Power local security communication method
Through the RF fingerprint comparison and authentication of distributed gateway systems and neural networks, the problem of tampering in traditional wireless communication systems is solved, and efficient security authentication and illegal device identification of wireless WiFi devices are achieved, which significantly improves the security of the system. sex.
Patent Information
- Application Number
- CN202310678505.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-08
- Publication Date
- 2024-12-17
AI Technical Summary
The identity authentication method of traditional wireless communication systems is easily tampered with, and the device MAC address based on the OSI upper layer protocol is difficult to ensure the integrity of the communication link, and there is a hidden danger of illegal device access.
A distributed gateway system is adopted to slide the received signal through a WiFi local gateway, calculate the radio frequency fingerprint, and use a neural network to compare and authenticate the device RF fingerprint, and perform device authentication and abnormal device blocking step by step to ensure system security.
Effectively identifying and blocking illegal access devices, improving the security of wireless WiFi device access, enhancing the protection function of the system, and avoiding loopholes in traditional methods.
Smart Images
Figure CN119155678A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wireless communications, and in particular to a method for local safe power communication. Background Art
[0002] Due to its openness, wireless communication systems are more vulnerable to malicious attacks than traditional wired communication systems. In view of the hidden dangers of wireless communication environments, the classic method to solve the problem of air eavesdropping is to encrypt air data. An effective way to solve the problem of illegal device access is to authenticate the device. Traditional encryption and authentication methods for protecting the security of wireless communication systems are based on the physical layer above the OSI seven-layer model. By designing a security protocol based on an encryption mechanism, the integrity and confidentiality of the data are protected, and identity authentication is provided for both parties in communication. However, actual wireless network security protocols often have loopholes. At present, the identity authentication of wireless communication systems mainly relies on the identity authentication of devices or identity authentication through protocols. For example, the control device authenticates the access device through the MAC address of the access device or the pre-shared key. However, the MAC address of the device based on the OSI upper layer protocol is easy to be tampered with, and the identity authentication based on the pre-shared key requires additional communication channel support.
[0003] "Wireless device fingerprint" and "wireless device fingerprint identification" are new concepts proposed by J. Hall et al. in Canada in 2003 in the research on wireless network device identification such as Bluetooth. Since then, this concept has mainly appeared in the research on wireless network device identification. Technologies related to "wireless device fingerprint identification" include specific radiation source identification technologies such as "radar radiation source identification" and "communication radio station individual identification", which can be included in the research scope based on physical unclonable functions.
[0004] The traditional security access gateway is a multifunctional device that not only has the function of a network firewall, but also can realize intrusion monitoring, VPN, data packet identification and forwarding, etc. Therefore, on the basis of the traditional security access gateway function, adding wireless physical layer fingerprint recognition technology can greatly enhance the protection function of the access gateway. The present invention proposes a local power security communication method based on radio frequency fingerprint technology. Summary of the invention
[0005] The purpose of the present invention is to provide a local safety communication method for electric power, which greatly improves the safety of the system and effectively identifies illegal access devices.
[0006] The technical solution to achieve the purpose of the present invention is: a method for local power security communication, which adopts distributed access gateway deployment, and is provided with a core network side gateway, an edge side gateway and a WiFi local gateway, wherein the WiFi local gateway and the edge side gateway use the radio frequency fingerprint of the device as the access basis, and perform device authentication step by step; and can realize the blocking operation of abnormal devices, and the specific steps are as follows:
[0007] Step 1: The WiFi local gateway performs sliding correlation on the received signal and calculates the correlation value. When a correlation peak is obtained, it is considered that the WiFi signal is received, and the device physical fingerprint is obtained through the WiFi training sequence;
[0008] Step 2: Obtain the device MAC address by decoding the data;
[0009] Step 3: Use a neural network to compare and authenticate the device's radio frequency fingerprint, and send the MAC address and fingerprint information of the abnormal device to the edge gateway through the 4G terminal, and then to the gateway on the core network side.
[0010] Furthermore, when the received signal is conjugate correlated in step 1 and a correlation peak is obtained, it is considered that the operation of receiving the WiFi signal is completed on an independent listening device, rather than being modified on a traditional WiFi access node. Therefore, the independent device can listen to the physical layer signals of devices in multiple WiFi access points.
[0011] Furthermore, the physical fingerprint of the device obtained through the training sequence in step 1 refers to the characteristic parameters of the RF signal caused by the difference in physical components of the device, including frequency deviation and IQ imbalance.
[0012] Furthermore, obtaining the device MAC address in step 2 specifically refers to the independent listening device decoding the MAC address using the received information instead of using the MAC information of the WiFi access point, which can ensure that the original transmission link is not changed.
[0013] Furthermore, the neural network structure in step 3 is as follows:
[0014] Using the Keras framework, the stochastic gradient descent method was adopted, and the Adam adaptive learning rate algorithm and regularization method were introduced to build a CNN neural network; the CNN neural network has 3 convolutional layers followed by 1 pooling layer, which is looped 4 times and then followed by 3 fully connected layers, and the activation function uses the softmax function.
[0015] Compared with the prior art, the present invention has the following significant advantages: (1) when the received signal is conjugate correlated and a correlation peak is obtained, it is considered that a WiFi signal has been received, and the physical fingerprint of the device is obtained through a training sequence; (2) the MAC address of the device is obtained by decoding the received data, and fingerprint comparison is performed locally on the WiFi. A specific neural network is used to compare feature values, and the MAC address of the abnormal device and its fingerprint information are sent to the edge gateway through a 4G terminal; (3) the device is identified by a physical radio frequency fingerprint authentication method, and a neural network is used for fingerprint comparison, which can effectively identify illegal access devices and improve the security of access to power WiFi devices. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 This is the implementation scenario of the local safety communication method for electric power of the present invention.
[0017] Figure 2 It is a working flow chart of the local safety communication method of electric power of the present invention.
[0018] Figure 3 It is a neural network model used for radio frequency fingerprint comparison in the local power safety communication method of the present invention. DETAILED DESCRIPTION
[0019] The present invention provides a method for local power security communication. The gateway system is deployed in a distributed manner and includes a core network side gateway, an edge side gateway and a WiFi local gateway. These gateways use the radio frequency fingerprint of the device as the authentication basis, perform device authentication step by step, and can realize the blocking operation of abnormal devices. The specific steps are as follows:
[0020] Step 1: Conjugate correlation is performed on the received signal. When a correlation peak is obtained, it is considered that a WiFi signal is received, and the physical fingerprint of the device is obtained through the training sequence;
[0021] Step 2: Obtain the device MAC address by decoding the received data;
[0022] Step 3: Use neural network to perform fingerprint comparison at the WiFi local gateway, send the MAC address of the abnormal device and its fingerprint information to the edge gateway through the 4G terminal, and then send the abnormal information to the gateway on the core network side for full network decision-making.
[0023] Furthermore, when the correlation peak is obtained by conjugate correlation of the received signal in step 1, it is considered that the operation of receiving the WiFi signal is completed on an independent listening device, rather than being modified on a traditional WiFi access node. Therefore, the independent device can listen to the physical layer signals of devices in multiple WiFi access points.
[0024] Furthermore, the physical fingerprint of the device obtained through the training sequence in step 1 refers to characteristic parameters of the radio frequency signal caused by differences in physical components of the device, such as frequency deviation, IQ imbalance, etc.
[0025] Furthermore, the obtaining of the device MAC address in step 2 specifically refers to an independent listening device decoding the MAC address using the received information instead of using the MAC information of the WiFi access point, which ensures that no change is made to the original transmission link.
[0026] Furthermore, the neural network structure described in step 3 is as follows:
[0027] Using the Keras framework, the stochastic gradient descent method, the Adam adaptive learning rate algorithm and the regularization method are introduced to build a CNN neural network. The CNN neural network has three convolutional layers followed by one pooling layer, four cycles followed by three fully connected layers, and the activation function uses the softmax function.
[0028] The first few layers of the CNN model mainly use convolution and pooling layers, while the last three layers use a fully connected approach, and the BP algorithm is used for parameter and weight adjustment. The specific description of the algorithm is as follows:
[0029] (1) Convolutional layer: In a convolutional layer, the feature map is obtained by convolving several different convolution kernels with the output feature map of the previous layer, and then passing through an activation function. f(·) represents the activation function. Each output map may be the weighted sum of multiple input maps convolved:
[0030]
[0031] (2) Downsampling layer sampling: The function of the downsampling layer is to reduce the dimension of the output map of the previous layer connected to it by a certain ratio, but does not change the number of output feature maps. Its formula is:
[0032]
[0033] where down(·) represents a downsampling function that reduces the output image by n 2 The activation function here is the same as that of other layers.
[0034] (3) Back propagation BP algorithm adjustment parameters:
[0035] (a) Forward propagation: For multi-class problems, there are C classes and N training samples. The cost function is expressed as a square error function, that is:
[0036]
[0037]
[0038] in, represents the ideal label value of the nth graph in the kth class sample, Indicates the network output label value corresponding to the input graph.
[0039] (b) Back propagation: The back propagation error is the partial derivative of the cost function with respect to the weight W and the bias b. The parameter update definition is as follows, where α is the learning rate,
[0040]
[0041]
[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 are within the scope of protection of the present invention.
[0043] Example
[0044] Combination Figure 1 to Figure 3 The electric power local security communication method of the present invention, the gateway system is distributedly deployed, including a core network side gateway, an edge side gateway and a WiFi local gateway. These gateways use the radio frequency fingerprint of the device as the authentication basis, perform device authentication step by step, and can realize the blocking operation of abnormal devices. The specific steps are as follows:
[0045] Step 1: Conjugate correlation is performed on the received signal. When a correlation peak is obtained, it is considered that a WiFi signal has been received, and the physical fingerprint of the device is obtained through the training sequence. This operation is completed on an independent listening device, not on a traditional WiFi access node. Therefore, the independent device can listen to the physical layer signals of devices in multiple WiFi access points. If no correlation peak is detected, prepare to enter the next cycle of correlation peak detection.
[0046] Step 2: Obtain the device MAC address by decoding the received data. The MAC address is a MAC address decoded by an independent listening device using the received information instead of the MAC information of the WiFi access point. This ensures that no changes are made to the original transmission link.
[0047] Step 3: Use the MAC address obtained in step 2 to determine the corresponding fingerprint in the local fingerprint database, use a neural network to compare and authenticate the device's RF fingerprint, and send the MAC address of the abnormal device and its fingerprint information to the edge gateway through the 4G terminal, and then to the gateway on the core network side. Figure 3As shown in the figure, the neural network uses the Keras framework, adopts the stochastic gradient descent method, introduces the Adam adaptive learning rate algorithm and the regularization method, and builds a CNN neural network. The CNN neural network has 3 convolutional layers followed by 1 pooling layer, which is looped 4 times and then followed by 3 fully connected layers. The activation function uses the softmax function.
[0048] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A method for local power safety communication, characterized in that: The method adopts distributed deployment of access gateways, and is equipped with a core network side gateway, an edge side gateway and a WiFi local gateway. The WiFi local gateway and the edge side gateway use the radio frequency fingerprint of the device as the access basis to perform device authentication step by step; and can realize the blocking operation of abnormal devices. The specific steps are as follows: Step 1: The WiFi local gateway performs sliding correlation on the received signal and calculates the correlation value. When a correlation peak is obtained, it is considered that the WiFi signal is received, and the device physical fingerprint is obtained through the WiFi training sequence; Step 2: Obtain the device MAC address by decoding the data; Step 3: Use a neural network to compare and authenticate the device's radio frequency fingerprint, and send the MAC address and fingerprint information of the abnormal device to the edge gateway through the 4G terminal, and then to the gateway on the core network side.
2. The method for local power safety communication according to claim 1, characterized in that: In step 1, when the received signal is conjugate correlated and a correlation peak is obtained, it is considered that the operation of receiving the WiFi signal is completed on an independent listening device, and the independent listening device can listen to the physical layer signals of devices in multiple WiFi access points.
3. The method for local power safety communication according to claim 2, characterized in that: The physical fingerprint of the device obtained through the training sequence in step 1 refers to the characteristic parameters of the RF signal caused by the differences in the physical components of the device, including frequency deviation and IQ imbalance.
4. The method for local power safety communication according to claim 1, characterized in that: Obtaining the device MAC address in step 2 specifically refers to the independent listening device decoding the MAC address using the received information to ensure that the original transmission link is not changed in any way.
5. The method for local power safety communication according to claim 1, characterized in that: The neural network structure in step 3 is as follows: Using the Keras framework, the stochastic gradient descent method was adopted, and the Adam adaptive learning rate algorithm and regularization method were introduced to build a CNN neural network; the CNN neural network has 3 convolutional layers followed by 1 pooling layer, which is looped 4 times and then followed by 3 fully connected layers, and the activation function uses the softmax function.