Wireless HART (Highway Addressable Remote Transducer) equipment security access method based on radio frequency fingerprint
Through the dual-branch convolutional neural network model based on radio frequency fingerprint, the problems of security threats and resource consumption in wireless communications of WirelessHART devices are solved, efficient identity recognition and secure access are achieved, and the security and real-time nature of the system are improved.
Patent Information
- Application Number
- CN202411526829.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-30
- Publication Date
- 2025-01-07
AI Technical Summary
WirelessHART devices have security threats in wireless communication, such as device disguise and malicious access. Traditional secure access methods rely on complex cryptographic algorithms, resulting in resource consumption and real-time problems, making it difficult to meet the rapidly changing industrial environment needs.
The dual-branch convolutional neural network model based on radio frequency fingerprint is adopted to extract and identify wireless radio frequency fingerprint information to achieve efficient identity recognition and secure access to the device, and use physical layer features to avoid software counterfeiting, simplify the key management and authentication process.
It improves the security and real-time nature of the system, reduces the consumption of equipment resources, ensures low power consumption and efficient identity recognition, and enhances network security and identification accuracy.
Smart Images
Figure CN119277393A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of cyberspace security technology, and in particular to a WirelessHART device security access method based on radio frequency fingerprint. Background Art
[0002] In wireless industrial control systems, the WirelessHART (Highway Addressable Remote Transducer) protocol, as a technology based on industrial wireless communication standards, is widely used in the field of industrial automation. With the continuous development of wireless communication technology, more and more industrial equipment has begun to use wireless methods for data transmission. However, due to the openness and complexity of wireless communication, it is vulnerable to network attacks, such as device spoofing, man-in-the-middle attacks, and malicious device access. These security threats may not only lead to data leakage and tampering of industrial control systems, but also affect the normal operation of the system, causing serious economic losses and safety hazards.
[0003] The current status of WirelessHART device access methods is:
[0004] 1) Traditional WirelessHART device security access methods mostly rely on cryptographic algorithms and key management mechanisms, and key management solutions involve key generation, distribution, storage, and update steps, which require complex management mechanisms. In addition, they require additional hardware and software support, increasing the cost and difficulty of system maintenance;
[0005] 2) WirelessHART devices are usually resource-constrained, such as sensor nodes, which usually have limited computing power and storage space. Running complex cryptographic algorithms on resource-constrained devices will lead to performance degradation, increased power consumption, or even failure to work properly;
[0006] 3) Traditional authentication methods usually require multiple interactions, such as key negotiation and identity authentication, which makes the authentication process time-consuming and reduces the real-time performance of the system. In some application scenarios, such as rapidly changing environments or applications with high real-time requirements, traditional authentication methods are difficult to meet the needs. Summary of the invention
[0007] Purpose of the invention: The purpose of the present invention is to provide a secure access method for WirelessHART devices based on radio frequency fingerprints, to establish an accurate and efficient identity recognition method according to the actual WirelessHART scenario, and to achieve secure access authentication for node devices in the network.
[0008] Technical solution: A secure access method for WirelessHART devices based on radio frequency fingerprint, comprising the following steps:
[0009] S1, extract multiple wireless radio frequency fingerprints for each legitimate device and establish a legitimate device fingerprint library; use the legitimate device fingerprint library to train the constructed dual-branch convolutional neural network model and save the best dual-branch convolutional neural network model;
[0010] S2, the network access point extracts the wireless radio frequency fingerprint information according to the network access request signal sent by the device to access the network;
[0011] S3, the network access point forwards the obtained RF fingerprint information to the WirelessHART network manager through routing;
[0012] S4, the WirelessHART network manager uses a dual-branch convolutional neural network model to identify the received RF fingerprint and obtain a legal device number or determine it as an illegal device;
[0013] S5, the WirelessHART network manager decides whether to allow the device to access the network based on the identification result, and returns a network access response to the device that wishes to access the network.
[0014] Furthermore, the implementation steps for establishing a legal device fingerprint library are as follows:
[0015] S11, collecting the wireless radio frequency signal data frame sent by the WirelessHART device and performing energy normalization processing;
[0016] S12, for the data frame after energy normalization processing, intercepting its 8-byte preamble code RF signal time domain amplitude image as a fingerprint, and the time domain amplitude of each preamble code includes I / Q two-channel RF information;
[0017] S13, executing according to steps S11 to S12, extracting fingerprints for each legal device multiple times, and establishing a legal device fingerprint library containing fingerprint features.
[0018] Further, in step S11, energy normalization processing is performed on the signal data frame, and the expression is as follows:
[0019]
[0020] Wherein, s(t) represents the original sampling point signal amplitude at time t, and z represents the root mean square of the amplitude values of all sampling points of the RF signal.
[0021] Furthermore, the dual-branch convolutional neural network includes two convolution branches, which are used to input the I / Q signals in the fingerprint respectively. Each convolution branch includes two convolution layers with ReLU activation functions; a pooling layer is executed after each convolution layer to perform downsampling; and both convolution branches use the Flat layer to convert multi-dimensional nodes into one-dimensional nodes as the input of the fully connected layer, and finally output the predicted device number through the Softmax layer.
[0022] Further, in step S2, after the network access point device receives the network access request signal sent by the device to access the network, it performs energy normalization processing on the network access request signal and extracts the 8-byte preamble code time domain amplitude image as the radio frequency fingerprint information of the device to access the network.
[0023] Furthermore, in step S3, the routing forwarding refers to forming a multi-hop forwarding path through other joined devices or access points in the network.
[0024] Further, in step S5, the WirelessHART network manager returns a network access response signal according to the fingerprint determination result of the dual-branch convolutional network model:
[0025] If the result is a legitimate device, the response signal contains the device nickname and session key for initial configuration of the device to be added to the network;
[0026] If it is an illegal device, the illegal device flag in the response signal is 1, indicating that the device is an illegal device; if it needs to join the network, its RF fingerprint should be registered in the legal device fingerprint database first.
[0027] Compared with the prior art, the present invention has the following significant effects:
[0028] 1. The present invention improves the traditional WirelessHART device security access method, overcomes the complexity and high cost problems caused by its multiple reliance on cryptographic algorithms and key management mechanisms, and utilizes the uniqueness and non-clonability of the radio frequency fingerprint of wireless devices in the WirelessHART network to achieve efficient identity recognition of access devices from the perspective of the physical layer, thereby enhancing the security and real-time performance of the system;
[0029] 2. The present invention improves the existing WirelessHART device security authentication technology, and overcomes the performance degradation and power consumption increase problems faced by resource-constrained devices when running complex cryptographic algorithms. It uses convenient preamble codes as RF fingerprint information to quickly collect and extract fingerprint data, avoiding excessive consumption of device resources by complex algorithms, and ensuring that the device can achieve effective identity recognition and secure communication while maintaining low power consumption and efficient operation.
[0030] 3. The present invention improves the existing fingerprint data processing technology and overcomes the problems of insufficient fusion and low processing efficiency that may exist when processing fingerprint I / Q two-way signal data. The dual-branch convolutional neural network structure of the present invention can make full use of the complementarity of the I / Q two-way signals, and efficiently fuse the signals through a deep learning algorithm, thereby improving the accuracy and efficiency of fingerprint recognition. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 It is the overall flow chart of the present invention;
[0032] Figure 2 This is the architecture diagram of the convolutional neural network model;
[0033] Figure 3 This is a schematic diagram of the device access process. DETAILED DESCRIPTION
[0034] Radio Frequency Fingerprint (RFF) technology is based on the physical layer characteristics generated by wireless devices during communication. These physical layer characteristics are determined by the hardware differences of the devices and are difficult to counterfeit through software means. Therefore, RF fingerprint is considered to be an efficient device identity authentication technology that can enhance the device's identity recognition capability without adding additional computing burden. In this context, the present invention proposes a WirelessHART device secure access method based on RF fingerprint.
[0035] In order to more intuitively demonstrate the technical solutions and advantages of the present invention, the specific implementation modes of the present invention will be described in detail below with reference to the accompanying drawings.
[0036] A secure access method for WirelessHART devices based on RF fingerprints. The overall flow chart is as follows: Figure 1 As shown, the following steps are included:
[0037] Step 1: extract multiple wireless radio frequency signal data frames for each legitimate device and establish a legitimate device fingerprint library for training a dual-branch convolutional neural network.
[0038] First, the wireless RF signal data frames sent by the WirelessHART device are collected and energy normalized. The signal acquisition tool is the USRP software radio platform, and the signal sampling frequency is 10MS / s; the energy normalization process aims to eliminate the interference that may be caused by the distance and transmission power differences between the transceiver devices, so that signals of different strengths reach a consistent level in energy, thereby improving the accuracy and reliability of subsequent identification. The specific implementation method of energy normalization is to divide the amplitude of each sampling point of the data frame by the root mean square of the amplitude values of all sampling points of the RF signal. The energy normalization formula is:
[0039]
[0040] Wherein, s(t) represents the original sampling point signal amplitude at time t, and z represents the root mean square of the amplitude values of all sampling points of the RF signal.
[0041] Then, for the energy-normalized data frame, the 8-byte preamble RF signal time domain amplitude image is intercepted as the fingerprint. The time domain amplitude of each preamble includes I / Q two-way RF information. According to this method, fingerprints are extracted multiple times for each legitimate device. For example, 100 fingerprints are extracted 100 times for each of 100 devices, and a legal device fingerprint library containing 10,000 fingerprint features can be established.
[0042] like Figure 2 As shown, a dual-branch convolutional neural network model specifically for classification and recognition is established. The dual-branch convolutional neural network model includes two convolution branches, which are used to input the I / Q signals in the fingerprint respectively; each convolution branch includes two convolution layers with ReLU activation function, and the convolution kernel size is 3*3, and the number is 16 and 32 respectively; a pooling layer is executed after each convolution layer for downsampling, and the pooling window is 2*2; and both convolution branches use the Flat layer to convert multi-dimensional nodes into one-dimensional nodes as the input of the fully connected layer D1. The fully connected layer D1 uses 256 nodes and outputs the predicted device number through the Softmax layer. Finally, the dual-branch convolutional neural network model is trained using the legal device fingerprint library established in this implementation. After the training is completed, the best model is saved. The best model will be able to determine the device number for the known RF fingerprint, or output the illegal device category.
[0043] like Figure 3 As shown, the following steps describe in detail the device access and authentication process.
[0044] Step 2: The network access point extracts the wireless radio frequency fingerprint information according to the network access request signal sent by the device that wants to access the network.
[0045] After receiving the network access request signal sent by the device to access the network, the network access point device performs energy normalization processing on the network access request signal according to formula (1), and extracts the 8-byte preamble code time domain amplitude image as the radio frequency fingerprint information of the device to access the network.
[0046] Step 3: The network access point forwards the obtained RF fingerprint information to the WirelessHART network manager through routing.
[0047] After the network access point device completes the fingerprint extraction of the device to be connected to the network, it forwards the RF fingerprint information to the WirelessHART network manager through the established WirelessHART topology network. The WirelessHART network manager decides whether to allow the device to join the network.
[0048] Routing forwarding refers to forming a multi-hop forwarding path through other joined devices or access points in the network.
[0049] In step 4, the WirelessHART network manager uses a dual-branch convolutional neural network model to identify the received RF fingerprint and obtain a legal device number or determine it as an illegal device.
[0050] After the WirelessHART network manager receives the fingerprint information sent by the access point device, it uses the time domain amplitude images of the I / Q signals in the fingerprint as the input of the dual-branch convolutional network model. The dual-branch convolutional network model will output the device number corresponding to the fingerprint information or judge it as an illegal device.
[0051] Step 5: The WirelessHART network manager decides whether to allow the device to access the network based on the identification result, and returns a network access response to the device that wishes to access the network.
[0052] The WirelessHART network manager returns a network access response signal based on the fingerprint determination result of the dual-branch convolutional network model:
[0053] If the result is a legitimate device, the response signal contains the device nickname and session key for initial configuration of the device to be added to the network;
[0054] If it is an illegal device, the illegal device flag in the response signal is 1, indicating that the device is an illegal device. If it needs to join the network, its RF fingerprint should be registered in the legal device fingerprint database first.
Claims
1. A secure access method for WirelessHART devices based on radio frequency fingerprint, characterized in that: The steps include: S1, extract multiple wireless radio frequency fingerprints for each legitimate device and establish a legitimate device fingerprint library; use the legitimate device fingerprint library to train the constructed dual-branch convolutional neural network model and save the best dual-branch convolutional neural network model; S2, the network access point extracts the wireless radio frequency fingerprint information according to the network access request signal sent by the device to access the network; S3, the network access point forwards the obtained RF fingerprint information to the WirelessHART network manager through routing; S4, the WirelessHART network manager uses a dual-branch convolutional neural network model to identify the received RF fingerprint and obtain a legal device number or determine it as an illegal device; S5, the WirelessHART network manager decides whether to allow the device to access the network based on the identification result, and returns a network access response to the device that wishes to access the network.
2. The secure access method for WirelessHART devices based on radio frequency fingerprint according to claim 1, characterized in that: The steps to establish a legal device fingerprint database are as follows: S11, collecting the wireless radio frequency signal data frame sent by the WirelessHART device and performing energy normalization processing; S12, for the data frame after energy normalization processing, intercepting its 8-byte preamble code RF signal time domain amplitude image as a fingerprint, and the time domain amplitude of each preamble code includes I / Q two-channel RF information; S13, executing according to steps S11 to S12, extracting fingerprints for each legal device multiple times, and establishing a legal device fingerprint library containing fingerprint features.
3. The secure access method for WirelessHART devices based on radio frequency fingerprint according to claim 2 is characterized in that: In step S11, energy normalization processing is performed on the signal data frame, and the expression is as follows: Wherein, s(t) represents the original sampling point signal amplitude at time t, and z represents the root mean square of the amplitude values of all sampling points of the RF signal.
4. The WirelessHART device secure access method based on radio frequency fingerprint according to claim 1 is characterized in that: The dual-branch convolutional neural network includes two convolution branches, which are used to input the I / Q signals in the fingerprint respectively. Each convolution branch includes two convolution layers with ReLU activation function; a pooling layer is executed after each convolution layer to perform downsampling; and both convolution branches use the Flat layer to convert multi-dimensional nodes into one-dimensional nodes as the input of the fully connected layer, and finally output the predicted device number through the Softmax layer.
5. The secure access method for WirelessHART devices based on radio frequency fingerprint according to claim 1, characterized in that: In step S2, after the network access point device receives the network access request signal sent by the device to access the network, it performs energy normalization processing on the network access request signal and extracts the 8-byte preamble code time domain amplitude image as the radio frequency fingerprint information of the device to access the network.
6. The secure access method for WirelessHART devices based on radio frequency fingerprint according to claim 1, characterized in that: In step S3, the routing forwarding refers to forming a multi-hop forwarding path through other joined devices or access points in the network.
7. The secure access method for WirelessHART devices based on radio frequency fingerprint according to claim 1, characterized in that: In step S5, the WirelessHART network manager returns a network access response signal based on the fingerprint determination result of the dual-branch convolutional network model: If the result is a legitimate device, the response signal contains the device nickname and session key for initial configuration of the device to be added to the network; If it is an illegal device, the illegal device flag in the response signal is 1, indicating that the device is an illegal device; if it needs to join the network, its RF fingerprint should be registered in the legal device fingerprint database first.