nth iteration, b n Representing a neural network bias value matrix at the nth iteration; setting an activation function, namely 6.3: pre-training a neural network model by using a training set, and firstly calculating a neural network activation value:
where n represents the number of iterations,representing an output value of the hidden layer of the sample under the nth iteration; o (O) n Representing an activation function at the nth iteration to enhance model nonlinearity, outputting hidden layerThe values are concentrated to a probability value in the (0, 1) range and are used as the final output of the neural network model.
Step 6.4: and calculating errors, carrying out error reverse transmission on the neural network model according to the loss function value, and optimizing parameters and learning rate by adopting a gradient descent method. Let the model predicted output value of the sample data be P (where the model output value of the normal sample data p=1, the model output value of the abnormal sample data p=0), the output layer error Err for the nth iteration n Can be expressed as:
the updating weight is as follows by error reverse transmission:
wherein, beta represents the learning rate,the neuron weights at the nth iteration are represented. When the output result of the proposed neural network is higher than 0.5, the sample to be tested is considered to be normal, namely the sensor to be tested has no weak point in the out-of-band signal of the current frequency point; otherwise, the sample is abnormal, namely the out-of-band signal of the current frequency point is the out-of-band fragile frequency point of the sensor. The training process is not stopped until the predicted error rate is lower than a certain threshold value, so that a trained network is obtained; b n+1 Is the neural network bias value, b at the n+1th iteration n Is the neural network bias value at the nth iteration.
Step 7: out-of-band vulnerability detection, in the detection stage, a trained neural network model is adopted to convert the abnormal detection problem into the classification problem, and a dichotomy is combined to search the optimal vulnerability frequency point, so that the detection efficiency is effectively improved, and a specific process is shown in a flowchart 5.
Step 7.1: first in the wide frequency band range [ L, R]Setting m equally-spaced frequency sweeping frequency points, preprocessing output values of to-be-detected sensors, vibration amplitude and working temperature data of equipment installed by the to-be-detected sensors, which are obtained in the frequency band range, according to the method of step 5, dividing the preprocessed data into m sections at equal intervals according to the frequency points, wherein the length of to-be-detected samples of each section is consistent with the length of training samples when training a neural network model, sequentially inputting m pieces of to-be-detected sample data obtained under the m frequency points into the trained neural network model to obtain m model output O= [ O ] 1 ,O 2 ,...,O m-1 ,O m ]M is an even number.
Step 7.2: the improved binary search method searches for fragile frequency points. Because the vulnerability frequency points appear in a certain narrow bandwidth range and a plurality of vulnerability points possibly exist, the workload is increased if each narrow-band frequency point is subjected to one-by-one frequency sweep detection in a full frequency band, and the frequency sweep frequency can be effectively reduced by adopting an improved binary search method, the search speed is improved, the vulnerability points can be accurately positioned, and the detection performance is enhanced.
The specific implementation method comprises the following steps: and (3) calculating the average value of the front half part output and the average value of the rear half part output of the model, comparing the average value of the front half part output and the average value of the rear half part output with the average value of the normal sample model output, if the front half part (or the rear half part) is smaller than the average value of the normal sample model output and the difference value of the front half part (or the rear half part) is larger than a detection threshold value, redefining the frequency sweep range in the front half part (or the rear half part), for example, setting the frequency sweep range as [ L, (L+R)/2 ] if the average value of the front half part output is smaller than the average value of the normal sample model output, setting the frequency sweep range as [ (L+R)/2, R ] if the average value of the rear half part output is smaller than the average value of the normal sample model output, repeating the steps 7.1 to 7.2 until the frequency sweep range is lower than the threshold value, and outputting the corresponding fragile frequency point, namely outputting the final frequency sweep range. Note that in the detection process, there may be a plurality of fragile frequency points, so if both the front and rear portions are smaller than the output average value of the normal sample model, it is necessary to perform dichotomy detection on both the front and rear portions.
In this embodiment, through out-of-band vulnerability detection in step 7, the sensor is swept from a frequency range of 1-100KHz, the number of equally spaced frequency sweeping points is set to 10, it is found by a binary search method that the difference between the output average value of the model output average value in the first half frequency range and the output average value of the normal sample model output average value exceeds a detection threshold, and the difference between the output average value in the second half frequency range and the output average value of the normal sample model output average value does not exceed the detection threshold, so that the vulnerability can be determined to be in the range of 1-50 KHz. Similarly, frequency subdivision is further carried out on [1KHz,50KHz ], frequency sweeping and judgment are carried out, and finally the vulnerability frequency point of the out-of-band sound wave signal of the accelerometer is detected to be about 0.8KHz and about 3KHz, the vulnerability of the accelerometer at about 0.8KHz is caused by the resonance of the Y axis, and the vulnerability at about 3KHz is caused by the resonance of the X axis, so that the vulnerability type is sound wave resonance vulnerability.
Step 8: a test report is generated. After the sweep test is finished, a complete test report is generated. The test report content comprises a sweep frequency detection object (a triaxial acceleration sensor), a test frequency point range (1 Hz-100 KHz), test accuracy (100 Hz), vulnerability frequency points (0.8 KHz and 3 KHz) and sensor output values under the vulnerability frequency points (when the frequency is 0.8KHz, the Y-axis output value of the sensor is 0.19g/s, and when the frequency is 3KHz, the X-axis output value of the sensor is 0.42 g/s). Through further professional analysis, the test report can also comprise the type of the loopholes of the sensor to be tested, the application field, the test principle, the protection suggestion and the like, and the specific content is shown in fig. 6. Aiming at the out-of-band acoustic wave signal vulnerability detected by the acceleration sensor in the case is an acoustic wave resonance vulnerability, the protection proposal is provided by adopting an acoustic wave blocking method, namely, adding a layer of acoustic wave shielding material on the periphery of the sensor to shield or attenuate malicious acoustic wave signals.
Corresponding to the foregoing embodiment of a method for detecting sensor security based on sweep frequency technology, the present application further provides an embodiment of a system for detecting sensor security based on sweep frequency technology, which includes:
the frequency sweep signal source module generates a periodic excitation electric signal according to the excitation signal type, the frequency sweep starting frequency, the frequency sweep cut-off frequency, the number of frequency points of equidistant frequency sweep, the amplitude of the excitation signal and the detection duration parameter of the same-frequency signal.
The out-of-band signal transmitting module is used for converting the periodic excitation electric signal into an out-of-band signal to be tested, and applying the out-of-band signal to be tested to the sensor and equipment for installing the sensor according to the contact type of the out-of-band signal to be tested and the sensor.
And the sensor data acquisition platform acquires output data of the sensor and characteristic data of equipment provided with the sensor, and preprocesses the data.
And the neural network model training module is used for training the model by using the normal sample data and the abnormal sample data.
And the detection module is used for carrying out safety detection by utilizing the trained model, and in the safety detection process, firstly, frequency sweep is carried out in a wide frequency band range, and the optimal vulnerability frequency point is searched according to the frequency sweep result and the dichotomy.
And the display module is used for displaying the detection result.
For system embodiments, reference is made to the description of method embodiments for the relevant points, since they essentially correspond to the method embodiments. The system embodiments described above are merely illustrative, and the detection modules may or may not be physically separate. In addition, each functional module in the present invention may be integrated in one processing unit, each module may exist alone physically, or two or more modules may be integrated in one unit. The integrated modules or units can be realized in a hardware form or a software functional unit form, so that part or all of the modules can be selected according to actual needs to realize the purpose of the scheme.
The foregoing list is only illustrative of specific embodiments of the invention. Obviously, the invention is not limited to the above embodiments, but many variations are possible. All modifications directly derived or suggested to one skilled in the art from the present disclosure should be considered as being within the scope of the present invention.
Claims (8)
1. The sensor safety detection method based on the sweep frequency technology is characterized by comprising the following steps of:
step one, an excitation signal generation stage: setting excitation signal types, sweep initial frequency, sweep cut-off frequency, equal interval sweep frequency point number, excitation signal amplitude and same frequency signal detection duration parameters by adopting a digital sweep frequency technology, and outputting a periodic excitation electric signal by an arbitrary waveform generator;
step two, an out-of-band signal generation stage: converting the periodic excitation electric signal into an out-of-band signal to be tested by using a transducer;
step three, an out-of-band signal testing stage: according to the contact type of the out-of-band signal to be tested and the sensor, the out-of-band signal to be tested is acted on the sensor and equipment for installing the sensor;
step four, test data collection stage: acquiring output data of a sensor and characteristic data of equipment provided with the sensor, and preprocessing the output data to obtain normal sample data;
step five, adding interference in the step three, and repeating the step three and the step four to obtain abnormal sample data;
step six, constructing a neural network model, and training the model by using normal sample data and abnormal sample data;
step seven, aiming at a sensor to be detected, carrying out safety detection by utilizing a trained model, in the safety detection process, firstly, executing frequency sweep in a wide frequency band range, and searching an optimal vulnerability frequency point according to a frequency sweep result and a dichotomy; the seventh step is specifically as follows:
7.1 Firstly, setting the number of m equally-spaced frequency sweeping frequency points in a broadband range [ L, R ], wherein L is the frequency sweeping initial frequency, R is the frequency sweeping cut-off frequency, and m is an even number; taking the output value of the to-be-detected sensor, the vibration amplitude of equipment installed by the to-be-detected sensor and the working temperature data which are obtained in the frequency band range as raw data, and carrying out filtering and normalization preprocessing on the raw data;
7.2 Dividing the preprocessed data into m sections according to the number of the equidistant frequency sweeping points, wherein the length of a sample to be tested of each section is consistent with the length of a training sample when training a neural network model, and m sample data to be tested are obtained;
7.3 Taking m pieces of sample data to be detected as input of a trained neural network model according to the time sequence of segmentation to obtain m output results;
7.4 And (3) taking the average value of the output of the normal sample in the training process as a standard value, respectively calculating the average value of the output of the first half part and the average value of the output of the second half part of the model, comparing the average value with the standard value, resetting the frequency band range in the first half part or the second half part by a bisection method if the average value of the first half part or the second half part is smaller than the standard value and the difference value of the average value and the average value exceeds the detection threshold value, setting the frequency sweep range of the first half part as [ L, (L+R)/2 ], or setting the frequency sweep range of the second half part as [ (L+R)/2, R ], and repeating the steps 7.1) to 7.4) until the frequency sweep range is lower than the frequency sweep threshold value, and outputting the corresponding fragile frequency point.
2. The method of claim 1, wherein the type of out-of-band signal comprises any of acoustic, optical, electrical, magnetic, thermal, chemical signals.
3. The method for safely detecting the sensor based on the sweep frequency technology according to claim 1, wherein the characteristic data of the equipment for installing the sensor comprises the working temperature and the vibration amplitude of the equipment.
4. The method for detecting the safety of the sensor based on the sweep frequency technology according to claim 1, wherein the preprocessing process of the data is as follows:
4.1 Taking the sensor output data and the characteristic data of the equipment provided with the sensor as raw data, and carrying out filtering and normalization processing on the raw data;
4.2 Dividing the normalized data into a plurality of samples according to the number of the frequency sweeping frequency points at equal intervals.
5. The method for detecting sensor safety based on the sweep frequency technique according to claim 4, wherein the length of the abnormal sample is the same as the length of the normal sample.
6. The method for detecting the safety of a sensor based on the sweep frequency technique according to claim 1, wherein the probability value in the range of (0, 1) is outputted by the model during the training of the sixth model, and the probability value is smaller the greater the possibility of abnormality of the sensor.
7. The method for safely detecting a sensor based on the sweep frequency technology according to claim 1, further comprising the step of generating a detection report according to the detection result; the detection report comprises a sweep frequency detection object, a sweep frequency range, test precision, a vulnerability frequency point and a sensor output value under the vulnerability frequency point.
8. A sensor safety detection system based on sweep frequency technology, which is used for realizing the sensor safety detection method of claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111394732.XA CN114121305B (en) | 2021-11-23 | 2021-11-23 | Sensor safety detection method and system based on sweep frequency technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111394732.XA CN114121305B (en) | 2021-11-23 | 2021-11-23 | Sensor safety detection method and system based on sweep frequency technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114121305A CN114121305A (en) | 2022-03-01 |
| CN114121305B true CN114121305B (en) | 2024-02-23 |
Family
ID=80440248
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111394732.XA Active CN114121305B (en) | 2021-11-23 | 2021-11-23 | Sensor safety detection method and system based on sweep frequency technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114121305B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115021971B (en) * | 2022-05-10 | 2023-06-09 | 电子科技大学 | Hardware security testing method for FPGA cloud platform |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104374468A (en) * | 2014-10-24 | 2015-02-25 | 哈尔滨工业大学深圳研究生院 | Intrinsic frequency measuring method for long and thin object |
| CN105136281A (en) * | 2015-10-14 | 2015-12-09 | 基康仪器股份有限公司 | Vibrating wire transducer frequency measurement device and method |
| CN107525849A (en) * | 2017-08-25 | 2017-12-29 | 北京航空航天大学 | A kind of single-input single-output test modal analysis system and method based on fiber grating |
| CN111157099A (en) * | 2020-01-02 | 2020-05-15 | 河海大学常州校区 | Distributed optical fiber sensor vibration signal classification method and identification classification system |
| CN111325095A (en) * | 2020-01-19 | 2020-06-23 | 西安科技大学 | Intelligent equipment health state detection method and system based on sound wave signals |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10271233B2 (en) * | 2013-03-15 | 2019-04-23 | DGS Global Systems, Inc. | Systems, methods, and devices for automatic signal detection with temporal feature extraction within a spectrum |
| JP6707670B2 (en) * | 2016-06-08 | 2020-06-10 | ザイゴ コーポレーションＺｙｇｏ Ｃｏｒｐｏｒａｔｉｏｎ | Precision positioning system using tunable laser |
-
2021
- 2021-11-23 CN CN202111394732.XA patent/CN114121305B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104374468A (en) * | 2014-10-24 | 2015-02-25 | 哈尔滨工业大学深圳研究生院 | Intrinsic frequency measuring method for long and thin object |
| CN105136281A (en) * | 2015-10-14 | 2015-12-09 | 基康仪器股份有限公司 | Vibrating wire transducer frequency measurement device and method |
| CN107525849A (en) * | 2017-08-25 | 2017-12-29 | 北京航空航天大学 | A kind of single-input single-output test modal analysis system and method based on fiber grating |
| CN111157099A (en) * | 2020-01-02 | 2020-05-15 | 河海大学常州校区 | Distributed optical fiber sensor vibration signal classification method and identification classification system |
| CN111325095A (en) * | 2020-01-19 | 2020-06-23 | 西安科技大学 | Intelligent equipment health state detection method and system based on sound wave signals |
Non-Patent Citations (2)
| Title |
|---|
| Taxonomy and Challenges of Out-of-Band Signal Injection Attacks and Defenses;Ilias Giechaskiel等;《IEEE Communications Surveys & Tutorials》;全文 * |
| 正弦动态压力校准中扫频速率对校准结果的影响研究;邰寒松;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114121305A (en) | 2022-03-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | A high-efficiency multiple events discrimination method in optical fiber perimeter security system | |
| CN105261136B (en) | The method and device of weather interference is shielded in a kind of fiber-optic monitoring warning system | |
| CN103116957B (en) | A kind of method of optical fiber perimeter security and protection system shielding climate impact | |
| Chen et al. | A real-time distributed deep learning approach for intelligent event recognition in long distance pipeline monitoring with DOFS | |
| CN114121305B (en) | Sensor safety detection method and system based on sweep frequency technology | |
| CN106404148A (en) | Portable micro vibration testing device and method | |
| CN102628960B (en) | Velocity and acceleration two-parameter digital geophone | |
| Sun et al. | Man-made threat event recognition based on distributed optical fiber vibration sensing and SE-WaveNet | |
| CN107153139A (en) | A kind of optical fibre gyro detection circuit board noise characteristic method of testing | |
| CN114510960A (en) | Method for recognizing distributed optical fiber sensor system mode | |
| CN104713575A (en) | Method for testing frequency characteristic of closed loop fiber optic gyroscope | |
| CN106767960A (en) | Fibre Optical Sensor Dynamic Signal demodulation method based on spectrum simulation Yu difference algorithm | |
| CN115265750A (en) | Optical fiber distributed acoustic wave sensing system and method | |
| CN101782562A (en) | Greenhouse gas on-line monitoring system and monitoring method thereof | |
| Jiang et al. | Identification Technique for the Intrusion of Airport Enclosure Based on Double Mach-Zehnder Interferometer. | |
| CN106094009B (en) | Reflection wave velocity observation system based on photoelectric sensor and acceleration transducer | |
| Ruzicka et al. | Classification of events violating the safety of physical layers in fiber-optic network infrastructures | |
| CN103985213B (en) | A kind of configurable continuous-flow type optical fiber disturbance alarm detection procedure and detector thereof | |
| CN109682457A (en) | Incident optical fiber recognition methods based on angle detection | |
| CN113776644A (en) | Optical fiber fence intrusion signal simulation device based on Mach-Zehnder interferometer | |
| Wu et al. | Real-time correction for sensor's dynamic error based on DSP | |
| CN108344493A (en) | A kind of measuring apparatus and measurement method of phase generated carrier optical fiber sensing system signal demodulation factor | |
| CN219038172U (en) | Integrated cable and fence detection system | |
| Chen | Condition monitoring system of electromechanical equipment based on fiber Bragg grating sensors and artificial neural network | |
| CN103868747A (en) | Structural damage recognition system and recognition method based on time series symbolization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |