An online automobile CAN network anomaly detection system

Through the online automotive CAN network abnormality detection system, the model training and update is used to use cloud servers to solve the problem of large computing burden and low model training efficiency in the existing technology, and high accuracy and diversity recognition are achieved, suitable for different models and and real-time upgrade and scalability.

CN111988342BActive Publication Date: 2024-11-22DALIAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010988464.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-09-18
Publication Date
2024-11-22
Estimated Expiration
2040-09-18

AI Technical Summary

Technical Problem

The information security detection technology of existing automotive CAN networks has problems such as high computing burden, poor communication real-time performance, low model training efficiency, and inability to upgrade and expand.

Method used

Adopt the online automotive CAN network abnormality detection system, train the CAN network abnormality detection model through the cloud server, and transmit the trained model to the on-board terminal for updates. Using the powerful computing power and multi-terminal data samples of the cloud server, real-time Detection and model updates.

Benefits of technology

It improves the accuracy and variety of types of security attack identification, is suitable for different models and regions, saves computing resources of on-board terminals, and realizes real-time upgrade and scalability of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111988342B_ABST
    Figure CN111988342B_ABST
Patent Text Reader

Abstract

The invention discloses an online automobile CAN network anomaly detection system, which belongs to the technical field of information security and intrusion detection of networked vehicles. The system comprises: a cloud server for training a CAN network anomaly detection model according to regional characteristics, user characteristics, vehicle product characteristics and characteristics of changes in security attacks on the entire network to obtain a trained CAN network anomaly detection model; the cloud server transmits the trained CAN network anomaly detection model to different vehicle-mounted terminals for performing CAN network anomaly detection of the vehicle, and at the same time, the vehicle-mounted terminal transmits real-time operation data of the vehicle to the cloud server to update the CAN network anomaly detection model, which can be upgraded in real time and has good scalability; when a single machine is used to train a model library, once the training is completed and the model library is loaded on the vehicle, it is finalized and cannot be upgraded or expanded; the system has a broad application prospect and value in the technical field of information security and intrusion detection of networked vehicles.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of networked vehicle information security and intrusion detection, and in particular to an online automobile CAN network anomaly detection system. Background Art

[0002] With the development of the four trends of automobiles (electrification, intelligence, networking, and sharing), automobiles as means of transportation are no longer an isolated system, but have gradually developed into a complex mobile network terminal. Most of the communication between the control units inside the car is through the CAN bus. With the in-depth research on intelligent networked cars, the number of the car's own control units and the number of communication ports and external devices with the outside world is increasing, and the complexity of the automobile communication network is increasing. With the increasing number of cars, the degree of information sharing is getting higher and higher, but the problem of automobile information security is becoming increasingly prominent.

[0003] At present, there are mainly the following CAN bus information security detection technologies: identity authentication, data encryption, and intrusion detection. The first two mainly use encryption technology and authentication methods to protect CAN network data and isolate it from the outside world to prevent information that does not comply with the protocol from entering. Intrusion monitoring selects corresponding algorithms and detects real-time data based on data characteristics. The relevant technologies are as follows:

[0004] Application publication number CN106143364A discloses an information security method for distributed controllers of electric vehicles. This method adds an embedded MCU and CAN transceiver inside each control unit. The embedded MCU has a built-in encryption and decryption algorithm to ensure that valid CAN information is not stolen and prevent attacks from malicious commands, thus playing a role of isolation and protection.

[0005] Application publication number CN108040082A discloses a connection device and a data transmission method. The invention provides a connection device to physically isolate the vehicle terminal and the vehicle CAN network, avoiding the information security issues caused by the vehicle network terminal directly reading the CAN bus message, and isolating and protecting the CAN bus data. The connection device includes an offline download module, a vehicle CAN bus reading module, a programmable logic module and an open CAN protocol module.

[0006] Application publication number CN105279421A discloses an information security detection method based on vehicle networking access to OBDⅡ, the method steps: establish an OBDⅡ command whitelist; filter and parse CAN messages (from outside to inside); determine whether it is a malicious command based on the whitelist and the sending frequency; make corresponding alarms and handle. The filtering device has a command judgment module, a command control module, a command whitelist, a communication module, an MCU and a power supply.

[0007] Application publication number CN106878130A discloses a method for detecting abnormalities in the CAN network of electric vehicles. The method comprises the following steps: determining whether an external device has sent a virtual message by receiving and sending a message ID; and disconnecting the power supply of a data terminal to determine whether the data terminal has been invaded. The invention designs a set of detection devices, with detection modules distributed in each CAN network.

[0008] Application publication number CN110149345A discloses a vehicle network intrusion detection method based on message sequence prediction. The steps of the method are: collecting vehicle CAN bus data through the T-box vehicle terminal to obtain a standard data set; analyzing whether the vehicle is under security threat based on the operation scenario composed of three attack paths; learning and forming an evaluation detector based on the normal message data set and the abnormal message feature library; and detecting and evaluating the input message through the evaluation detector.

[0009] Application publication number CN110149258A discloses a method for detecting abnormal data in a vehicle CAN bus network based on isolation forests. The method first establishes an isolation tree and an isolation forest based on word samples in a training set; then, the abnormal score of the data to be tested is calculated based on the isolation forest to determine whether it is abnormal.

[0010] Application publication number CN109067773A discloses a vehicle-mounted CAN network intrusion detection method and system based on neural network. He uses the transmission frequency of various CAN network data packets as the input of BP neural network, uses principal component analysis PCA to reduce the data, and detects the transmission frequency of various CAN data packets; uses the correctness of relevant data such as engine speed and intake volume as the input of BP neural network, and detects their correctness in real time. As long as there is an abnormality, it will be judged as abnormal.

[0011] Combined with the specific conditions of vehicle driving and vehicle network, the above technologies have several shortcomings: encryption technology will increase the information composition of CAN frames, change the CAN communication protocol, and is not conducive to actual implementation; the data encryption and decryption process and identity authentication process will increase the ECU computing burden, affect the real-time communication, and also cause a certain message to occupy the CAN network for too long; intrusion detection system, the amount of calculation required for model training is large, and the computing power of the vehicle ECU is insufficient to support it. Model training requires a large number of samples to support, and single-machine training takes a lot of time, and the training efficiency is low. When training the model library on a single machine, once the training is completed and loaded on the vehicle, it is finalized and cannot be upgraded or expanded. Summary of the invention

[0012] In view of the problems existing in the prior art, the present invention discloses an online automobile CAN network anomaly detection system, comprising:

[0013] According to the regional characteristics, user characteristics, vehicle product characteristics and the characteristics of the changes in the security attacks on the entire network, the CAN network anomaly detection model is trained to obtain a cloud server for the trained CAN network anomaly detection model;

[0014] The cloud server transmits the trained CAN network anomaly detection model to different vehicle-mounted terminals for CAN network anomaly detection of the vehicle. At the same time, the vehicle-mounted terminal transmits the real-time operation data of the vehicle to the cloud server to update the CAN network anomaly detection model.

[0015] Furthermore, the vehicle-mounted terminal includes a remote data interaction module for transmitting the trained CAN network anomaly detection model;

[0016] A detection model storage module that receives and saves the trained CAN network anomaly detection model transmitted by the remote data interaction module;

[0017] A CAN protocol analysis module connected to the vehicle's CAN bus and parsing CAN bus messages;

[0018] A central processing unit module that receives the CAN bus message signal transmitted by the CAN protocol analysis module and the CAN network anomaly detection model transmitted by the detection model storage module, and detects whether the CAN network is abnormal or attacked according to the CAN network anomaly detection model, and transmits an alarm signal when the CAN network is abnormal or attacked;

[0019] A power module that provides power to the remote data interaction module, CAN protocol analysis module, detection model storage module, and central processing unit module.

[0020] Furthermore, the system also includes an alarm module which receives the alarm signal transmitted by the central processing unit module and issues an alarm.

[0021] Due to the adoption of the above technical scheme, the present invention provides an online automobile CAN network anomaly detection system, in which the vehicle terminal uploads the CAN bus message to the cloud server, the cloud server trains the CAN network anomaly detection model, and updates the trained CAN network anomaly detection model in the vehicle terminal. This method has the following advantages: 1) There are many types of security attack identification and high accuracy; the cloud server has powerful computing power and can obtain big data samples of multiple terminals, which can accurately establish a sample library of security attack models such as DoS, tampering, discarding, replaying, flooding, and theft. The entire system has the advantages of many types of security attack identification and high accuracy;

[0022] 2) Strong applicability to different vehicle models and regions, short sample library training time, and saving of computing resources for vehicle terminals. The training of the CAN network anomaly detection model requires a large number of samples and powerful computing resource support. The cloud server can simultaneously receive big data samples from vehicle terminals of multiple regions and different models of vehicles, and conduct targeted training of security attack sample libraries based on the characteristics of the regions and models. The entire system has the advantages of strong applicability to different vehicle models and regions, short model training time, and saving of computing resources for vehicle terminals.

[0023] 3) It can be upgraded in real time and has good scalability; when training the model library on a single machine, once the training is completed and loaded on the vehicle, it is finalized and cannot be upgraded or expanded. However, by combining the vehicle terminal with the cloud server, the system can combine the changing characteristics and trends of security attacks on the entire network, and update the detection model of the vehicle terminal in a targeted and real-time manner. The entire system has good scalability and maintainability; based on the above reasons, the present invention has a broad application prospect and value in the field of information security and intrusion detection technology of networked vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 A schematic diagram of an online automobile CAN network anomaly detection system provided by the present invention;

[0026] Figure 2 A schematic diagram of the connection between the vehicle-mounted terminal provided by the present invention and the vehicle CAN bus;

[0027] Figure 3 The circuit schematic diagram of the online automobile CAN bus vehicle terminal provided by the present invention;

[0028] Figure 4 It is the circuit diagram of the CAN protocol analysis module of the present invention;

[0029] Figure 5 A circuit diagram of a central processing unit module of the present invention;

[0030] Figure 6 A circuit diagram of a remote data interaction module of the present invention;

[0031] Figure 7 A circuit diagram of a detection model storage module of the present invention;

[0032] Figure 8A circuit diagram of an alarm module of the present invention;

[0033] Fig. 9 is a circuit diagram of a power module of the present invention;

[0034] Fig.10 The present invention provides an online automobile CAN bus abnormality detection process flow diagram.

[0035] In the figure: 1. CAN protocol analysis module, 2. Central processing unit module, 3. Remote data interaction module, 4. Detection model storage module, 5. Alarm module; 6. Power module; 7. Cloud server; 10. Vehicle terminal. DETAILED DESCRIPTION

[0036] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0037] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] Figure 1 The overall framework of the system includes a cloud server 7 and a vehicle terminal 10;

[0039] The cloud server 7 trains the CAN network anomaly detection model according to regional characteristics, user characteristics, vehicle product characteristics, and the characteristics of changes in network-wide security attacks;

[0040] The vehicle terminal 10 transmits the trained CAN network anomaly detection model to many different vehicle terminals 10 for CAN network anomaly detection of the vehicle. At the same time, the vehicle terminal 10 transmits the real-time operation data of the vehicle to the cloud server 7 to update the CAN network anomaly detection model.

[0041] The vehicle terminal 10 includes: a CAN protocol analysis module 1, a central processing unit module 2, a remote data interaction module 3, a sample library storage module 4, an early warning module 5 and a power supply module 6;

[0042] The remote data interaction module 3 receives the trained CAN network anomaly detection model transmitted by the cloud server 7 and sends sample data to the cloud server 7;

[0043] The detection model storage module 4 receives the trained CAN network anomaly detection model transmitted by the remote data interaction module 3 and saves it;

[0044] The CAN protocol analysis module 1 is connected to the CAN bus of the vehicle and analyzes the CAN bus message;

[0045] The central processing unit module 2 receives the CAN bus message signal transmitted by the CAN protocol analysis module 1 and the CAN network anomaly detection model transmitted by the detection model storage module 4, and detects whether the CAN network has an anomaly or is under attack according to the CAN network anomaly detection model;

[0046] The power supply module 6 provides power for the remote data interaction module vehicle 3 , the CAN protocol analysis module 1 , the detection model storage module 4 , the central processing unit module 2 and the alarm module 5 .

[0047] The alarm module 5 receives the alarm signal transmitted by the central processing unit module 2 and issues an alarm.

[0048] Figure 2 This is a schematic diagram of the connection between the vehicle-mounted terminal provided by the present invention and the vehicle CAN bus. Figure 2 Two connection schemes are shown in Figure 2 The vehicle terminal can be embedded in the automobile CAN bus network. The data flow of the bus passes through the CAN protocol analysis module 1, and after data analysis, the information is understood and important data is extracted, and then flows into the central processing unit module 2 for real-time detection; this solution also provides a connection method, and the CAN protocol analysis module 1 can also be connected to the CAN bus through the OBD interface, and then the CAN bus data flow is obtained to complete abnormality detection.

[0049] The circuit schematic diagram of the vehicle-mounted terminal 10 is as follows Figure 3As shown, the CAN bus protocol analysis module 1 is composed of TJA1050T, where TJA1050T is responsible for the communication level conversion and data reception and transmission of the CAN bus. The specific protocols include: ISO14230-4 address mode, ISO14230-4 fast mode, ISO9141-2, ISO15765-4 standard 11bit500K, ISO15765-4 extended 29bit500K, ISO15765-4 standard 11bit 250K, ISO15765-4 extended 29bit 250K.

[0050] Figure 4 The present invention is a circuit diagram of a CAN protocol analysis module; the chip model used in the CAN protocol analysis module 1 is TAJ1050, the RS pin of the chip TJA1050T is connected to the anode of the diode D7, the anode of the diode D8, one end of the capacitor C3 and one end of the capacitor C4, the CANH pin of the chip TJA1050T is connected to the other end of the capacitor C3 and the other end of the capacitor C4, and the CANL pin of the chip TJA1050T is connected to the cathode of the diode D7 and the cathode of the diode D8.

[0051] Figure 5 The circuit diagram of the central processing unit module of the present invention is as follows; the central processing unit module 2 is composed of an STM32F103CBT6 chip, the NRST pin of the STM32F103CBT6 is connected to the ground through a capacitor C18, the PB10 pin and the PB11 pin of the STM32F103CBT6 are connected to resistors R1 and R22 respectively, and the VSS_1 pin, the VSS_2 pin and the VSS_3 pin of the STM32F103CBT6 are connected to the ground.

[0052] Figure 6 The circuit diagram of the remote data interaction module of the present invention is shown in Figure 3. The remote data interaction module 3 adopts ZTE's ZM9000 module, which has NSA / SA dual-mode 5G technology and is compatible with 5G / 4G / 3G / 2G and China Mobile / China Unicom / China Telecom communication technologies. The ZM9000 module is connected to the main processor STM32F103 using a serial communication interface.

[0053] Figure 7The circuit diagram of the detection model storage module of the present invention; the detection model storage module 4 is composed of a W25N512 chip, and its working principle is a FLASH type flash memory chip, and the storage capacity can be expanded to 512M. The W25N512 is connected to the main processor STM32F7 using an SPI communication interface, and the V_BCKP pin of the W25N512 chip is connected to one end of the capacitor C9, one end of the capacitor C11, and one end of the capacitor C13; the other end of the capacitor C9, the other end of the capacitor C11, and the other end of the capacitor C13 are all connected to the ground.

[0054] Figure 8 The circuit diagram of the alarm module of the present invention is as follows; the alarm module 5 adopts the ESP32-U4WDH chip, the 1Y pin of the ESP32-U4WDH is connected through one end of the resistor R4, and the 2Y pin of the ESP32-U4WDH is connected through one end of the resistor R3. The Bluetooth communication part and the WIFI communication part are integrated in the ESP32-U4WDH chip, and the dual-antenna and dual-frequency communication modes are supported. The ESP32-U4WDH is connected to the main processor STM32F103CBT6 using the SPI communication interface.

[0055] Fig. 9 The circuit diagram of the power module of the present invention is shown in FIG. 6 . The power module 6 adopts a 12V to 5V power module, whose input end is powered by the vehicle-mounted 12V and whose output is 5V, providing power supply for the entire vehicle-mounted terminal. The chip model used by the power module 6 is TPS5430. The VIN pin of the TPS5430 is connected to one end of the capacitor C6, the pin of the diode D1 and one end of the resistor R27. The BOOT pin of the TPS5430 is connected to one end of the capacitor C5 and the anode of the diode D3.

[0056] Fig.10 The schematic diagram of the online automobile CAN bus anomaly detection processing flow provided by the present invention is that the central processing unit module 2 performs safety detection based on the real-time CAN bus data to be tested and the CAN network anomaly detection model. After the anomaly is detected, an alarm is triggered. The real-time dynamic data of the automobile will also be transmitted to the cloud server to improve the sample library, so that the sample library is more in line with the user's driving habits, and a user-personalized sample library is established to prevent false detection and missed detection; the updated CAN network anomaly detection model will also be transmitted to the vehicle terminal 10 and stored in the detection model storage module for use.

[0057] The cloud server 7 trains the CAN network anomaly detection model based on the principle of information entropy, that is, when there is an anomaly or intrusion, the entropy value will change: when there is a replay attack, the information entropy of the CAN bus system increases; when the replay starts, the relative distance of the message changes to a positive value; when the replay ends, the relative distance of the message changes to a negative value. In the process of building the CAN network anomaly detection model, according to the collected original data, that is, the sample data used in the CAN network anomaly detection model training model, after pre-processing, the entropy value under normal circumstances, the threshold for detecting anomalies, and the monitoring sampling cycle are determined to complete the establishment of the CAN network anomaly detection model. The model has abnormal data transmission frequency, such as replay and flooding attacks;

[0058] The CAN network anomaly detection model solution of the cloud server 7 can also be based on the decision tree principle, that is, the construction of the model constructs a tree with the fastest data purity increase based on data purity, which includes three stages: decision tree model attribute selection, decision tree model generation and decision tree model pruning; random tree generation adopts binary method or random hyperplane method. Then, the anomaly score is constructed according to the isolated tree to make anomaly judgment. The random forest is a tree group composed of isolated decision trees with different IDs. The model can detect message data anomalies, such as CAN frame tampering, illegal message injection and other attacks.

[0059] The CAN network anomaly detection model solution of the cloud server 7 can also be based on the neural network principle, that is, by constructing a neural network model to detect whether an event is abnormal, whether a fault occurs, or to predict the development of things; this method can be jointly detected from two aspects: the data packet sending frequency of the CAN network is used as the input of a BP neural network to train the model, and the CAN network anomaly detection model can detect denial of service attacks, replay and injection attacks; related data (such as engine speed, intake volume, vehicle speed, etc.) is used as the input of another BP neural network to train the model. If one of the data is abnormal and there is an error in the correlation with other data, it is judged as an abnormality. The model can detect tampering attacks and man-in-the-middle attacks.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An online automobile CAN network anomaly detection system, characterized by: include: According to the regional characteristics, user characteristics, vehicle product characteristics and the characteristics of the changes in the security attacks on the entire network, the CAN network anomaly detection model is trained to obtain a cloud server (7) of the trained CAN network anomaly detection model; The cloud server (7) transmits the trained CAN network anomaly detection model to different vehicle-mounted terminals (10) to perform CAN network anomaly detection on the vehicle, and the vehicle-mounted terminal (10) transmits the real-time operation data of the vehicle to the cloud server (7) to update the CAN network anomaly detection model; The vehicle-mounted terminal (10) comprises a remote data interaction module (3) for transmitting a trained CAN network anomaly detection model; A detection model storage module (4) that receives and stores the trained CAN network anomaly detection model transmitted by the remote data interaction module (3); A CAN protocol analysis module (1) connected to a CAN bus of a vehicle and analyzing CAN bus messages; A central processing unit module (2) receives the CAN bus message signal transmitted by the CAN protocol analysis module (1) and the CAN network anomaly detection model transmitted by the detection model storage module (4), and detects whether the CAN network has an anomaly or is under attack according to the CAN network anomaly detection model; A power supply module (6) for providing power to the remote data interaction module (3), the CAN protocol analysis module (1), the detection model storage module (4), and the central processing unit module (2); The system also includes an alarm module (5) for receiving an alarm signal transmitted by the central processing module (2) and issuing an alarm; The central processing unit module (2) performs safety detection based on the real-time CAN bus data to be tested and the CAN network anomaly detection model. After detecting an anomaly, an alarm is triggered. The real-time dynamic data of the car will also be transmitted to the cloud server to improve the sample library, so that the sample library is more in line with the user's driving habits, and a user-personalized sample library is established to prevent false detection and missed detection; the updated CAN network anomaly detection model will also be transmitted to the vehicle terminal 10 and stored in the detection model storage module for future use; The cloud server (7) trains the CAN network anomaly detection model based on the principle of information entropy, that is, when there is an anomaly or intrusion, the entropy value will change: when there is a replay attack, the information entropy of the CAN bus system increases; when the replay starts, the relative distance of the information entropy change is positive; when the replay ends, the relative distance of the message changes to a negative value. In the process of building the CAN network anomaly detection model, the entropy value under normal conditions, the threshold for detecting anomalies, and the monitoring sampling period are determined after preprocessing based on the collected original data, that is, the sample data used in the CAN network anomaly detection model training model, to complete the establishment of the CAN network anomaly detection model. The model has abnormal data transmission frequency, including replay and flooding attacks; The CAN network anomaly detection model solution of the cloud server (7) may be based on the decision tree principle, that is, the construction of the model constructs a tree with the fastest data purity increase based on data purity, which includes three stages: decision tree model attribute selection, decision tree model generation and decision tree model pruning; random tree generation adopts binary method or random hyperplane method, and then anomaly judgment is performed based on the anomaly score constructed according to the isolated tree. The random forest is a tree group composed of isolated decision trees with different IDs. The model can detect message data anomalies, including CAN frame tampering and message illegal injection attacks; The CAN network anomaly detection model solution of the cloud server (7) may adopt the basic neural network principle, that is, by constructing a neural network model to detect whether an event is abnormal, whether a fault occurs, or to predict the development of things; the method can perform joint detection in two aspects: using the data packet transmission frequency of the CAN network as an input of a BP neural network to train the model, and the CAN network anomaly detection model can detect denial of service attacks, replay and injection attacks; using data with correlation as the input of another BP neural network to train the model, if a certain data is abnormal and there is an error in the correlation with other data, it is judged as abnormal, and the model includes detection of tampering attacks and man-in-the-middle attacks.