A system and method for detecting remote vehicle state estimation tampering attacks.

Abstract

This invention discloses a system and method for detecting tampering attacks on vehicle remote state estimation. The invention uses event triggers and anomaly detectors to detect attacks on the vehicle sensor transmission channels. Event triggers are deployed at the vehicle state transmitter and the remote estimator state receiver. Through the event triggering mechanism, the data transmitted in the channel is set to either 0 or 1. At this time, an attacker can only launch an attack at the triggering moment and can only tamper with the data from 0 to 1 or from 1 to 0, reducing the attacker's attack capability. Furthermore, when tampering occurs, it is equivalent to injecting a lower-bound attack, causing the detector's detection signal to change, improving the attack detection effect. The data transmission and detection mechanism of this invention not only reduces the data transmission volume between the vehicle and the remote estimator but also ensures timely detection of tampering attacks.

Description

Technical Field

[0001] This invention relates to a vehicle attack detection system and method in the field of vehicle communication security, and more particularly to a vehicle remote state estimation tampering attack detection system and method. Background Technology

[0002] Connected electric vehicles (EVs) are equipped with advanced onboard sensors, controllers, and actuators, integrating modern communication and network technologies to achieve V2X intelligent information exchange and sharing. They possess complex environmental perception, intelligent decision-making, collaborative control, and execution functions, enabling safe, comfortable, energy-efficient, and high-performance driving—a new generation of automobiles. However, while technologies like vehicle-to-everything (V2X) enhance the intelligence and convenience of vehicles, they also bring serious security challenges. Due to the increased number of onboard sensors, controllers, and actuators, and the existence of issues such as insufficient communication protocol security and incomplete encryption coverage at the network level, security vulnerabilities such as remote attacks, malicious control, and the theft of user privacy data can endanger the lives and property of vehicles, roads, the environment, and even national security.

[0003] Remote estimation in connected electric vehicles aims to remotely estimate the vehicle's operating status using sensor data for purposes such as monitoring, path planning, and control. However, remote estimators are vulnerable to attacks. DoS (Denial of Service) attacks and spoofing attacks are two common types of attacks on remote estimators. The main goal of a DoS attack is to disrupt the transmission channel with limited energy, minimizing estimation performance. DoS attacks are unspy because they can be detected as long as the estimator is not receiving data. Compared to DoS attacks, spoofing attacks have a greater impact on remote estimators because they can reduce estimation performance while avoiding detection. A major factor in the success of spoofing attacks is the attacker's ability to arbitrarily modify data in the communication channel. Sensor data can be altered to any data designed by the attacker to achieve their objectives. Summary of the Invention

[0004] To address the problems in the background art, this invention provides a vehicle remote state estimation tampering attack detection system and method. This invention introduces an event-triggered mechanism to reduce the attacker's tampering ability and improve the detection rate when tampering occurs. Specifically, firstly, it ensures that attackers can only tamper with data when an event is triggered; secondly, the event-triggered mechanism can binarize the transmitted data, and the binarized data can only be tampered with in a specific direction. Furthermore, once tampering occurs, the detector will find it easier to detect, thus reducing the attacker's tampering ability and improving the detection rate when tampering occurs.

[0005] The technical solution adopted in this invention is as follows:

[0006] I. A Vehicle Remote State Estimation Tampering Attack Detection System

[0007] The detection system includes vehicle sensors, event triggers, encoders, communication networks, decoders, state estimators, and detectors. The vehicle sensors are connected to the event triggers, the event triggers are connected to the encoders, the encoders transmit encoded signals to the decoders through the communication networks, the decoders are connected to the state estimators, and the state estimators are connected to the detectors.

[0008] In the event trigger, the system determines whether to trigger based on the actual front and rear state information of the vehicle sensors. When the absolute value of the difference between the actual front and rear states of the vehicle sensors exceeds a preset trigger threshold, a trigger signal is sent to the encoder; otherwise, no action is taken.

[0009] In the encoder, when a trigger signal is received from an event trigger, the encoded value is set to 1; otherwise, it is set to 0, and the obtained encoded signal is transmitted to the decoder through the communication network.

[0010] The decoder decodes the received encoded signal. When the encoded value is 1, the state reference value is added to a preset trigger threshold and used as the vehicle's transmission state information. The vehicle's transmission state information is then updated to the state reference value. When the encoded value is 0, the state reference value is subtracted from the preset trigger threshold and used as the vehicle's restored state information. Finally, the vehicle's restored state information is sent to the state estimator.

[0011] The state estimator performs remote estimation of the vehicle state based on the restored state information of the vehicle, and obtains the estimated state information of the vehicle.

[0012] In the state estimator, the vehicle model is used to predict the historical estimated state information to obtain the predicted state information of the vehicle. Then, based on the restored state information of the vehicle, the Kalman filter method is used to correct the predicted state information of the vehicle to obtain the estimated state information of the vehicle. At the same time, the estimated state information of the vehicle is used as the historical estimated state for the next moment.

[0013] The detector determines whether the restored state information is abnormal based on the estimated state information of the vehicle. When the difference between the estimated state information and the restored state information exceeds a preset state threshold, it is determined that the received data is abnormal and sends a message to the decoder not to update the state reference value; otherwise, it does not send a message.

[0014] II. A method for detecting remote vehicle state estimation tampering attacks

[0015] Step 1: After the vehicle sensor collects the vehicle's status, it sends the vehicle's actual status information to the event trigger;

[0016] Step 2: Based on the actual status information sent by the vehicle sensors, the event trigger determines whether to trigger and then sends the event trigger result to the encoder;

[0017] Step 3: When the encoder receives the trigger signal sent by the event trigger, it sets the encoded value to 1; otherwise, it sets it to 0 and transmits the obtained encoded signal to the decoder through the communication network.

[0018] Step 4: The decoder decodes the received encoded signal. When the encoded value is 1, the state reference value is added to the preset trigger threshold and used as the vehicle's transmission state information. The vehicle's transmission state information is then updated to the state reference value. When the encoded value is 0, the state reference value is subtracted from the preset trigger threshold and used as the vehicle's restored state information. Finally, the vehicle's restored state information is sent to the state estimator.

[0019] Step 5: Based on the restored state information of the vehicle, the state estimator performs remote estimation of the vehicle state to obtain the estimated state information of the vehicle.

[0020] Step 6: Based on the estimated state information of the vehicle, the detector determines whether the restored state information is abnormal. When the difference between the estimated state information and the restored state information exceeds the preset state threshold, it is determined that the received data is abnormal, and the detector sends a message to the decoder not to update the state reference value; otherwise, it does not send a message.

[0021] In step 3, the event trigger determines whether to trigger based on the front and rear state information of the vehicle sensor. When the absolute value of the difference between the front and rear states of the vehicle sensor exceeds the preset trigger threshold, a trigger signal is sent to the encoder; otherwise, no action is taken.

[0022] In step 5, the state estimator uses the vehicle model to predict the historical estimated state information to obtain the predicted state information of the vehicle. Then, based on the restored state information of the vehicle, the Kalman filter is used to correct the predicted state information of the vehicle to obtain the estimated state information of the vehicle. At the same time, the estimated state information of the vehicle is used as the historical estimated state for the next moment.

[0023] The beneficial effects of this invention are:

[0024] This invention simplifies the data transmitted in the data transmission channel to 0 or 1 through an event triggering mechanism. At this time, the attacker can only launch an attack at the triggering moment and can only tamper with the data from 0 to 1 or from 1 to 0, which reduces the attacker's attack capability.

[0025] Furthermore, when tampering occurs, it is equivalent to injecting a lower-bound attack, and the detector's detection signal will send a jump, improving the detection effect of the attack.

[0026] By utilizing the data transmission and detection mechanism of this invention, not only can the data transmission volume between the vehicle and the remote estimator be reduced, but also timely detection of tampering attacks can be ensured. Attached Figure Description

[0027] Figure 1 This is an overall block diagram of the present invention.

[0028] Figure 2 It is a simulation diagram of the vehicle in its actual state.

[0029] Figure 3 This is a simulation diagram of the estimated state of the remote estimator.

[0030] Figure 4 This is a simulation diagram of the detector's performance.

[0031] In the diagram: 1. Vehicle sensor; 2. Event trigger; 3. Encoder; 4. Communication network; 5. Decoder; 6. State estimator; 7. Detector. Detailed Implementation

[0032] The following is a further explanation with reference to the accompanying drawings.

[0033] like Figure 1 As shown, the system includes a vehicle sensor 1, an event trigger 2, an encoder 3, a communication network 4, a decoder 5, a state estimator 6, and a detector 7. The vehicle sensor 1, event trigger 2, and encoder 3 are all deployed within the vehicle, while the decoder 5, state estimator 6, and detector 7 are located remotely. The vehicle sensor 1 is connected to the event trigger 2, which in turn is connected to the encoder 3. The encoder 3 transmits encoded signals to the decoder 5 via the communication network 4, which is a wireless communication network. The decoder 5 is connected to the state estimator 6, which in turn is connected to the detector 7. The vehicle sensor is used to detect the vehicle's actual state information in real time; this information includes the vehicle's position and attitude. The event trigger determines whether to send the latest state data based on the actual state information from the vehicle sensor. The encoder encodes the data that the event trigger decides to send. The decoder decodes the received encoded data. The state estimator uses the data decoded by the decoder for remote vehicle state estimation. The detector detects whether the data received by the state estimator is abnormal.

[0034] Event trigger 2 determines whether to trigger based on the actual status information sent by vehicle sensor 1, and then sends the event trigger result to encoder 3.

[0035] In event trigger 2, it is determined whether to trigger based on the actual state information of the vehicle sensor 1 before and after. When the absolute value of the difference between the actual state of the vehicle sensor 1 before and after exceeds the preset trigger threshold, that is, when the absolute value of the difference between the actual sensor output and the state recorded by the event trigger is greater than or equal to the preset trigger threshold, and each state has a corresponding trigger threshold, a trigger signal is sent to encoder 3; otherwise, no action is taken.

[0036] In encoder 3, when a trigger signal is received from event trigger 2, the encoded value is set to 1; otherwise, it is set to 0, and the obtained encoded signal is transmitted to decoder 5 through communication network 4.

[0037] Decoder 5 decodes the received coded signal. When the coded value is 1, the state reference value is stored in decoder 5. The state reference value is added to a preset trigger threshold to obtain the vehicle's transmission state information. The vehicle's transmission state information is then updated to the state reference value and stored in decoder 5. When the coded value is 0, the state reference value is subtracted from the preset trigger threshold to obtain the vehicle's restored state information. Finally, the vehicle's restored state information is sent to state estimator 6. Through the cooperation of the encoder and decoder, the real data transmitted by the sensor can be approximately restored.

[0038] The state estimator 6 remotely estimates the vehicle state based on the restored state information of the vehicle, and obtains the estimated state information of the vehicle.

[0039] In state estimator 6, the vehicle model is used to predict the historical estimated state information to obtain the predicted state information of the vehicle. Then, based on the restored state information of the vehicle, the Kalman filter method is used to correct the predicted state information of the vehicle to obtain the estimated state information of the vehicle. At the same time, the estimated state information of the vehicle is used as the historical estimated state for the next moment.

[0040] Detector 7 determines whether the restored state information is abnormal based on the estimated state information of the vehicle. When the difference between the estimated state information and the restored state information exceeds the preset state threshold, it is determined that the received data is abnormal and sends a message to decoder 5 to not update the state reference value; otherwise, it does not send a message.

[0041] The method includes the following steps:

[0042] Step 1: After collecting the vehicle's status, vehicle sensor 1 sends the vehicle's actual status information to event trigger 2.

[0043] Step 2: Based on the actual status information sent by vehicle sensor 1, event trigger 2 determines whether to trigger, and then sends the event trigger result to encoder 3;

[0044] In event trigger 2, it is determined whether to trigger based on the front and rear state information of vehicle sensor 1. When the absolute value of the difference between the front and rear states of vehicle sensor 1 exceeds the preset trigger threshold, that is, the absolute value of the difference between the actual sensor output and the state recorded by the event trigger is greater than or equal to the preset trigger threshold, each state has a corresponding trigger threshold, then a trigger signal is sent to encoder 3; otherwise, no action is taken.

[0045] Step 3: When the trigger signal sent by event trigger 2 is received, encoder 3 sets the encoded value to 1; otherwise, it sets it to 0 and transmits the obtained encoded signal to decoder 5 through communication network 4.

[0046] Step 4: Decoder 5 decodes the received coded signal. When the coded value is 1, the state reference value is stored in decoder 5. The state reference value is added to a preset trigger threshold and used as the vehicle's transmission state information. The vehicle's transmission state information is updated to the state reference value and stored in decoder 5. When the coded value is 0, the state reference value is subtracted from the preset trigger threshold and used as the vehicle's restored state information. Finally, the vehicle's restored state information is sent to state estimator 6. Through the cooperation of encoder and decoder, the real data transmitted by the sensor can be approximately restored.

[0047] Step 5: Based on the restored state information of the vehicle, the state estimator 6 performs remote estimation of the vehicle state to obtain the estimated state information of the vehicle.

[0048] In step 5, the state estimator 6 uses the vehicle model to predict the historical estimated state information to obtain the predicted state information of the vehicle. Then, based on the restored state information of the vehicle, the Kalman filter method is used to correct the predicted state information of the vehicle to obtain the estimated state information of the vehicle. At the same time, the estimated state information of the vehicle is used as the historical estimated state for the next moment.

[0049] Step 6: Based on the estimated state information of the vehicle, the detector 7 determines whether the restored state information is abnormal. When the difference between the estimated state information and the restored state information exceeds the preset state threshold, it is determined that the received data is abnormal, and the detector 7 sends a message to the decoder 5 to not update the state reference value; otherwise, it does not send a message.

[0050] In order to reduce the attacker's ability to tamper with the data and to make tampering attacks easier to detect, this invention uses an event-triggered transmission mechanism for each channel.

[0051] The event triggering mechanism is designed as follows:

[0052]

[0053] Where, k i and k i+1 Represents the kth and kth i+1 time, For k i The output of the system state at any time, This is an internal variable of the event trigger, where the trigger time is:

[0054]

[0055] Where, ki It is the last point in time when the event was triggered, and until k. i+1 , Keep it constant. d is a constant called the trigger threshold, and inf{.} denotes taking the lower bound.

[0056] The encoder encodes the trigger result as either 1 or 0. The basic idea of ​​the encoding is that each time an event is triggered, there are only two possibilities:

[0057] 1. This means an update It can be approximated as

[0058] 2. in this case, It can be approximated as

[0059] This means that the estimator can receive the same information by transmitting the trigger direction instead of the actual sensor data.

[0060] To transmit the trigger direction, the encoder on the system side is designed as follows:

[0061]

[0062] Here, Θ represents the actual data transmitted. Since Θ∈{0,1}, only one bit bandwidth is needed for transmission.

[0063] To decode the received trigger direction, the decoder on the estimator side is designed as follows:

[0064]

[0065] in, It is the received data. This indicates that the estimator is at time k. i The data decoded and recorded (at the previous moment) For time k i+1 (Latest moment) Decoded latest data, This provides the decoder with output data. This invention can detect whether the transmitted data in the channel follows a trigger direction, i.e., "0" or "1," which is why this method is called a one-bit transmission mechanism. Furthermore, if there is no attack in the channel, then... otherwise, Possible. A block diagram of a remote estimator using an event-triggered mechanism is shown below. Figure 1 As shown.

[0066] Secondly, the design of the remote state estimator is as follows. The remote state estimator requires the vehicle model, so the vehicle kinematic model is first established as follows:

[0067]

[0068]

[0069]

[0070] Where (x, y) are the global coordinates of the center of the vehicle's rear wheel. Let x be the first derivative. y is the first derivative of y, v is the velocity at the center of the rear wheel of the vehicle, and θ is the rotation angle of the vehicle's coordinates relative to the global coordinates (counterclockwise is positive). Let δ be the first derivative of the rotation angle, L be the distance from the center of the rear wheel to the center of the front wheel, and δ be the rotation angle of the front wheel relative to the vehicle's coordinate system (counterclockwise is positive).

[0071] Discretizing the above model and considering the vehicle's uniform motion, while treating the vehicle's speed as a state variable, we obtain:

[0072]

[0073] Where, x k =[x k y k θ k v k ] T Where the subscript k represents time and Δt is the discrete period. Linearizing it and considering process noise, we get:

[0074] x k+1 =Ax k +w k

[0075] Where A is the system matrix, and the formula is as follows:

[0076]

[0077] Process noise w k It follows a Gaussian distribution with a mean of 0 and a variance of Q.

[0078] Assume the vehicle can obtain x and y using lidar, that is:

[0079] y k =Cx k +v k

[0080] Where C is the observation matrix, and the formula is as follows:

[0081]

[0082] v kThe measurement noise follows a Gaussian distribution with a mean of 0 and a variance of R, which is related to w. k independent.

[0083] The remote estimator employs an extended Kalman filter.

[0084]

[0085] M k+1|k =AM k A T +Q

[0086] L k+1 =M k+1|k C T (CM k+1|k C T +R) -1

[0087]

[0088] M k+1 =M k+1|k -L k+1 CM k+1|k

[0089] in, It is an estimated state, M k It is the error covariance, L k It is the gain matrix. and M k+1|k They are and M k The prediction. The initial state of the Kalman filter is And M0 = 0, R represents the model prediction function of the system, and R represents the variance of the measurement noise.

[0090]

[0091] If (A,C) is detectable and It is stable, and the estimated error covariance M k and gain matrix L k Ultimately, it converges to a unique value. Therefore, this invention assumes that the Kalman filter has reached a steady state before time 0, and defines the covariance convergence matrix M and the gain convergence matrix L, i.e.

[0092]

[0093] Among them, C T This represents the transpose of matrix C.

[0094] Thus, the estimator in a stable state is:

[0095]

[0096] Among them, z k+1 For health and new information, to meet

[0097] This invention proposes installing a fault detector next to the estimator to detect anomalous data from the communication network. The detector will be triggered if the following conditions occur:

[0098] g k >α

[0099] Where α is a constant threshold, and g k Selected as χ 2 The detector takes the following form:

[0100]

[0101] Σ represents the covariance of the steady-state estimation error. When there is no attack in the communication channel, g k Satisfying χ 2 It has a low probability of triggering an alarm.

[0102] In practical applications, data transmission networks may be attacked, causing the estimator to be driven by the attacked input data. This invention assumes that the transmission network has suffered a spoofed data injection attack, i.e., when time k, the attacking data a... k Injected and y k The data received by the estimator becomes Therefore, the attacked estimator is:

[0103]

[0104]

[0105] in, and These are the estimated attack and the new information, respectively.

[0106] To describe the impact of data injection attacks, the difference in innovation between estimators with and without attacks is:

[0107]

[0108] This means injecting a k It can also affect health information. k The difference between attacking new information Note that if ||Δz k The size is so small that attack detectors are unlikely to be able to distinguish it. and z k According to the triangle inequality Difference Δzk+1 It should be as small as possible.

[0109] Since the attacker only has the resource to tamper with the communication channel, the attack can only be launched when an event is triggered. For convenience, the steps to launch an attack are represented as k. a The analysis of the tampering attack is as follows.

[0110] Attackers can only modify data from "0" to "1" or from "1" to "0", with a trigger threshold of d. Once an attack is launched, a(k) must be available. a )=2d or a(k a ) = -2d.

[0111] Assume a(k) a = 2d. Since the attack is initiated step-by-step, k a Therefore, the present invention can obtain

[0112]

[0113] and

[0114]

[0115]

[0116] According to the new interest rate spread formula, we can obtain:

[0117]

[0118] Due to similarity, when a(k) a From )=-2d, we can obtain |Δz(k) a )|=2d.

[0119] Furthermore, through Δz(k) a )=-2d

[0120]

[0121] and

[0122]

[0123] Where I represents the identity matrix, and ε1 and ε2 are the errors caused by linearization. Since a k It is an integer multiple of 2d, so we know It can be approximated as an integer multiple of 2d, thus yielding Δz. k It can also be approximated as an integer multiple of 2d, i.e., |Δz(k) a +i)|≈2Υd, i=1,2,3,..., where Υ is an integer related to the triggering time and direction. It can be seen that the new interest rate differential is proportional to the triggering interval.

[0124] Simulation results are as follows Figure 2-4 As shown, Figure 2 Vehicle status. Figure 3 As can be seen from the state estimation of the estimator, under the transmission mechanism proposed in this invention, the estimator can accurately estimate the vehicle's state. Figure 4 As shown in the detector output, this invention adds a tampering attack when k=66. It can be seen that the detector output changes significantly and the attack can be successfully detected.

Claims

1. A vehicle remote state estimation tampering attack detection system, characterized in that, It includes a vehicle sensor (1), an event trigger (2), an encoder (3), a communication network (4), a decoder (5), a state estimator (6), and a detector (7); the vehicle sensor (1) is connected to the event trigger (2), the event trigger (2) is connected to the encoder (3), the encoder (3) transmits the encoded signal to the decoder (5) through the communication network (4), the decoder (5) is connected to the state estimator (6), and the state estimator (6) is connected to the detector (7); In the event trigger (2), whether to trigger is determined based on the actual front and rear state information of the vehicle sensor (1). When the absolute value of the difference between the actual front and rear states of the vehicle sensor (1) exceeds the preset trigger threshold, a trigger signal is sent to the encoder (3); otherwise, no action is taken. In the encoder (3), when a trigger signal sent by the event trigger (2) is received, the encoding value is set to 1; otherwise, it is set to 0, and the obtained encoding signal is transmitted to the decoder (5) through the communication network (4). The decoder (5) decodes the received encoded signal. When the encoded value is 1, the state reference value is added to the preset trigger threshold and used as the transmission state information of the vehicle. The transmission state information of the vehicle is then updated to the state reference value. When the encoding value is 0, the state reference value is subtracted from the preset trigger threshold and used as the restored state information of the vehicle. Finally, the restored state information of the vehicle is sent to the state estimator (6). The state estimator (6) performs remote estimation of the vehicle state based on the restored state information of the vehicle to obtain the estimated state information of the vehicle. In the state estimator (6), the vehicle model is used to predict the historical estimated state information to obtain the predicted state information of the vehicle. Then, based on the restored state information of the vehicle, the Kalman filter method is used to correct the predicted state information of the vehicle to obtain the estimated state information of the vehicle. At the same time, the estimated state information of the vehicle is used as the historical estimated state at the next moment. The detector (7) determines whether the restored state information is abnormal based on the estimated state information of the vehicle. When the difference between the estimated state information and the restored state information exceeds the preset state threshold, it is determined that the received data is abnormal and sends a non-updated state reference value to the decoder (5); otherwise, it does not send it.

2. A method for detecting remote vehicle state estimation tampering attacks, characterized in that, Includes the following steps: Step 1: After the vehicle sensor (1) collects the vehicle's status, it sends the vehicle's actual status information to the event trigger (2); Step 2: Based on the actual status information sent by the vehicle sensor (1), the event trigger (2) determines whether to trigger and then sends the event trigger result to the encoder (3). Step 3: When the trigger signal sent by the event trigger (2) is received, the encoder (3) sets the encoded value to 1; otherwise, it sets it to 0 and transmits the obtained encoded signal to the decoder (5) through the communication network (4); Step 4: The decoder (5) decodes the received encoded signal. When the encoded value is 1, the state reference value is added to the preset trigger threshold and used as the vehicle's transmission state information. The vehicle's transmission state information is then updated to the state reference value. When the encoding value is 0, the state reference value is subtracted from the preset trigger threshold and used as the restored state information of the vehicle. Finally, the restored state information of the vehicle is sent to the state estimator (6). Step 5: Based on the restored state information of the vehicle, the state estimator (6) performs remote estimation of the vehicle state to obtain the estimated state information of the vehicle; Step 6: Based on the estimated state information of the vehicle, the detector (7) determines whether the restored state information is abnormal. When the difference between the estimated state information and the restored state information exceeds the preset state threshold, it is determined that the received data is abnormal and sends a non-updated state reference value to the decoder (5); otherwise, it does not send it.

3. The method for detecting remote vehicle state estimation tampering attacks according to claim 2, characterized in that, In step 3, the event trigger (2) determines whether to trigger based on the front and rear state information of the vehicle sensor (1). When the absolute value of the difference between the front and rear states of the vehicle sensor (1) exceeds the preset trigger threshold, a trigger signal is sent to the encoder (3); otherwise, no action is taken.

4. The method for detecting remote vehicle state estimation tampering attacks according to claim 2, characterized in that, In step 5, the state estimator (6) uses the vehicle model to predict the historical estimated state information to obtain the predicted state information of the vehicle. Then, based on the restored state information of the vehicle, it uses the Kalman filter method to correct the predicted state information of the vehicle to obtain the estimated state information of the vehicle. At the same time, the estimated state information of the vehicle is used as the historical estimated state at the next moment.

Images