Hypothesis testing-based laser radar deception detection method and system, and storage medium

Abstract

The invention discloses a hypothesis testing-based laser radar deception detection method and system, and a storage medium. The method comprises the steps of constructing a hypothesis testing variable composed of multiple parameters; estimating a hypothesis test variable value by using a maximum likelihood method; calculating a sample number threshold value according to the speed and / or the acceleration and / or the attack data proportion; calculating an optimal sampling window value according to a relationship between a sample number threshold value and power consumption; comparing and verifying whether the mean value of the sample variables of the laser radar has significant change according to a double-sample hypothesis test method, and if so, judging that the data in the current sampling window is subjected to spoofing attack; and otherwise, judging that the data in the current sampling window is not subjected to spoofing attack. According to the invention, the problem that the existing laser radar deception detection technology does not consider the acceleration and adopts a fixed sample number threshold value, resulting in low detection accuracy and even failure is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of information security, and particularly relates to a lidar spoofing detection method, system and storage medium based on hypothesis testing. Background Art

[0002] The lidar solution is one of the core solutions for current intelligent driving and autonomous driving. A lidar consists of a transmitter and a receiver. When it works, the transmitter emits laser beams outward. After the laser beams encounter an object and are reflected, they are received by the receiver. The lidar can calculate the distance to the measured target through the time difference measurement of the emitted and received laser beams, and plays a key role in the obstacle detection system of autonomous driving.

[0003] However, lidar may become a potential target of attack, thus endangering the safety of the autonomous driving system. In particular, the lidar system may be subject to spoofing attacks. An attacker emits carefully designed sensor signals to the lidar receiver, making it think that there is an obstacle ahead or removing an actually existing obstacle. When the above situation occurs, the vehicle may make wrong decisions such as emergency braking, starting failure or automatic braking failure, which will cause many traffic hazards.

[0004] To solve the problem of detecting whether the lidar is under a spoofing attack, the current solution is to judge whether the lidar is under a spoofing attack by comparing the sample values calculated by the Doppler shift and the time of flight; however, this method has problems: (1) Due to the existence of acceleration, the Doppler spectrum obtained by Fourier transforming the signal data has a certain width, affecting the accuracy. This situation is more obvious when the object speed is small and the acceleration is large. At this time, the above test may judge more normal data as attack data, resulting in the invalidation of the test. (2) The set sample number threshold will seriously reduce the accuracy of judgment in some driving scenarios.

[0005] To solve the problem of low or even invalid test accuracy rate caused by the existing lidar spoofing detection technology not considering acceleration and using a fixed sample number threshold, a lidar spoofing detection method, system and storage medium based on hypothesis testing are proposed. Summary of the Invention

[0006] The present invention provides a lidar spoofing detection method, system and storage medium based on hypothesis testing to at least solve the problem of low or even invalid test accuracy rate caused by the existing lidar spoofing detection technology not considering acceleration and using a fixed sample number threshold.

[0007] According to an embodiment of the present invention, a lidar spoofing detection method based on hypothesis testing is provided, including the steps:

[0008] Construct a hypothesis testing variable composed of multiple parameters;

[0009] Estimate the value of the hypothesis testing variable using the maximum likelihood method;

[0010] Calculate the sample quantity threshold according to the speed and / or acceleration and / or proportion of attack data;

[0011] Calculate the optimal sampling window value according to the relationship between the sample quantity threshold and power consumption;

[0012] Compare and verify whether there is a significant change in the mean value of the sample variables of the lidar according to the two-sample hypothesis testing method. If so, it is determined that the data in the current sampling window is under spoofing attack; otherwise, it is determined that the data in the current sampling window is not under spoofing attack.

[0013] Optionally, the constructing a hypothesis testing variable composed of multiple parameters includes the steps of:

[0014] Obtain the moving speed and acceleration of the moving body at each moment;

[0015] Calculate the signal-to-noise ratio at each moment according to the influence of environmental light intensity on the signal;

[0016] Construct a hypothesis testing variable according to the moving speed, acceleration and signal-to-noise ratio at each moment.

[0017] Optionally, the estimating the value of the hypothesis testing variable using the maximum likelihood method includes the steps of:

[0018] Estimate the moving speed, acceleration and signal-to-noise ratio in the Doppler frequency shift from the Fourier transform using the maximum likelihood method;

[0019] Calculate the moving speed, acceleration and signal-to-noise ratio under the flight time according to the time difference between the laser beam emission and reception of the lidar;

[0020] Calculate the sample mean and variance of the moving speed, acceleration and signal-to-noise ratio under the Doppler frequency shift and flight time.

[0021] Optionally, the calculating the sample quantity threshold according to the speed and / or acceleration and / or proportion of attack data includes the steps of:

[0022] Calculate the allowable range of the first type of error and the allowable range of the second type of error according to different operating environments and operating modes of the moving body; the first type of error is the error of identifying normal data as spoofing data, and the second type of error is the error of failing to identify spoofing data;

[0023] Calculate the non-central parameter according to the speed and / or acceleration and / or proportion of attack data;

[0024] Calculate the sample quantity threshold value that meets the allowable range of type I error and the allowable range of type II error according to the non-central parameter.

[0025] Further optionally, the calculating the non-central parameter according to the speed and / or acceleration and / or proportion of attack data includes the steps of:

[0026] Calculate the speed influence value according to the average speed and / or speed change value and / or maximum speed value;

[0027] Calculate the acceleration influence value according to the average acceleration and / or acceleration change value and / or maximum acceleration value;

[0028] Calculate the proportion of attack data influence value according to the influence of the proportion of attack data on the power test;

[0029] Calculate the non-central parameter according to the speed influence value and / or acceleration influence value and / or proportion of attack data influence value.

[0030] Optionally, the calculating the sample quantity threshold value that meets the allowable range of type I error and the allowable range of type II error according to the non-central parameter includes the steps of:

[0031] Substitute the non-central parameter into the alternative hypothesis of different attack scenarios;

[0032] Calculate the expressions of type I error and type II error under the alternative hypothesis of different attack scenarios;

[0033] Calculate the functional relationship between the sample quantity threshold value and the test value according to the expressions of type I error and type II error under the alternative hypothesis of different attack scenarios;

[0034] Calculate the sample quantity threshold value that meets the allowable range of type I error and the allowable range of type II error according to the functional relationship between the sample quantity threshold value and the test value.

[0035] Optionally, the calculating the optimal sampling window value according to the relationship between the sample quantity threshold value and the power consumption includes the steps of:

[0036] Calculate the functional relationship between the sample quantity threshold value and the power consumption according to the power consumption corresponding to different sample quantity threshold values;

[0037] Calculate the functional relationship between the sample quantity threshold value and the power according to the functional relationship between the sample quantity threshold value and the test value and the functional relationship between the sample quantity threshold value and the power consumption

[0038] Calculate the sample quantity threshold value corresponding to the maximum power according to the functional relationship between the sample quantity threshold value and the power and use this as the optimal sampling window value.

[0039] Optionally, comparing and verifying whether the mean value of the sample variables of the lidar according to the two-sample hypothesis testing method includes the steps of:

[0040] Construct the statistic of the two-sample Hotelling T 2 testing method;

[0041] Calculate the quantile with the significance level within the allowable range of the first type of error according to the probability distribution of the statistic;

[0042] Judge whether the mean value of the sample variables of the lidar has a significant change according to whether the statistic is greater than the quantile with the significance level within the allowable range of the first type of error.

[0043] According to another embodiment of the present invention, there is provided a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program causes a computer to execute the above method.

[0044] According to another embodiment of the present invention, there is provided a lidar spoofing detection system based on hypothesis testing, including:

[0045] A processor;

[0046] A memory;

[0047] And

[0048] One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the signal processing unit, and the programs cause a computer to execute the above method.

[0049] The advantages of the lidar spoofing detection device, method, system and storage medium based on hypothesis testing of the present invention are:

[0050] (1) Construct hypothesis testing variables according to the moving speed, acceleration and signal-to-noise ratio at each moment. Compared with the traditional two-sample testing technical solution that only considers the speed factor, it can not only effectively solve the problem of test failure caused by large errors in Doppler frequency shift data, but also take into account the influence of environmental factors during the driving of the moving body, effectively improving the test accuracy.

[0051] (2) Calculate the non-central parameter according to the speed and / or acceleration and / or the proportion of attack data, and calculate the sample quantity threshold value that meets the allowable range of the first type of error and the allowable range of the second type of error. Compared with the traditional technical solution of obtaining a fixed sample quantity threshold value through power testing, it can effectively ensure that the error rate of the test is within an acceptable range in various driving environments, improving the environmental adaptability of the lidar spoofing detection system.

[0052] (3) Calculate the optimal sampling window value according to the relationship between the sample quantity threshold and power consumption. Compared with the traditional deception detection technical solution, it can effectively reduce the detection power consumption on the basis of ensuring the inspection accuracy rate.

[0053] (4) Compare and verify whether there is a significant change in the mean value of the sample variables of the lidar according to the two-sample Hotelling T 2 hypothesis testing method. Compared with the traditional deception detection technical solution, it can meet the requirement of detecting multiple variables simultaneously and effectively improve the inspection accuracy rate. Description of the Drawings

[0054] Figure 1 is the flow chart of the lidar deception detection method based on hypothesis testing according to the embodiment of the present invention;

[0055] Figure 2 is the flow chart of step S01 of the lidar deception detection method based on hypothesis testing according to the embodiment of the present invention;

[0056] Figure 3 is the flow chart of step S02 of the lidar deception detection method based on hypothesis testing according to the embodiment of the present invention;

[0057] Figure 4 is the flow chart of step S03 of the lidar deception detection method based on hypothesis testing according to the embodiment of the present invention;

[0058] Figure 5 is the flow chart of sub-step S032 of the lidar deception detection method based on hypothesis testing according to the embodiment of the present invention;

[0059] Figure 6 is the flow chart of sub-step S033 of the lidar deception detection method based on hypothesis testing according to the embodiment of the present invention;

[0060] Figure 7 is the flow chart of step S04 of the lidar deception detection method based on hypothesis testing according to the embodiment of the present invention;

[0061] Figure 8 is the flow chart of step S05 of the lidar deception detection method based on hypothesis testing according to the embodiment of the present invention;

[0062] Figure 9 is the structural schematic diagram of the lidar deception detection system based on hypothesis testing according to the embodiment of the present invention; Detailed Embodiment

[0063] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the invention, but do not limit the invention in any form. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several changes and improvements can still be made. These all belong to the protection scope of the present invention.

[0064] The lidar described in this embodiment is deployed on a moving body, including but not limited to automobiles, ships, industrial machinery, etc.

[0065] According to an embodiment of the present invention, a lidar spoofing detection method based on hypothesis testing is provided, and the flowchart is as Figure 1 shown, including:

[0066] Step S01, construct a hypothesis testing variable composed of multiple parameters;

[0067] Step S02, estimate the value of the hypothesis testing variable by the method of maximum likelihood;

[0068] Step S03, calculate the sample number threshold according to the speed and / or acceleration and / or attack data ratio;

[0069] Step S04, calculate the optimal sampling window value according to the relationship between the sample number threshold and the power consumption;

[0070] Step S05, compare and verify whether the mean value of the sample variable of the lidar has a significant change according to the two-sample hypothesis testing method. If so, it is determined that the data in the current sampling window is under spoofing attack; otherwise, it is determined that the data in the current sampling window is not under spoofing attack.

[0071] In an exemplary embodiment, the step S01, construct a hypothesis testing variable composed of multiple parameters, and the flowchart is as Figure 2 shown, including the steps:

[0072] Step S011, obtain the moving speed and acceleration of the moving body at each moment;

[0073] Step S012, calculate the signal-to-noise ratio at each moment according to the influence of the environmental light intensity on the signal;

[0074] Step S013, construct a hypothesis testing variable according to the moving speed, acceleration and signal-to-noise ratio at each moment.

[0075] This embodiment is a lidar system mounted on a vehicle, and the lidar is used for obstacle detection. Select the relative speed v and acceleration a between the lidar and the obstacle, and calculate the signal-to-noise ratio n at each moment according to the influence of the environmental light intensity on the signal, and use this as the variable of the hypothesis test. The speed unit is m / s, and the acceleration unit is m / s2 .

[0076] Construct the hypothesis test variable x=[v,a,n] based on the moving speed v, acceleration a and signal-to-noise ratio n.

[0077] In an exemplary embodiment, the step S02 uses the maximum likelihood method to estimate the hypothesis test variable value, and the flow chart is as follows: Figure 3 As shown, the steps include:

[0078] Step S021: using the maximum likelihood method to estimate the moving speed, acceleration and signal-to-noise ratio in the Doppler frequency shift from the Fourier transform;

[0079] Step S022: Calculate the moving speed, acceleration, and signal-to-noise ratio under the time of flight based on the time difference between the laser radar transmitting and receiving the laser beam;

[0080] Step S023: Calculate the sample mean and variance of the moving speed, acceleration, and signal-to-noise ratio under the Doppler frequency shift and flight time.

[0081] In this embodiment, the maximum likelihood method is used to estimate the moving speed v in the Doppler frequency shift from the Fourier transform. Dop , acceleration a Dop and signal-to-noise ratio n Dop , that is, the hypothesis test variable x under Doppler frequency shift Dop =[v Dop ,a Dop ,n Dop ];

[0082] The moving speed v under the flight time is calculated based on the time difference between the laser radar transmitting and receiving the laser beam ToF , acceleration a ToF And calculate the signal-to-noise ratio n under the flight time ToF , that is, the hypothesis test variable x under flight time ToF =[v ToF ,a ToF ,n ToF ];

[0083] Obtain the moving speed, acceleration and signal-to-noise ratio samples under the Doppler frequency shift and flight time at multiple moments in the preset time period, calculate the sample mean and variance, and hypothesis test variable x under the Doppler frequency shift Dop The sample mean is denoted as μ1 and the variance is denoted as s1 2 , hypothesis test variable x under flight time ToF The mean is denoted as μ2 and the variance is denoted as s2 2 .

[0084] In an exemplary embodiment, the step S03, calculating a sample quantity threshold according to speed and / or acceleration and / or the proportion of attack data, has a flowchart as Figure 4 shown, including steps:

[0085] Step S031, calculating a first type of error tolerance range and a second type of error tolerance range according to different operating environments and operating modes of the moving body;

[0086] Step S032, calculating a non-central parameter according to speed and / or acceleration and / or the proportion of attack data;

[0087] Step S033, calculating a sample quantity threshold that conforms to the first type of error tolerance range and the second type of error tolerance range according to the non-central parameter.

[0088] Optionally, the calculating of the first type of error tolerance range and the second type of error tolerance range according to different operating environments and operating modes of the moving body is to adjust the preset first type of error tolerance range and the second type of error tolerance range according to the different influences of different operating environments and operating modes of the moving body on the error tolerance range to obtain the first type of error tolerance range and the second type of error tolerance range.

[0089] Optionally, the step S032, calculating a non-central parameter according to speed and / or acceleration and / or the proportion of attack data, has a flowchart as Figure 5 shown, including steps:

[0090] Step S0321, calculating a speed influence value according to the average speed and / or the speed change value and / or the maximum speed;

[0091] Step S0322, calculating an acceleration influence value according to the average acceleration and / or the acceleration change value and / or the maximum acceleration;

[0092] Step S0323, calculating an attack data proportion influence value according to the influence of the proportion of attack data on the power test;

[0093] Step S0324, calculating a non-central parameter according to the speed influence value and / or the acceleration influence value and / or the attack data proportion influence value.

[0094] In this embodiment, calculating the speed influence value according to the average speed and / or the speed change value and / or the maximum speed means obtaining a positive correlation between the average speed and / or the speed change value and / or the maximum speed and the influence degree of the non-central parameter based on the influence of speed on the non-central parameter (the greater the average speed and / or the speed change value and / or the maximum speed, the greater the influence degree of the non-central parameter, that is, the greater the speed influence value), and calculating the speed influence value according to the positive correlation between the average speed and / or the speed change value and / or the maximum speed and the speed influence value. The speed influence value is represented by the variable l;

[0095] Calculating the acceleration influence value according to the average acceleration and / or the acceleration change value and / or the maximum acceleration means obtaining a positive correlation between the average acceleration and / or the acceleration change value and / or the maximum acceleration and the influence degree of the non-central parameter based on the influence of acceleration on the non-central parameter (the greater the average acceleration and / or the acceleration change value and / or the maximum acceleration, the greater the influence degree of the non-central parameter, that is, the greater the acceleration influence value), and calculating the acceleration influence value according to the positive correlation between the average acceleration and / or the acceleration change value and / or the maximum acceleration and the acceleration influence value. The acceleration influence value is represented by the variable w;

[0096] Calculating the attack data ratio influence value according to the influence of the attack data ratio on the efficacy test means obtaining a positive correlation between the attack data ratio and the attack data ratio influence value based on the influence of the attack data ratio on the efficacy test (the greater the attack data ratio, the greater the influence on the efficacy test, that is, the greater the attack data ratio influence value), and calculating the attack data ratio influence value according to the positive correlation between the attack data ratio and the attack data ratio influence value. The attack data ratio influence value is represented by the variable r;

[0097] Calculating the non-central parameter according to the speed influence value and / or the acceleration influence value and / or the attack data ratio influence value means calculating the non-central parameter correction value according to the positive correlation between the speed influence value and / or the acceleration influence value and / or the attack data ratio influence value and the non-central parameter correction value, and calculating the non-central parameter according to the preset initial non-central parameter and the non-central parameter correction value in different scenarios. The non-central parameter correction value is represented by the variable s.

[0098] Examples A1 to A7 are used to represent different implementation manners of calculating the non-central parameter correction value.

[0099] Example A1: Calculate the non-central parameter correction value according to the speed influence value.

[0100] Specifically, the speed influence value, denoted as l, is calculated based on the positive correlation relationship between the average speed and / or the speed change value and / or the maximum speed and the speed influence value; the non - central parameter correction value s is calculated based on the positive correlation relationship between the speed influence value l and the non - central parameter correction value. In a preferred embodiment, the non - central parameter correction value s = o1·l o2 + o3, where o1, o2 (o2>0), o3 are calculation coefficients obtained through prior training. In this embodiment, the average speed of the moving object within a certain period is obtained as 5 m / s, normalized to d = 0.6 according to the preset speed threshold, and the speed influence value l = k1·d k2 + k3 = 1×0.6 + 0 = 0.6 (k1, k2, k3 are calculation coefficients obtained through prior training, where k1 = 1, k2 = 1, k3 = 0 in this embodiment), and the calculation coefficients o1 = 1, o2 = 1, o3 = 0 obtained through prior training, then the non - central parameter correction value s = o1·l o2 + o3 = 1×0.6 + 0 = 0.6.

[0101] Example A2: Calculate the non - central parameter correction value according to the acceleration influence value.

[0102] Specifically, the acceleration influence value, denoted as w, is calculated based on the positive correlation relationship between the average acceleration and / or the acceleration change value and / or the maximum acceleration and the acceleration influence value; the non - central parameter correction value s is calculated based on the positive correlation relationship between the acceleration influence value w and the non - central parameter correction value. In a preferred embodiment, the non - central parameter correction value s = o4·w o5 [[ID=?]] 2 According to the preset acceleration threshold, it is normalized to h = 0.7, and the acceleration influence value w = k4·h k5 + k6 = 1×0.7 + 0 = 0.7 (k4, k5, k6 are calculation coefficients obtained through prior training, where k4 = 1, k5 = 1, k6 = 0 in this embodiment), and the calculation coefficients o4 = 1, o5 = 1, o6 = 0 obtained through prior training, then the non - central parameter correction value s = o4·w o5 + o6 = 1×0.7 + 0 = 0.7.

[0103] Example A3: Calculate the non - central parameter correction value according to the proportion influence value of the attack data.

[0104] Note: There seems to be a formatting issue in the original text where the unit of acceleration in line 14 is incomplete. It should probably be "1.5m / s²" for a complete and correct physical quantity representation.Specifically, calculate the attack data proportion impact value, denoted as r, according to the positive correlation between the attack data proportion and the attack data proportion impact value; calculate the non-central parameter correction value s according to the positive correlation between the attack data proportion impact value r and the non-central parameter correction value. In a preferred embodiment, calculate the non-central parameter correction value s = o7·r o8 + o9, where o7, o8 (o8>0), o9 are calculation coefficients obtained through prior training. In this embodiment, detect the attack data proportion data of the moving body within a period of time, and normalize it according to the preset attack proportion threshold to obtain the attack data proportion j = 0.8. Calculate the attack data proportion impact value r = k7·j k8 + k9 = 1×0.8 + 0 = 0.8 (k7, k8, k9 are calculation coefficients obtained through prior training. In this embodiment, k7 = 1, k8 = 1, k9 = 0), and the calculation coefficient o7 = 1, o8 = 1, o9 = 0 obtained through prior training. Calculate the non-central parameter correction value s = o7·r o8 + o9 = 1×0.8 + 0 = 0.8.

[0105] Example A4: Calculate the non-central parameter correction value according to the speed impact value and the acceleration impact value.

[0106] Specifically, calculate the speed impact value, denoted as l, according to the positive correlation between the average speed and / or the speed change value and / or the maximum speed and the speed impact value; calculate the acceleration impact value, denoted as w, according to the positive correlation between the average acceleration and / or the acceleration change value and / or the maximum acceleration and the acceleration impact value; calculate the non-central parameter correction value s according to the positive correlation between the speed impact value l and the acceleration impact value w and the non-central parameter correction value. In a preferred embodiment, calculate the non-central parameter correction value s = o10·l o11 + o12·w o13 , where o10, o11 (o11>0), o12, o13 (o13>0) are calculation coefficients obtained through prior training. In this embodiment, obtain the average speed of the moving body within a period of time as 5 m / s, normalize it according to the preset speed threshold to d = 0.6, and calculate the speed impact value l = k1·d k2 + k3 = 1×0.6 + 0 = 0.6 (k1, k2, k3 are calculation coefficients obtained through prior training. In this embodiment, k1 = 1, k2 = 1, k3 = 0); obtain the average acceleration of the moving body within a period of time as 1.5 m / s 2 , normalize it according to the preset acceleration threshold to h = 0.7, and calculate the acceleration impact value w = k4·h k5+k6 = 1×0.7 + 0 = 0.7 (k4, k5, k6 are calculation coefficients obtained through prior training. In this embodiment, k4 = 1, k5 = 1, k6 = 0); the calculation coefficients o10 = 0.7, o11 = 1, o12 = 0.3, o13 = 1 obtained through prior training, calculate the non - central parameter correction value s = o10·l o11 +o12·w o13 = 0.7×0.6 + 0.3×0.7 = 0.63. In another preferred embodiment, calculate the non - central parameter correction value s = o14·l o15 ·w o16 +o17, where o14, o15 (o15>0), o16 (o16>0), o17 are calculation coefficients obtained through prior training. In this embodiment, obtain the average speed of the moving object within a period of time as 5m / s, normalize it to d = 0.6 according to the preset speed threshold, and calculate the speed influence value l = k1·d according to the positive correlation between the average speed and the speed influence value k2 +k3 = 1×0.6 + 0 = 0.6 (k1, k2, k3 are calculation coefficients obtained through prior training. In this embodiment, k1 = 1, k2 = 1, k3 = 0); obtain the average acceleration of the moving object within a period of time as 1.5m / s 2 , normalize it to h = 0.7 according to the preset acceleration threshold, and calculate the acceleration influence value w = k4·h according to its positive correlation with the acceleration influence value k5 +k6 = 1×0.7 + 0 = 0.7 (k4, k5, k6 are calculation coefficients obtained through prior training. In this embodiment, k4 = 1, k5 = 1, k6 = 0); the calculation coefficients o14 = 1.5, o15 = 1, o16 = 1, o17 = 0 obtained through prior training, calculate the non - central parameter correction value s = o14·l o15 ·w o16 +o17 = 1.5×0.6×0.7 + 0 = 0.63.

[0107] Example A5: Calculate the non - central parameter correction value according to the speed influence value and the attack data ratio influence value.

[0108] Specifically, calculate the speed influence value l according to the positive correlation between the average speed and / or the speed change value and / or the maximum speed and the speed influence value; calculate the attack data ratio influence value r according to the positive correlation between the attack data ratio and the attack data ratio influence value; calculate the non - central parameter correction value s according to the positive correlation between the speed influence value l and the attack data ratio influence value r and the non - central parameter correction value. In a preferred embodiment, calculate the non - central parameter correction value s = o18·l o19 +o20·r o21, where o18, o19 (o19>0), o20, and o21 (o21>0) are calculation coefficients obtained through prior training. In this embodiment, the average speed of the moving body over a period of time is obtained as 5 m / s, normalized to d = 0.6 according to a preset speed threshold, and the speed influence value l = k1·d is calculated based on the positive correlation between the average speed and the speed influence value k2 +k3 = 1×0.6 + 0 = 0.6 (k1, k2, k3 are calculation coefficients obtained through prior training. In this embodiment, k1 = 1, k2 = 1, k3 = 0); the proportion data of the attack data of the moving body over a period of time is detected, normalized according to a preset attack proportion threshold to obtain the attack data proportion j = 0.8, and the attack data proportion influence value r = k7·j is calculated based on its positive correlation with the attack data proportion influence value k8 +k9 = 1×0.8 + 0 = 0.8 (k7, k8, k9 are calculation coefficients obtained through prior training. In this embodiment, k7 = 1, k8 = 1, k9 = 0); the calculation coefficients obtained through prior training are o18 = 0.8, o19 = 1, o20 = 0.2, o21 = 1, and the non - central parameter correction value s = o18·l o19 +o20·r o21 = 0.8×0.6 + 0.2×0.8 = 0.64. In another preferred embodiment, the non - central parameter correction value s = o22·l o23 ·r o23 +o25, where o22, o23 (o23>0), o24 (o24>0), and o25 are calculation coefficients obtained through prior training. In this embodiment, the average speed of the moving body over a period of time is obtained as 5 m / s, normalized to d = 0.6 according to a preset speed threshold, and the speed influence value l = k1·d k2 +k3 = 1×0.6 + 0 = 0.6 (k1, k2, k3 are calculation coefficients obtained through prior training. In this embodiment, k1 = 1, k2 = 1, k3 = 0); the proportion data of the attack data of the moving body over a period of time is detected, normalized according to a preset attack proportion threshold to obtain the attack data proportion j = 0.8, and the attack data proportion influence value r = k7·j is calculated based on its positive correlation with the attack data proportion influence value k8 +k9 = 1×0.8 + 0 = 0.8 (k7, k8, k9 are calculation coefficients obtained through prior training. In this embodiment, k7 = 1, k8 = 1, k9 = 0); the calculation coefficients obtained through prior training are o22 = 1.35, o23 = 1, o24 = 1, o25 = 0, and the non - central parameter correction value s = o22·l o23 ·r o23 +o25 = 1.35×0.6×0.8 + 0 = 0.648.

[0109] Embodiment A6: Calculate the non - central parameter correction value according to the acceleration influence value and the attack data ratio influence value.

[0110] Specifically, calculate the acceleration influence value, denoted as w, according to the positive correlation between the average acceleration and / or the acceleration change value and / or the maximum acceleration and the acceleration influence value; calculate the attack data ratio influence value, denoted as r, according to the positive correlation between the attack data ratio and the attack data ratio influence value; calculate the non - central parameter correction value s according to the positive correlation between the acceleration influence value w and the attack data ratio influence value r and the non - central parameter correction value. In a preferred embodiment, calculate the non - central parameter correction value s = o26·w o27 +o28·r o29 , where o26, o27 (o27>0), o28, o29 (o29>0) are calculation coefficients obtained through prior training. In this embodiment, the average acceleration of the moving body within a period of time is obtained as 1.5 m / s 2 , normalized to h = 0.7 according to the preset acceleration threshold, and calculate the acceleration influence value w = k4·h according to its positive correlation with the acceleration influence value k5 +k6 = 1×0.7 + 0 = 0.7 (k4, k5, k6 are calculation coefficients obtained through prior training. In this embodiment, k4 = 1, k5 = 1, k6 = 0); detect the attack data ratio data of the moving body within a period of time, normalize it according to the preset attack ratio threshold to obtain the attack data ratio j = 0.8, and calculate the attack data ratio influence value r = k7·j according to its positive correlation with the attack data ratio influence value k8 +k9 = 1×0.8 + 0 = 0.8 (k7, k8, k9 are calculation coefficients obtained through prior training. In this embodiment, k7 = 1, k8 = 1, k9 = 0); the calculation coefficients obtained through prior training are o26 = 0.7, o27 = 1, o28 = 0.3, o29 = 1, and calculate the non - central parameter correction value s = o26·w o27 +o28·r o29 = 0.7×0.7 + 0.3×0.8 = 0.73. In another preferred embodiment, calculate the non - central parameter correction value s = o30·w o31 ·r o32 , where o30, o31 (o31>0), o32 (o32>0), o33 are calculation coefficients obtained through prior training. In this embodiment, the average acceleration of the moving body within a period of time is obtained as 1.5 m / s 2 , normalized to h = 0.7 according to the preset acceleration threshold, and calculate the acceleration influence value w = k4·h according to its positive correlation with the acceleration influence value k5+k6 = 1×0.7 + 0 = 0.7 (k4, k5, k6 are calculation coefficients obtained through prior training. In this embodiment, k4 = 1, k5 = 1, k6 = 0); Detect the proportion data of the attack data of the moving body within a period of time, and normalize it according to the preset attack proportion threshold to obtain the attack data proportion j = 0.8, and calculate the attack data proportion influence value r = k7·j according to its positive correlation with the attack data proportion influence value k8 +k9 = 1×0.8 + 0 = 0.8 (k7, k8, k9 are calculation coefficients obtained through prior training. In this embodiment, k7 = 1, k8 = 1, k9 = 0); The calculation coefficients o30 = 1.2, o31 = 1, o32 = 1, o33 = 0 obtained through prior training, calculate the non - central parameter correction value s = o30·w o31 ·r o32 +o33 = 1.2×0.7×0.8 + 0 = 0.672.

[0111] Example A7: Calculate the non - central parameter correction value according to the speed influence value, acceleration influence value and attack data proportion influence value.

[0112] Specifically, calculate the speed influence value according to the positive correlation between the average speed and / or speed change value and / or maximum speed and the speed influence value, denoted as l; Calculate the acceleration influence value according to the positive correlation between the average acceleration and / or acceleration change value and / or maximum acceleration and the acceleration influence value, denoted as w; Calculate the attack data proportion influence value according to the positive correlation between the attack data proportion and the attack data proportion influence value, denoted as r; Calculate the non - central parameter correction value s according to the positive correlation between the speed influence value l, acceleration influence value w and attack data proportion influence value r and the non - central parameter correction value. In a preferred embodiment, calculate the non - central parameter correction value s = o34·l o35 +o36·w o37 +o38·r o39 , where o34, o35 (o35 > 0), o36, o37 (o37 > 0), o38, o39 (o39 > 0) are calculation coefficients obtained through prior training. In this embodiment, the average speed of the moving body within a period of time is obtained as 5 m / s, and it is normalized to d = 0.6 according to the preset speed threshold, and the speed influence value l = k1·d is calculated according to the positive correlation between the average speed and the speed influence value k2 +k3 = 1×0.6 + 0 = 0.6 (k1, k2, k3 are calculation coefficients obtained through prior training. In this embodiment, k = 1, k2 = 1, k3 = 0); The average acceleration of the moving body within a period of time is obtained as 1.5 m / s 2 , and it is normalized to h = 0.7 according to the preset acceleration threshold, and the acceleration influence value w = k4·h is calculated according to its positive correlation with the acceleration influence valuek5 +k6 = 1×0.7 + 0 = 0.7 (k4, k5, k6 are calculation coefficients obtained through prior training. In this embodiment, k4 = 1, k5 = 1, k6 = 0); Detect the proportion data of the attack data of the moving body within a period of time, and normalize it according to the preset attack proportion threshold to obtain the attack data proportion j = 0.8. Calculate the attack data proportion influence value r = k7·j according to its positive correlation with the attack data proportion influence value k8 +k9 = 1×0.8 + 0 = 0.8 (k7, k8, k9 are calculation coefficients obtained through prior training. In this embodiment, k7 = 1, k8 = 1, k9 = 0); The calculation coefficients o34 = 0.5, o35 = 1, o26 = 0.3, o27 = 1, o28 = 0.2, o29 = 1 obtained through prior training. Calculate the non - central parameter correction value s = o34·l o35 +o36·w o37 +o38·r o39 = 0.5×0.6 + 0.3×0.7 + 0.2×0.8 = 0.67. In another preferred embodiment, calculate the non - central parameter correction value s = o40·l o41 ·w o42 ·r o43 +o44, where o40, o41 (o41>0), o42 (o42>0), o43 (o43>0), o44 are calculation coefficients obtained through prior training. In this embodiment, obtain the average speed of the moving body within a period of time as 5 m / s, normalize it according to the preset speed threshold to d = 0.6, and calculate the speed influence value l = k1·d according to the positive correlation between the average speed and the speed influence value k2 +k3 = 1×0.6 + 0 = 0.6 (k1, k2, k3 are calculation coefficients obtained through prior training. In this embodiment, k1 = 1, k2 = 1, k3 = 0); Obtain the average acceleration of the moving body within a period of time as 1.5 m / s 2 , normalize it according to the preset acceleration threshold to h = 0.7, and calculate the acceleration influence value w = k4·h according to its positive correlation with the acceleration influence value k5 +k6 = 1×0.7 + 0 = 0.7 (k4, k5, k6 are calculation coefficients obtained through prior training. In this embodiment, k4 = 1, k5 = 1, k6 = 0); Detect the proportion data of the attack data of the moving body within a period of time, and normalize it according to the preset attack proportion threshold to obtain the attack data proportion j = 0.8. Calculate the attack data proportion influence value r = k7·j according to its positive correlation with the attack data proportion influence value k8+k9 = 1×0.8 + 0 = 0.8 (k7, k8, and k9 are calculation coefficients obtained through prior training. In this embodiment, k7 = 1, k8 = 1, and k9 = 0); the calculation coefficients obtained through prior training are o40 = 2, o41 = 1, o42 = 1, o43 = 1, o44 = 0, and the non - central parameter correction value s is calculated as s = o40·l o41 ·w o42 ·r o43 +o44 = 2×0.6×0.7×0.8 + 0 = 0.672.

[0113] The calculation of the non - central parameter based on the preset initial non - central parameter and non - central parameter correction value under different scenarios is to calculate the non - central parameter as the product superposition value or accumulation value of the preset initial non - central parameter and the non - central parameter correction value s obtained by the method described in any one of Embodiments A1 - A7 under different scenarios.

[0114] Optionally, in step S033, calculating the sample number threshold value that meets the first - type error tolerance range and the second - type error tolerance range according to the non - central parameter, the flowchart is as Figure 6 shown, including the steps:

[0115] Step S0331: Substitute the non - central parameter into the alternative hypothesis of different attack scenarios;

[0116] Step S0332: Calculate the expressions of the first - type error and the second - type error under the alternative hypothesis of different attack scenarios;

[0117] Step S0333: Calculate the functional relationship between the sample number threshold value and the test value according to the expressions of the first - type error and the second - type error under the alternative hypothesis of different attack scenarios;

[0118] Step S0334: Calculate the sample number threshold value that meets the first - type error tolerance range and the second - type error tolerance range according to the functional relationship between the sample number threshold value and the test value.

[0119] In this embodiment, the setting of the sample number threshold value should ensure that in any attack scenario, the hypothesis test can correctly reject the null hypothesis with a high probability, and both the first - type and second - type errors are within the acceptable range. Considering the extreme operating scenarios that may lead to serious consequences, according to the attacker and the attack consequences, the spoofing attacks are divided into the following three attack scenarios:

[0120] (1) A stationary attacker inserts a non - existent obstacle in front of the lidar, causing the moving object to brake suddenly.

[0121] (2) A moving attacker inserts a non - existent obstacle in front of the lidar, causing the moving object to brake suddenly.

[0122] (3) A stationary attacker inserts an object moving at the same speed as the lidar in front of the lidar, covering the existing obstacle, resulting in the failure of the moving body to brake.

[0123] In the first attack scenario, a stationary attacker inserts a non-existent object in front of the lidar, causing the moving body to brake suddenly.

[0124] In this case, both the attacker and the forged object are stationary. Therefore, the alternative hypothesis is H 1,a : 2x Dop = x ToF , and the corresponding test statistic is:

[0125]

[0126] where n is the sample size, μ1 is the sample mean of x DoP sample mean, μ2 is the sample mean of x ToF sample variance of s1 2 for x DoP sample variance of s2 2 for x ToF sample variance, and the pooled variance S of the two groups of samples pooled = (s1 2 + s2 2 ) / 2.

[0127] This test statistic follows a non-central F distribution, and the initial non-central parameter is calculated as:

[0128]

[0129] where σ 2 is the mean square error. The first degree of freedom of this F distribution is 2, the second degree of freedom is 5n - 6, and the non-central parameter is n.c.p.1. The non-central parameter is calculated based on the product superposition value or cumulative value of the initial non-central parameter n.c.p.1 and the non-central parameter correction value s obtained by the method described in any one of Examples A1 to A7. The β quantile corresponding to its significance level can be obtained by statistical software.

[0130] The type II error is expressed as:

[0131]

[0132] In the second attack scenario, a moving attacker forges an object in front of the lidar, causing it to brake suddenly. In this scenario, it is assumed that the attacker and the victim vehicle are moving at the same speed, and the attacker wants to forge an object in front of the lidar to cause the victim vehicle to brake urgently.

[0133] In this scenario, the alternative hypothesis is H2,a : x Dop = 0, x ToF ≠ 0, the corresponding test statistic is:

[0134]

[0135] This test statistic follows a non - central F - distribution, and its initial non - central parameter is:

[0136]

[0137] The first degree of freedom of this F - distribution is 2, the second degree of freedom is 2n - 3, and the non - central parameter is n.c.p.2. The non - central parameter is calculated based on the product superposition value or cumulative value of the initial non - central parameter n.c.p.2 and the non - central parameter correction value s obtained by the method described in any one of Examples A1 - A7. Its β - quantile corresponding to the significance level can be obtained by calculation using statistical software.

[0138] The type II error is expressed as:

[0139]

[0140] In the third attack scenario, a stationary attacker forges an object moving at the same speed as the lidar in front of the lidar, resulting in the failure of automatic braking.

[0141] The alternative hypothesis in this scenario is H 3,a : x Dop ≠ 0, x ToF = 0, and the corresponding test statistic is:

[0142]

[0143] This statistic follows a non - central F - distribution, and its initial non - central parameter is

[0144]

[0145] The first degree of freedom of this F - distribution is 2, the second degree of freedom is 2n - 3, and the non - central parameter is n.c.p.3. The non - central parameter is calculated based on the product superposition value or cumulative value of the initial non - central parameter n.c.p.3 and the non - central parameter correction value s obtained by the method described in any one of Examples A1 - A7. Its β - quantile corresponding to the significance level can be obtained by calculation using statistical software.

[0146] The type II error is expressed as:

[0147]

[0148] The functional relationship between the sample size threshold value and the test value is calculated based on the expressions of the first and second type errors under the alternative hypotheses of different attack scenarios, where the sample threshold value is represented by the variable m, the test value is represented by the variable T, and the functional relationship between the sample size threshold value and the test value is recorded as function T = f(m).

[0149] The test value needs to meet the tolerance range of the first and second types of errors. Therefore, according to the expressions of the first and second types of errors under the alternative hypotheses of different attack scenarios and the functional relationship between the sample size threshold and the test value, the inequality related to the sample size threshold is obtained and the minimum sample size threshold that meets the tolerance range of the first and second types of errors is calculated, that is, the sample size threshold that meets the tolerance range of the first and second types of errors.

[0150] In another exemplary embodiment, the step S04 is to calculate the optimal sampling window value according to the relationship between the sample quantity threshold value and the power consumption, as shown in the flowchart. Figure 7 As shown, the steps include:

[0151] Step S041: Calculate the functional relationship between the sample quantity threshold and the power consumption according to the power consumption corresponding to different sample quantity thresholds;

[0152] Step S042: Calculate the functional relationship between the sample quantity threshold value and the power efficiency based on the functional relationship between the sample quantity threshold value and the test value and the functional relationship between the sample quantity threshold value and the power consumption;

[0153] Step S043: Calculate the sample quantity threshold value corresponding to the maximum efficacy according to the functional relationship between the sample quantity threshold value and the efficacy and use it as the optimal sampling window value.

[0154] In this embodiment, a good detector should control both the first and second type errors within a reasonable range, setting the first type error α = 0.05 and the second type error β = 0.05.

[0155] Generate multiple sets of attack data through simulation:

[0156] Consider different driving situations: (1) continuous high-speed driving (speed 33m / s, acceleration 0.5m / s 2 ); (2) High-speed ramp driving (speed 20m / s, acceleration 1.5m / s 2 ), which also applies to deceleration scenarios; (3) encountering a red light on an urban road (speed 11m / s, acceleration 5m / s 2 ).

[0157] Considering that it is difficult for attackers to continuously deceive lidar and they only have the ability to interfere with some signals. Hypothesis testing should ensure good accuracy even when normal data and attack data coexist. Therefore, the proportion of spoofing signals is set to 20% to 50%.

[0158] Considering the influence of different environmental illuminations on lidar, the signal-to-noise ratio is set to -60 dB to -30 dB.

[0159] The corresponding number of samples is 38, and the test statistic is 6.3. According to different attack scenarios, power analysis is performed on the simulation data in the above driving conditions.

[0160] Under the simulated attack data, the functional relationship between the sample number threshold and power consumption is fitted and calculated according to the power consumption corresponding to different sample number thresholds, denoted as G = g(m);

[0161] According to the positive correlation between the functional relationship T = f(m) between the sample number threshold and the test value and the power, and the negative correlation between the functional relationship G = g(m) between the sample number threshold and the power consumption and the power, calculate the functional relationship between the sample number threshold and the power. The power value is represented by the variable y, and the functional relationship between the sample number threshold and the power is expressed as y = h(m) = e1·f(m) e2 +e3·g(m) e4 +e5 or y = h(m) = e6·f(m) e7 ·g(m) e8 +e9, where e1, e2 (e1·e2 > 0), e3, e4 (e3·e4 < 0), e5, e6, e7 (e6·e7 > 0), e8 (e7·e8 < 0), e9 are calculation coefficients obtained through training.

[0162] Calculate the maximum value of the function y = h(m) between the sample number threshold and the power, denoted as Y. The sample number threshold corresponding to the maximum power value Y is the optimal sample number threshold, and the optimal sample number threshold is used as the optimal sampling window value.

[0163] In an exemplary embodiment, in step S05, the means of the sample variables of the lidar are compared and verified according to the two-sample hypothesis testing method. The flowchart is as Figure 8 shown, including steps:

[0164] Step S051: Construct the statistic of the two-sample HotellingT 2 test method;

[0165] Step S052: Calculate the quantile with a significance level within the allowable range of the first type of error according to the probability distribution of the statistic;

[0166] Step S053: Determine whether the mean of the sample variables of the lidar has changed significantly by judging whether the statistic is greater than the quantile at the allowable range of the first type of error for the significance level.

[0167] In this embodiment, the power tests are sequentially performed on the three attack scenarios in the above embodiments, and the sample sizes required when both the first type and the second type of errors are within the acceptable range are calculated.

[0168] Since the distribution under the alternative hypothesis is related to the non-central parameter, the minimum sample size cannot be directly obtained from the power test. It is necessary to further consider typical real-world scenarios, such as common speed, acceleration range, proportion of attack data, signal-to-noise ratio, etc., generate simulation data, calculate the corresponding non-central parameter, and then obtain the minimum sample size required for the corresponding power level of 1-β from the power test. The threshold of the sample size is determined by the maximum value determined by the above power test.

[0169] Construct the two-sample Hotelling T 2 Statistic of the test method:

[0170]

[0171] It follows an F-distribution, Its first degree of freedom is 2, and the second degree of freedom is 2n - 3. Its 1-α quantile corresponding to the significance level of α Can be obtained by looking up the table or calculating with statistical software. It can be proved that the Hotelling T 2 Test is an unbiased uniformly most powerful test. When It is considered that the sample mean has changed significantly, and the H0 hypothesis is rejected, indicating that a spoofing attack may have occurred.

[0172] The first type of error is expressed as:

[0173]

[0174] Determine whether the mean of the sample variables of the lidar has changed significantly by judging whether the statistic is greater than the quantile at the allowable range of the first type of error for the significance level.

[0175] In a preferred embodiment, assume that the sample size threshold is 40. First, record the Doppler shift velocity, acceleration, signal-to-noise ratio, and the time-of-flight velocity, acceleration, signal-to-noise ratio, where the proportion of malicious data is 30%. The test statistic at the quantile of 0.05 is Since Therefore, the H0 hypothesis is rejected, and it is considered that the sample mean has changed significantly, and there is a spoofing attack.

[0176] According to another embodiment of the present invention, there is also provided a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program causes a computer to execute the lidar spoofing detection method based on hypothesis testing described in the above embodiments.

[0177] According to another embodiment of the present invention, there is also provided a lidar spoofing detection system based on hypothesis testing, the structural schematic diagram is as Figure 9 shown, including:

[0178] A processor;

[0179] A memory;

[0180] And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs cause a computer to execute the lidar spoofing detection method based on hypothesis testing described in the above embodiments.

[0181] The method according to the present invention described above can be implemented in hardware, firmware, or be implemented as software or computer code that can be stored in a recording medium (such as a CDROM, RAM, floppy disk, hard disk, or magneto-optical disk), or be implemented as computer code originally stored in a remote recording medium or a non-transitory machine-readable medium and to be downloaded through a network and stored in a local recording medium, so that the method described herein can be stored on such a software process on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component (such as a RAM, a ROM, a flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the process shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the process shown herein.

[0182] Of course, those of ordinary skill in the art in the technical field should recognize that the above embodiments are only used to illustrate the present invention and are not intended as a limitation of the present invention. As long as it is within the scope of the present invention, changes and variations of the above embodiments will fall within the protection scope of the present invention.

Claims

1. A lidar spoofing detection method based on hypothesis testing, characterized in that Including: Construct a hypothesis testing variable composed of multiple parameters; Estimate the value of the hypothesis testing variable using the maximum likelihood method; Calculate the sample quantity threshold according to the speed and / or acceleration and / or proportion of attack data; Calculate the optimal sampling window value according to the relationship between the sample quantity threshold and power consumption; According to the two-sample hypothesis testing method, compare and verify whether the mean value of the sample variables of the lidar has a significant change. If so, it is determined that the data in the current sampling window is under spoofing attack; Otherwise, it is determined that the data in the current sampling window is not under spoofing attack.

2. The lidar spoofing detection method based on hypothesis testing according to claim 1, wherein The constructing of the hypothesis testing variable composed of multiple parameters includes the steps of: Obtain the moving speed and acceleration of the moving body at each moment; Calculate the signal-to-noise ratio at each moment according to the influence of environmental light intensity on the signal; Construct a hypothesis testing variable according to the moving speed, acceleration and signal-to-noise ratio at each moment.

3. The method for lidar spoofing detection based on hypothesis testing according to claim 2, characterized in that, The estimating of the value of the hypothesis testing variable using the maximum likelihood method includes the steps of: Estimate the moving speed, acceleration and signal-to-noise ratio in the Doppler frequency shift from the Fourier transform using the maximum likelihood method; Calculate the moving speed, acceleration and signal-to-noise ratio under the flight time according to the time difference between the laser radar transmitting and receiving laser beams; Calculate the sample mean and variance of the moving speed, acceleration and signal-to-noise ratio under the Doppler frequency shift and flight time.

4. The lidar spoofing detection method based on hypothesis testing according to claim 3, wherein, The calculating of the sample quantity threshold according to the speed and / or acceleration and / or proportion of attack data includes the steps of: Calculate the allowable range of the first type of error and the allowable range of the second type of error according to different operating environments and operating modes of the moving body; The first type of error is the error of identifying normal data as spoofing data, and the second type of error is the error of not identifying spoofing data; Calculate the non-central parameter according to the speed and / or acceleration and / or proportion of attack data; Calculate the sample quantity threshold that meets the allowable range of the first type of error and the allowable range of the second type of error according to the non-central parameter.

5. The method for lidar spoofing detection based on hypothesis testing according to claim 4, wherein The calculating of the non-central parameter according to the speed and / or acceleration and / or proportion of attack data includes the steps of: Calculate the speed influence value according to the average speed and / or speed change value and / or maximum speed; Calculate the acceleration influence value according to the average acceleration and / or acceleration change value and / or maximum acceleration; Calculate the influence value of the proportion of attack data according to the influence of the proportion of attack data on the power test; Calculate the non-central parameter according to the speed influence value and / or acceleration influence value and / or influence value of the proportion of attack data.

6. The method for lidar spoofing detection based on hypothesis testing according to claim 4, wherein The calculating of the sample quantity threshold that meets the allowable range of the first type of error and the allowable range of the second type of error according to the non-central parameter includes the steps of: Substitute the non-central parameter into the alternative hypothesis of different attack scenarios; Calculate the expressions of the first type of error and the second type of error under the alternative hypothesis of different attack scenarios; Calculate the functional relationship between the sample quantity threshold and the test value according to the expressions of the first type of error and the second type of error under the alternative hypothesis of different attack scenarios; Calculate the sample quantity threshold that meets the allowable range of the first type of error and the allowable range of the second type of error according to the functional relationship between the sample quantity threshold and the test value.

7. The method for lidar spoofing detection based on hypothesis testing according to claim 6, wherein The calculating of the optimal sampling window value according to the relationship between the sample quantity threshold and power consumption includes the steps of: Calculate the functional relationship between the sample quantity threshold and power consumption according to the power consumption corresponding to different sample quantity thresholds; Calculate the functional relationship between the sample quantity threshold and efficacy according to the functional relationship between the sample quantity threshold and the test value and the functional relationship between the sample quantity threshold and power consumption; Calculate the sample quantity threshold corresponding to the maximum efficacy according to the functional relationship between the sample quantity threshold and efficacy, and use this as the optimal sampling window value.

8. The method for lidar spoofing detection based on hypothesis testing according to claim 1, wherein The comparison and verification of whether the mean value of the sample variables of the lidar has a significant change according to the two-sample hypothesis testing method includes the steps of: Construction of the statistic for the two-sample Hotelling T 2 test method; Calculate the quantile with the significance level within the allowable range of the type I error according to the probability distribution of the statistic; Judge whether the mean value of the sample variables of the lidar has a significant change according to whether the statistic is greater than the quantile with the significance level within the allowable range of the type I error.

9. A computer-readable storage medium storing a computer program for electronic data interchange, wherein, The computer program causes the computer to execute the method according to any one of claims 1-8.

10. A lidar spoofing detection system based on hypothesis testing, characterized in that, Comprising: A processor; A memory; And One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the processor, and the programs cause the computer to execute the method according to any one of claims 1-8.

Images