A method for extracting physical fingerprint information from CAN device signals

By collecting and analyzing signals on the CAN bus and extracting the physical fingerprint characteristics of the device, the problem of unrecognizing and traceable attack devices in the existing technology is solved, and the identity authentication and attack traceability of the CAN bus device are realized, which improves security performance and maintains the backward compatibility.

CN114154134BActive Publication Date: 2024-11-19SOUTHEAST UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111505886.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-10
Publication Date
2024-11-19
Estimated Expiration
2041-12-10

AI Technical Summary

Technical Problem

The existing CAN bus intrusion detection scheme cannot accurately identify and trace attack equipment, resulting in insufficient security performance.

Method used

By collecting and analyzing physical fingerprint information in the CAN device signal, the physical fingerprint features of the CAN device are extracted by using energy detection, clock synchronization, resampling, analysis and feature extraction steps to achieve identity authentication and attack traceability.

Benefits of technology

It realizes identity authentication and attack tracing of CAN bus devices, provides more effective access control, has data independence and real-time performance, does not affect the existing CAN protocols and circuits, and has good backward compatibility.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114154134B_ABST
    Figure CN114154134B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for extracting physical fingerprint information from a CAN device signal, the method comprising the following steps: using a signal acquisition device to collect a differential signal on a CAN gateway side bus to obtain an original waveform; locating the starting position of a frame signal; clock synchronizing the original waveform from the starting position of the frame; resampling the original waveform using a synchronization signal to obtain an ideal waveform; intercepting a rising signal part and a falling signal part in the original waveform with a symbol width, splicing them together to form a characteristic waveform, and similarly forming a reference waveform from the ideal waveform; calculating the autocorrelation function of the characteristic waveform and the reference waveform respectively, and then subtracting the two to obtain a difference W, at which time W is numerically compressed and smoothed to be used as the physical fingerprint of the device. The present invention can realize the extraction of physical fingerprint information from CAN device signals, and finally realize identity authentication and intrusion detection of ECU devices on the CAN bus, thereby improving the security of CAN bus access.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of communication and information security, and in particular to a method for extracting physical fingerprint information from a CAN device signal. Background Art

[0002] With the development of information modernization and industrial modernization, various devices in the automotive industry and industrial control field are equipped with ECU devices to achieve automation and intelligent control. At the same time, due to the simplicity and reliability of the CAN protocol, it has become the de facto standard for ECU communication. Although the CAN protocol has high performance and reliability in communication, it was originally designed only for the industrial control field. Therefore, when the CAN protocol was initially designed, only the communication content and format between modules were designed, and the security of communication was not considered. For example, data is directly transmitted in plain text on the CAN bus, and there is no relevant terminal authentication mechanism. Any device connected to the CAN bus can send data to other devices on the bus. It is for this reason that the vehicle-mounted CAN bus can be easily attacked. For example, an attacker can easily control any component on the vehicle CAN bus, including the engine, dashboard, radio, and even the ABS system, and can also interfere with the normal operation of all devices on the entire CAN bus.

[0003] At present, there are some methods to detect ECU devices that invade the CAN bus, but these methods mainly focus on the algorithm of designing the intrusion detection system. By real-time monitoring the content of message transmission on the CAN bus, when it is found that the message may be sent by the intrusion device, the user of the CAN bus system is warned so that corresponding measures can be taken. Although these methods can detect whether there is an intrusion on the vehicle CAN bus, these methods cannot locate the intrusion device, cannot determine which ECU devices are performing the intrusion or attack, and thus cannot identify the identity of the attacker and cannot trace the source. However, it is very important to be able to accurately, effectively and quickly locate the identity of the attacker, because it can provide more effective help for evidence collection, isolation and repair. Therefore, the traditional CAN bus intrusion detection solution cannot meet the current demand for security performance.

[0004] The present invention proposes a physical fingerprint extraction method based on terminal devices on the CAN bus, which can more effectively identify intrusion devices. Since the CAN controller and driver contain many electronic components, such as capacitors, inductors, and semiconductor transistors, and these electronic components will bring various differences during the production process due to factors such as different manufacturing processes and different manufacturing environments, it is precisely because of these differences that the output signals of the CAN controller and the driver are different. These output signals are collected, analyzed, and feature extracted, which can be used as the physical fingerprint of the current CAN device. Moreover, the physical fingerprint obtained by this method corresponds to the CAN device one by one and cannot be tampered with. And the fingerprint extraction method proposed in this article is independent of the content of the data sent by the CAN device and has data independence. Summary of the invention

[0005] In order to solve various security problems caused by the lack of identity authentication mechanism in the CAN protocol and the inability to trace the source of attacks, the present invention proposes a method for extracting physical fingerprint information from a CAN device signal, comprising the following steps:

[0006] Step 1, data acquisition: collect and store signals on the CAN bus to obtain the CAN signal sequence Y(n) and the original waveform y(n);

[0007] Step 2, locate the starting position of the frame: perform energy detection on the CAN signal sequence Y(n) to locate the starting position of the frame;

[0008] Step 3, clock synchronization: starting from the frame start position located in step 2, establish a window sequence, and use the window sequence to synchronize each symbol of the frame to form a synchronization signal;

[0009] Step 4, resampling: performing shaping and energy normalization processing on the original waveform y(n), then resampling the original waveform y(n) using the synchronization signal, and reconstructing the level of the resampling result according to the decision threshold to obtain the ideal waveform x(n);

[0010] Step 5, parsing CAN data: according to the CAN frame protocol format, the reconstructed CAN ideal waveform is parsed, and at the same time, it is verified whether it is a complete frame;

[0011] Step 6, intercepting the data domains of the original waveform y(n) and the ideal waveform x(n) respectively according to the analysis results;

[0012] Step 7, extract the characteristic waveform: extract the rising signal part and the falling signal part of the original waveform y(n) to obtain BTT y and TTB y, spliced ​​to form the characteristic waveform O(n); extract the rising signal part and the falling signal part of the ideal waveform x(n) to obtain BTT x and TTB x , spliced ​​to form the reference waveform R(n);

[0013] Step 8, feature extraction: perform autocorrelation operations on the feature waveform O(n) and the reference waveform R(n) respectively, and then calculate the difference to obtain the waveform W. Perform numerical compression and smoothing on the waveform W to obtain the physical fingerprint characteristics of the CAN device signal.

[0014] Preferably, in step 1, the signal on the bus is collected at the gateway position on the CAN bus, and the collected signal is the differential signal between the two signals of the high data line CAN_H and the low data line CAN_L on the CAN bus. The collected differential signal is stored to obtain the original waveform y(n) and signal sequence Y(n) of the CAN signal.

[0015] Preferably, in step 1, the signal collected is a piece of fixed content data or random content data. Fixed data means data with fixed content, such as a device repeatedly sending 0x1010, and random data means random content sent by the device.

[0016] Preferably, in step 2, the energy detection includes using a threshold level to determine the starting position of the frame, and the threshold level is 1 / 2 of the difference between the high level and the low level of the frame.

[0017] Preferably, in step 3, the length of the window sequence is the number of sampling points N used for one symbol, where N is the ratio of the sampling rate of the signal acquisition device to the data transmission rate of the CAN device, wherein the baud rate B of the CAN device and the sampling rate F of the signal acquisition device are known, and the position of each symbol in the signal sequence Y(n) is located according to the number of sampling points N and the starting position of the frame.

[0018] Preferably, in step 4, the decision threshold level is taken from the middle position of each symbol, that is, the level decision result at the position N / 2 is taken as the level of the current symbol.

[0019] Preferably, the data field is taken from a waveform sequence between the first symbol after the DLC segment and the first symbol before the CRC segment in each data frame.

[0020] Preferably, in step 7, by detecting the rising edge and the falling edge, the falling signal part of the original waveform y(n) is extracted to form TTB y Sequence, extract the rising signal part of the original waveform y(n) to form BTT y Sequence, TTB y Sequence and BTTy The sequence splicing forms the characteristic waveform O(n); the descending signal part of the ideal waveform x(n) is extracted to form TTB x Sequence, extract the rising signal part of the ideal waveform x(n) to form BTT x Sequence, TTB x Sequence and BTT x The sequences are concatenated to form the reference waveform R(n).

[0021] Preferably, in step 8, the formula for calculating waveform W is as follows:

[0022]

[0023] in is the autocorrelation function of the characteristic waveform O(n), is the autocorrelation function of the reference waveform R(n), l is the time shift, and W can be used as the physical fingerprint of the current CAN device signal by numerical compression and smoothing. W can then be smoothed to be the physical fingerprint feature of the device.

[0024] Preferably, the numerical compression process uses the formula: W′=log(|W|), l=1,2,3….

[0025] Compared with the prior art, the present invention has the following advantages:

[0026] 1. Compared with intrusion detection algorithms, it has unique identity authentication and attack tracing mechanisms, which can better implement more comprehensive access control.

[0027] 2. Using a separate rising edge and falling edge signal portion as a feature source weakens the dependence on the data transmission content.

[0028] 3. The device physical fingerprint extraction method has a simple calculation process, is real-time, and does not impose additional calculation load on other devices on the CAN bus.

[0029] 4. There is no need to modify the existing CAN protocol or the existing CAN bus circuit, and it has good backward compatibility.

[0030] 5. The physical fingerprint extraction method in the CAN device signal proposed in the present invention can extract the device physical fingerprint in the CAN signal emitted by the ECU terminal device on the CAN bus, thereby performing identity authentication and attack tracing on the terminal device on the CAN bus. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a schematic diagram of the physical fingerprint extraction process of a CAN device according to the present invention;

[0032] Figure 2This is a schematic diagram of the original waveform collected by the present invention;

[0033] Figure 3 This is a schematic diagram of the clock synchronization result of the present invention;

[0034] Figure 4 This is a schematic diagram of the data domain extraction result of the present invention;

[0035] Figure 5 This is a schematic diagram of the extraction of the descending signal part of the present invention;

[0036] Figure 6 This is a schematic diagram of extracting the rising signal portion of the present invention;

[0037] Figure 7 It is a schematic diagram of the autocorrelation difference between the characteristic waveform and the reference waveform of the present invention;

[0038] Figure 8 Schematic diagram of physical fingerprints of different devices in the present invention. DETAILED DESCRIPTION

[0039] In order to make the technical details and advantages of the present invention clearer, the technical solution is described in more detail below in conjunction with the accompanying drawings. It is particularly noted that the drawings and specific operations used herein are only used to explain the present invention, but are not limited to the present invention.

[0040] like Figure 1 As shown, the invention provides a method for extracting physical fingerprint information from a CAN device signal, comprising the following steps:

[0041] Step 1, data acquisition: collect the data frame signal on the bus at the gateway position on the CAN bus at a sampling rate of 20MHz, the device baud rate B is 1Mbps, store its differential signal, and obtain the original waveform y(n) of the CAN signal. The data acquisition is to collect the differential signal sent by the ECU terminal device on the CAN bus through the data acquisition device, a total of 110 symbols, 20 sampling points for each symbol, a total of 2200 sampling points. As shown in the attached figure Figure 2 The differential signal waveform shown in FIG. 1 is the difference between the CAN_H level value and the CAN_L level value. The use of differential signals can effectively suppress the noise impact caused by a single-channel signal.

[0042] Step 2, locate the starting position of the frame: use the threshold level to determine the starting position of the frame. The threshold level is 1 / 2 of the difference between the high and low levels. According to the CAN protocol requirements, there are at least 7 consecutive low levels between every two frames. This feature can be used to determine the end position of the frame. The collected CAN signal sequence y(n) is extracted through these two positions, and irrelevant data is removed to obtain the original waveform. After removing the last 7 consecutive low levels, the original waveform is 103 symbols, with a total of 2060 sampling points.

[0043] Step 3, clock synchronization: perform clock synchronization on the frame. Starting from the start position of the frame located in the previous step, establish a window sequence with a window size of one symbol length, that is, 20 sampling points. Then use the window sequence to synchronize each symbol of the frame to form a synchronization signal.

[0044] Step 4, resampling: First, reshape the original waveform, then normalize the energy, then use the synchronization signal to resample the original waveform, and finally reconstruct the level of the resampling result according to the decision threshold to obtain the ideal waveform x(n). The ideal waveform is shown in the attached figure. Figure 3 shown.

[0045] Step 5, parse CAN data: parse the reconstructed CAN ideal waveform according to the CAN frame protocol format, and verify whether it is a complete frame.

[0046] Step 6, extract the data domain waveform according to the analysis results: For the collected CAN frame original waveform y(n) and ideal waveform x(n), the data domain refers to all symbol sequences specified in the CAN protocol from the first symbol after the DLC segment to the first symbol before the CRC segment. If there are filling symbols, they should also be counted, totaling 66 symbols and 1320 sampling points. The specific extraction results are shown in the attached figure. Figure 4 shown.

[0047] Step 7, extract the characteristic waveform: extract the rising signal part and the falling signal part of the original waveform y(n) to obtain BTT y and TTB y , spliced ​​to form the characteristic waveform O(n). Similarly, the rising signal part and the falling signal part of the ideal waveform x(n) are extracted to obtain BTT x and TTB x , spliced ​​to form the reference waveform R(n). Specifically, for the falling edge process, it should include the second half cycle of the previous symbol of each falling edge, until the first half cycle of the next symbol. The specific extraction results are shown in the attached Figure 5The rising edge process should include the second half symbol cycle of the previous symbol of each rising edge, until the first half cycle of the next symbol. The specific extraction results are shown in the attached Figure 6 It should be noted that the length of each rising signal part and falling signal part is one symbol length, that is, 20 sampling points. Finally, all the extraction results need to be assembled into a complete sequence, with a total of 520 sampling points and 26 symbols.

[0048] Step 8, extract features: first calculate the autocorrelation function of the feature waveform O(n), then calculate the autocorrelation function of the reference waveform R(n), and then subtract the two to get W. The calculation formula is as follows:

[0049]

[0050] The calculation results of W are shown in the attached Figure 7 After obtaining W, numerical compression and smoothing are performed. The data compression formula is W′=log(|W|), l=1,2,3…. The final physical fingerprint extraction results of different devices are shown in the attached figure. Figure 8 shown.

[0051] The technical means disclosed in the scheme of the present invention are not limited to the technical means disclosed in the above-mentioned implementation mode. For technicians involved in the field of communication and information security technology, equal or similar changes made to the present invention still fall within the protection scope of the present invention.

Claims

1. A method for extracting physical fingerprint information from CAN device signals, characterized in that: The extraction method comprises the following steps: Step 1, data acquisition: collect and store signals on the CAN bus to obtain the CAN signal sequence Y(n) and the original waveform y(n); Step 2, locate the starting position of the frame: perform energy detection on the CAN signal sequence Y(n) to locate the starting position of the frame; Step 3, clock synchronization: starting from the frame start position located in step 2, establish a window sequence, and use the window sequence to synchronize each symbol of the frame to form a synchronization signal; Step 4, resampling: performing shaping and energy normalization processing on the original waveform y(n), then resampling the original waveform y(n) using the synchronization signal, and reconstructing the level of the resampling result according to the decision threshold to obtain the ideal waveform x(n); Step 5, parsing CAN data: according to the CAN frame protocol format, the reconstructed CAN ideal waveform is parsed, and at the same time, it is verified whether it is a complete frame; Step 6, intercepting the data domains of the original waveform y(n) and the ideal waveform x(n) respectively according to the analysis results; Step 7, extract the characteristic waveform: extract the rising signal part and the falling signal part of the original waveform y(n) to obtain BTT y and TTB y , spliced ​​to form the characteristic waveform O(n); extract the rising signal part and the falling signal part of the ideal waveform x(n) to obtain BTT x and TTB x , spliced ​​to form the reference waveform R(n); Step 8, feature extraction: perform autocorrelation operations on the feature waveform O(n) and the reference waveform R(n) respectively, and then calculate the difference to obtain the waveform W. Perform numerical compression and smoothing on the waveform W to obtain the physical fingerprint characteristics of the CAN device signal.

2. The method for extracting physical fingerprint information from CAN device signals according to claim 1, characterized in that: In the step 1, the signal on the bus is collected at the gateway position on the CAN bus. The collected signal is the differential signal between the high data line CAN_H and the low data line CAN_L on the CAN bus. The collected differential signal is stored to obtain the original waveform y(n) and signal sequence Y(n) of the CAN signal.

3. The method for extracting physical fingerprint information from CAN device signals according to claim 1, characterized in that: In step 1, the collected signal is a piece of data with fixed content or data with random content.

4. The method for extracting physical fingerprint information from CAN device signals according to claim 1, characterized in that: In step 2, the energy detection includes determining the start position of the frame using a threshold level, where the threshold level is 1 / 2 of the difference between the high and low levels of the frame.

5. The method for extracting physical fingerprint information from CAN device signals according to claim 1, characterized in that: In step 3, the length of the window sequence is the number of sampling points N used for one symbol, where N is the ratio of the sampling rate of the signal acquisition device to the data transmission rate of the CAN device, wherein the baud rate B of the CAN device and the sampling rate F of the signal acquisition device are known, and the position of each symbol in the signal sequence Y(n) is located according to the number of sampling points N and the starting position of the frame.

6. The method for extracting physical fingerprint information from CAN device signals according to claim 1, characterized in that: In step 4, the decision threshold level is taken from the middle position of each symbol, that is, the level decision result at the position N / 2 is taken as the level of the current symbol.

7. The method for extracting physical fingerprint information from CAN device signals according to claim 1, characterized in that: The data field is taken from a waveform sequence between the first symbol after the DLC segment and the first symbol before the CRC segment in each data frame.

8. The method for extracting physical fingerprint information from CAN device signals according to claim 1, characterized in that: In step 7, by detecting the rising edge and the falling edge, the falling signal part of the original waveform y(n) is extracted to form TTB y Sequence, extract the rising signal part of the original waveform y(n) to form BTT y Sequence, TTB y Sequence and BTT y The sequence splicing forms the characteristic waveform O(n); the descending signal part of the ideal waveform x(n) is extracted to form TTB x Sequence, extract the rising signal part of the ideal waveform x(n) to form BTT x Sequence, TTB x Sequence and BTT x The sequences are concatenated to form the reference waveform R(n).

9. The method for extracting physical fingerprint information from CAN device signals according to claim 1, characterized in that: In step 8, the formula for calculating waveform W is as follows: in is the autocorrelation function of the characteristic waveform O(n), is the autocorrelation function of the reference waveform R(n), l is the time shift, and the waveform W can be used as the physical fingerprint of the current CAN device signal by numerically compressing and smoothing it.

10. The method for extracting physical fingerprint information from CAN device signals according to claim 9, characterized in that: The numerical compression process uses the formula: W′=log(|W|), where the time shift l in the calculation formula for W takes a positive value, and W′ is the value after data compression.