A physical layer eavesdropping location method for software-defined optical networks

By using the BP neural network model to perform eavesdropping attack positioning in the fiber optic communication system, the problem of low positioning accuracy and large event blind spots in the fiber optic communication system in the eavesdropping attack is solved, more efficient eavesdropping detection and rapid response are achieved, and the optical network is enhanced security.

CN117478237BActive Publication Date: 2024-11-08CHONGQING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311525639.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-15
Publication Date
2024-11-08
Estimated Expiration
2043-11-15

AI Technical Summary

Technical Problem

Fiber optical communication systems, especially high-speed coherent optical communication systems, are difficult to quickly locate and process when faced with eavesdropping attacks, resulting in security challenges.

Method used

The eavesdropping attack positioning model based on BP neural network is adopted to locate the eavesdropping location through photoelectric conversion data, and send it to the network security administrator in the form of an alarm to quickly handle the network attack. This model simulates eavesdropping attacks, trains eavesdropping attack positioning models, and uses Optical Network Controller in SDN for signal transmission monitoring.

Benefits of technology

It realizes more efficient eavesdropping positioning, can detect trace signal leakage, reduce event blind spots, improve real-time and accuracy of detection, and is suitable for high-speed and complex coherent communication systems, enhancing the security of optical networks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117478237B_ABST
    Figure CN117478237B_ABST
Patent Text Reader

Abstract

The present invention belongs to the field of communication technology, and specifically relates to a physical layer eavesdropping positioning method for software-defined optical networks. The method utilizes the different influence mechanisms of eavesdropping attacks at different positions on transmission signals in the physical layer. After a normal signal is subjected to eavesdropping attacks at different positions in an optical fiber link, a signal after photoelectric conversion at a receiving end will undergo subtle changes to varying degrees. This spectral feature inputs the collected signal data into an eavesdropping positioning model based on a BP neural network to perform eavesdropping positioning. Since the photoelectrically converted data is used, no complex DSP processing is required, so the method has a simple model structure, high positioning accuracy, and low model cost.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of communication technology, and in particular relates to a physical layer eavesdropping positioning method for a software-defined optical network. Background Art

[0002] Optical fiber is an insulating medium that can effectively resist electromagnetic interference. Therefore, it is considered a safe transmission medium and is widely used in backbone networks and access networks. At present, optical networks based on optical fiber infrastructure carry more than 90% of the world's Internet data transmission. The emergence of a variety of hidden and advanced technical means, such as bending optical fiber, V-groove cutting, grating scattering, splitting and evanescent wave coupling, has made optical fiber transmission no longer as safe as it was once thought. Although these means will only cause a small part of the optical signal loss and will not affect the communication quality, it is difficult for both the server and the client to detect it. However, even if the above means only output 1% of the leaked light, the original signal can be recovered from it. In addition, the transmission rate of optical signals in optical fibers is very high. Even a short-term eavesdropping on the optical fiber will result in a large amount of data leakage. Therefore, the security of optical fiber communication systems, especially coherent optical communication systems with higher rates, is greatly challenged. Summary of the invention

[0003] To solve the above problems, the present invention provides a physical layer eavesdropping locating method for software-defined optical networks. The method defines an optical network physical layer eavesdropping locating network system, constructs and trains an eavesdropping attack locating model based on a BP neural network. When an Optical Network Controller detects that there is eavesdropping in the optical network physical layer eavesdropping locating network system, the photoelectric conversion data of the optical network physical layer eavesdropping locating network system is obtained and input into the eavesdropping attack locating model to locate the eavesdropping position, and sends it to a network security administrator in the form of an alarm trigger, thereby quickly handling the network attack.

[0004] The training process of the eavesdropping attack location model includes the following steps:

[0005] S1. Building a coherent optical communication system, the coherent optical communication system includes a coherent transmitter, a coherent receiver and an optical fiber link with a length of 100km;

[0006] S2. The coherent transmitter outputs a 200G-16QAM dual-polarization signal to the optical fiber link, and uses an optical splitter in the optical fiber link to simulate different degrees of eavesdropping attacks;

[0007] S3. The coherent receiver collects a piece of photoelectric conversion data every fixed time window, and obtains multiple pieces of photoelectric conversion data to form a data set;

[0008] S4. preprocess the data set, and divide the preprocessed data set into a training set, a validation set, and a test set in a ratio of 3:1:1;

[0009] S5. Use the training set to train the eavesdropping attack localization model, calculate the loss value and back propagate until the model converges or reaches the maximum number of iterations.

[0010] Furthermore, the optical network physical layer eavesdropping and positioning network system includes a transmitting end, a receiving end, an optical module, an OXC and an optical network; the working process of the optical network physical layer eavesdropping and positioning network system is as follows: the transmitting end signal enters the optical network through the optical module and the OXC uplink, and then reaches the receiving end through the OXC and the optical module downlink, and the receiving end performs DSP processing on the received signal; wherein the Optical Network Controller in the SDN is used to perform signal transmission monitoring on the optical network physical layer eavesdropping and positioning network system.

[0011] Furthermore, in the coherent optical communication system, the coherent transmitter outputs a 200G-16QAM dual-polarization signal to a 100km optical fiber link, and the coherent receiver performs digital signal processing on the received data to make the BER zero, thereby ensuring the transmission effectiveness of the coherent optical communication system.

[0012] Furthermore, the specific process of data collection includes:

[0013] S21. Starting from the left end of the 100 km optical fiber link, mark an eavesdropping position at 0 km, 10 km, 20 km, 30 km, 40 km, 50 km, 60 km, 70 km, 80 km and 90 km to the right;

[0014] S22. Use a 1% optical splitter to simulate an attacker at the i=1, 2, ..., 10th eavesdropping position, then use a coherent transmitter to output a 200G-16QAM dual-polarization signal to the optical fiber link, and the coherent receiver collects the signal and performs photoelectric conversion. A fixed time window is used to divide the photoelectric conversion data into multiple segments and sample and store them at intervals; each segment of photoelectric conversion data includes four signals: XI, XQ, YI, and YQ (X-pol In-phase, X-pol Quadrature, Y-pol In-phase, and Y-pol Quadrature);

[0015] S23. Use a 10% optical splitter to simulate an attacker at the i=1, 2, …, 10th eavesdropping position, then use a coherent transmitter to output a 200G-16QAM dual-polarization signal to the optical fiber link, and the coherent receiver collects the signal and performs photoelectric conversion. A fixed time window is used to divide the photoelectric conversion data into multiple segments and sample and store them at intervals; each segment of photoelectric conversion data includes four signals: XI, XQ, YI, and YQ;

[0016] S24. Combining multiple segments of photoelectric conversion data collected by the 1% splitter and the 10% splitter at different eavesdropping positions into a data set.

[0017] Furthermore, the optical fiber link is composed of polarization multiplexing standard single-mode optical fiber and EDFA.

[0018] Furthermore, an optical bandpass filter is provided at the coherent receiver, and the signal transmitted via the optical fiber link is first processed by the optical bandpass filter, and then the processed signal is transmitted to the coherent receiver.

[0019] Furthermore, step S4 preprocesses the data set, including:

[0020] S41. Extract the signal amplitude, signal frequency, signal phase, signal rise time, signal fall time, signal period, signal upper and lower limits, signal peak value and signal waveform of each signal in each section of photoelectric conversion data in the data set to form a dimensional data, and form a set of sample data with the dimensional data corresponding to the four signals in each section of photoelectric conversion data; perform data cleaning on all sample data to remove sample data with obvious errors;

[0021] S42. Normalizing and standardizing the cleaned sample data set to obtain a standard data set;

[0022] S43. According to the eavesdropping attack location, each set of standard sample data in the standard data set is labeled accordingly;

[0023] S44. After labeling, all standard sample data are shuffled to obtain the preprocessed data set.

[0024] Furthermore, the eavesdropping attack location detection model based on the BP neural network includes an input layer, four hidden layers and an output layer; the number of nodes in the input layer is 8192, the number of nodes in the four hidden layers are 4096, 2048, 512 and 256 respectively, and the number of nodes in the output layer is 10.

[0025] Beneficial effects of the present invention:

[0026] Compared with the traditional method of using OTDR (Optical Time-Domain Reflectometer) to achieve eavesdropping location, the method of eavesdropping location proposed in the present invention through data after photoelectric conversion has the advantages of small detection blind area, simultaneous monitoring and communication, and the ability to detect trace signal leakage. This is a more efficient detection method based on the existing optical network infrastructure.

[0027] Compared with the existing methods of realizing eavesdropping location through artificial intelligence algorithms, the method proposed in the present invention uses the signal after photoelectric conversion as the input of the eavesdropping attack location detection model, which does not require expensive equipment or complex DSP processing, and the input acquisition is simple, the detection is highly real-time, and it is convenient for large-scale deployment in the network, and can be applied to the current high-speed and complex coherent communication systems. The present invention uses a neural network as a detection model, verifies the feasibility of the signal after photoelectric conversion as an eavesdropping attack location indicator, and the BP neural network is applicable to eavesdropping attack location, ensuring faster detection of attacks while having a high accuracy rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 Positioning network framework for macroscopic optical network physical layer eavesdropping attacks of the present invention;

[0029] Figure 2 A 200G-16QAM dual-polarization coherent optical communication system for implementing eavesdropping attacks in the present invention;

[0030] Figure 3 It is a signal diagram after photoelectric conversion at the receiving end at different eavesdropping positions of the present invention;

[0031] Figure 4 It is a fiber optic eavesdropping attack positioning model based on BP neural network in the coherent communication system of the present invention;

[0032] Figure 5 This is a graph showing how the positioning accuracy and loss on the validation set change with the number of iterations during the training process of the present invention;

[0033] Figure 6 This is a flow chart of the eavesdropping attack positioning model training of the present invention. DETAILED DESCRIPTION

[0034] 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 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 are within the scope of protection of the present invention.

[0035] Traditional eavesdropping positioning mainly uses OTDR. OTDR uses Rayleigh scattering and Fresnel reflection to analyze the fiber link by measuring the reflected signal intensity and return time of the optical pulse. It can detect signal distortion caused by pressure, birefringence and other factors, so it can identify fiber bending, breakage and abnormal loss. In addition, the abnormal position of the fiber link can also be determined by the return time of the reflected light. However, there is an event blind spot in OTDR detection. When the light leakage is relatively small, OTDR may not be able to accurately locate, or even detect whether light leakage occurs; and for non-reflective events such as optical signal leakage, OTDR only detects continuous loss, without obvious discontinuous detection signal mutation, and the event blind spot is relatively large. With the improvement of eavesdropping technology, the loss generated by the eavesdropping device is extremely small, and the impact on the original communication system is relatively weak; at the same time, the status monitoring data obtained during the operation of the equipment is becoming more and more diverse, and the structure and operating environment of the equipment itself are becoming more and more complex. The difficulty of extracting features related to the eavesdropping status is also increasing, and OTDR cannot effectively monitor the changes in channel parameters caused by eavesdropping.

[0036] To this end, the present invention proposes a physical layer eavesdropping positioning method for software-defined optical networks to address the problems of high cost, low positioning accuracy and large event blind spots in existing positioning solutions for dealing with the problem of physical layer eavesdropping in optical networks. This method uses the rich time domain information of optical signals after photoelectric conversion and the ability of neural networks to analyze and process the connection between a large amount of information to make intelligent, efficient and accurate judgments on the operation of optical network links, which helps to enhance the security of optical networks and protect information security.

[0037] Specifically, the present invention first defines a macroscopic eavesdropping attack positioning network system, such as Figure 1As shown, the eavesdropping attack positioning network system includes terminal equipment, optical modules, OXC (Optical Cross-Connect), PD (PhotoDiode) and optical network, and the terminal equipment can be used as both a sender and a receiver; in this system, there may be attackers who steal part of the signal in the optical network link. Since the signal loss caused by eavesdropping is very small, it is difficult for both the sender and the receiver to find it. Therefore, when the signal is transmitted through ON (Optical Network), the Optical Network Controller in SDN (Software Defined Network) is used for transmission monitoring. In the eavesdropping attack positioning network system, the sender sends a signal through the optical module and OXC to the optical network, and then goes down through OXC and the optical module, and finally reaches the receiver for DSP processing. The eavesdropping attack locating process in the eavesdropping attack locating network system using the eavesdropping locating model includes: the optical network controller periodically obtains the data after photoelectric conversion through the PD deployed in the optical network, feeds the data after photoelectric conversion to the trained neural network (i.e., the eavesdropping locating model) through SDN uploading for eavesdropping locating detection, determines the location where the optical eavesdropping occurs, and if attacked, sends the network attack situation to the network security administrator in the form of an alarm trigger, so as to quickly deal with the network attack.

[0038] Based on the above eavesdropping attack positioning network system, such as Figure 6 As shown, the specific scheme of the present invention includes:

[0039] S1. Building a coherent optical communication system, the coherent optical communication system includes a coherent transmitter, a coherent receiver and an optical fiber link with a length of 100km;

[0040] S2. The coherent transmitter outputs a 200G-16QAM dual-polarization signal to the optical fiber link, and uses an optical splitter in the optical fiber link to simulate different degrees of eavesdropping attacks;

[0041] S3. The coherent receiver collects a piece of photoelectric conversion data every fixed time window, and obtains multiple pieces of photoelectric conversion data to form a data set;

[0042] Specifically, in this embodiment, the coherent receiver first receives the signal transmitted by the optical fiber link and performs optical-to-electrical conversion on it, and then divides it into fixed time windows.

[0043] Specifically, the specific process of data collection includes:

[0044] S21. Starting from the left end of the 100 km optical fiber link, mark an eavesdropping position at 0 km, 10 km, 20 km, 30 km, 40 km, 50 km, 60 km, 70 km, 80 km and 90 km to the right;

[0045] S22. Use a 1% optical splitter to simulate an attacker at the i=1, 2, ..., 10th eavesdropping position, then use a coherent transmitter to output a 200G-16QAM dual-polarization signal to the optical fiber link, and the coherent receiver collects the signal and performs photoelectric conversion. A fixed time window is used to divide the photoelectric conversion data into multiple segments and sample and store them at intervals; each segment of photoelectric conversion data includes four signals: XI (X-pol In-phase, X-polarization state in-phase component), XQ (X-pol Quadrature, X-polarization state orthogonal component), YI (Y-pol In-phase, Y-polarization state in-phase component), and YQ (Y-pol Quadrature, Y-polarization state orthogonal component);

[0046] S23. Use a 10% optical splitter to simulate an attacker at the i=1, 2, …, 10th eavesdropping position, then use a coherent transmitter to output a 200G-16QAM dual-polarization signal to the optical fiber link, and the coherent receiver collects the signal and performs photoelectric conversion. A fixed time window is used to divide the photoelectric conversion data into multiple segments and sample and store them at intervals; each segment of photoelectric conversion data includes four signals: XI, XQ, YI, and YQ;

[0047] S24. Combining multiple segments of photoelectric conversion data collected by the 1% splitter and the 10% splitter at different eavesdropping positions into a data set.

[0048] S4. preprocess the data set, and divide the preprocessed data set into a training set, a validation set, and a test set in a ratio of 3:1:1;

[0049] Specifically, step S4 preprocesses the data set, including:

[0050] S41. Extract the signal amplitude, signal frequency, signal phase, signal rise time, signal fall time, signal period, signal upper and lower limits, signal peak value and signal waveform of each signal in each section of photoelectric conversion data in the data set to form a dimensional data, and form a set of sample data by combining the dimensional data corresponding to the four signals in each section of photoelectric conversion data; perform data cleaning on all sample data to remove sample data with obvious errors;

[0051] S42. Normalizing and standardizing the cleaned sample data set to obtain a standard data set;

[0052] S43. According to the eavesdropping attack location, each set of standard sample data in the standard data set is labeled accordingly;

[0053] S44. After labeling, all standard sample data are shuffled to obtain the preprocessed data set.

[0054] S5. Use the training set to train the eavesdropping attack localization model, calculate the loss value and back propagate until the model converges or reaches the maximum number of iterations.

[0055] Specifically, in the data collection stage, the coherent optical communication system constructed by the present invention to realize the eavesdropping attack is as follows: Figure 2 As shown, it includes a coherent transmitter, a coherent receiver and an optical fiber link between the two. In the coherent optical communication system, the coherent transmitter transmits the optical signal to the optical link with an eavesdropper for transmission, and finally passes through an OBPF (Optical Band Pass Filter) with a bandwidth of 50GHz and a center frequency of the transmitted 16QAM signal for filtering. The coherent receiver receives the filtered optical signal and converts it into an electrical signal using a PD (PhotoDiode), and then divides the electrical signal data using a fixed time window.

[0056] The optical signal emitted by the coherent transmitter is a 200G-16QAM dual-polarization signal, and the optical fiber link consists of a 100km long polarization division multiplexing standard single-mode optical fiber and an EDFA. Each time, an eavesdropper is set at a certain position in the optical fiber link to simulate an eavesdropping situation, and then the coherent receiver collects the corresponding electrical signal data once; the specific setting is to use 1% and 10% splitters to couple and split at different positions (0km, 10km, ... 90km) in the optical fiber link to simulate different eavesdropping situations. Table 1 reflects the details of sample collection at different eavesdropping positions.

[0057] Table 1

[0058]

[0059]

[0060] Specifically, to ensure the transmission effectiveness of the coherent optical communication system, the coherent receiver receives the signal transmitted from the optical fiber link and performs digital signal processing on it so that the BER (Bit Error Rate) is zero.

[0061] Figure 3The following is a signal comparison diagram after the coherent receiver performs photoelectric conversion on the received signal when the eavesdropping position is 0km, 10km, ..., 90km. It can be seen that the signal comparison difference at different eavesdropping positions is small, and it is difficult for the naked eye to distinguish the two situations directly through the characteristics of the signal diagram. But in fact, there are subtle differences in the signals at different eavesdropping positions, including differences in amplitude, frequency, phase, rise time and fall time, periodicity, signal upper and lower limits, peak value and waveform symmetry, and they have rich characteristic information. These characteristic information will be affected by different eavesdropping positions, but due to the mutual influence of characteristic information, it is difficult to find the corresponding relationship between characteristic information and different eavesdropping situations through general methods. The continuous development of neural networks provides a new idea for this problem. Neural networks are calculated by connecting a large number of artificial neurons. In most cases, artificial neural networks can change their internal structure based on external information. They are adaptive systems, which generally speaking have learning functions. Modern neural networks are a nonlinear statistical data modeling tool. Neural networks are usually optimized through a learning method based on mathematical statistics. Artificial neural networks can have simple decision-making and simple judgment abilities similar to humans. This method has more advantages than formal logical reasoning calculations.

[0062] This paper uses the subtle difference characteristics in the time domain signal to input the collected photoelectric conversion data into the eavesdropping attack positioning model based on BP neural network proposed in this paper to perform eavesdropping attack positioning detection. In order to prevent the model from being affected by a single feature during the training process, falling into a local optimal solution and affecting the results, the data is normalized and standardized in the data preprocessing stage, and the features are fused. This processing method can not only improve the robustness of the model but also reduce the amount of calculation and accelerate the convergence of the model. At the same time, when the signal is transmitted in the optical network, the signal will be affected by nonlinear effects such as polarization mode dispersion, dispersion, and phase shift. The influence of different eavesdropping positions will produce coupling between different polarization states, between different signals in the same polarization state, and between different signals in different polarization states. In order to fully retain and utilize these features, this paper takes the stacking of the four-way signals XI, XQ, YI, and YQ, and processes them into a four-dimensional data to realize feature fusion, increase the feature vector distance of different eavesdropping positions, and improve the positioning accuracy of the model. The final input data dimension in the embodiment of the present invention is (8192, 4).

[0063] Specifically, the eavesdropping attack location detection model based on BP neural network proposed in this paper is a fully connected multi-layer feedforward neural network that adopts the error back propagation algorithm. It contains an input layer, a hidden layer and an output layer, in which the input layer is responsible for receiving input data, the hidden layer processes the input data, and the output layer outputs the final result. The training of BP neural network is also a supervised learning process, which is divided into two stages: forward propagation of information and backward propagation of error. During the forward propagation process, the input features are passed to the network, the predicted output of the model is calculated, and the value of the loss function is calculated by comparing the predicted value of the model with the true value, which is used to measure the difference between the predicted value and the actual value. During the back propagation process, according to the loss function, the gradient is calculated, and the optimization algorithm is used to update the weights and biases in the network. Such as Figure 4 As shown, the wiretap attack location detection model based on BP neural network of the present invention includes an input layer, four hidden layers and an output layer; wherein the number of nodes in the input layer is 8192, the number of nodes in the four hidden layers is 4096, 2048, 512 and 256 respectively, and the number of nodes in the output layer is 10. The model uses Adam optimizer in the training process, the activation function of the hidden layer neurons is selected as ReLU function, and the loss is minimized by gradient descent.

[0064] Finally, the 10 nodes in the output layer represent the positioning of different eavesdropping locations. Therefore, the model proposed in this paper can be regarded as a multi-classification problem. The unique hot encoding method is used to use 0000000001 to represent the situation of being eavesdropped at 0km, 0000000010 to represent the situation of being eavesdropped at 10km, and so on. 1000000000 represents the situation of being eavesdropped at 90km. The loss function selects the CategoricalCrossentropy function to calculate the cross entropy loss between the true value and the predicted value:

[0065]

[0066] where y i,k represents the actual one-hot encoded label, φ i,k Represents the actual output of the model, N represents the total number of samples, and K is the length after one-hot encoding.

[0067] Use the mean absolute error function MAE (Mean Absolute Error) as the evaluation function:

[0068]

[0069] Among them, y i represents the true value, y i Represents the predicted value of the model.

[0070] The output layer uses the sigmoid activation function:

[0071]

[0072] The entire network uses gradient descent to minimize the loss during the training process. The smaller the loss value, the closer the model prediction value is to the true value. The signal data after photoelectric conversion is input into the trained neural network, and the output of the model is calculated through forward propagation, which is the prediction result.

[0073] In this embodiment, VPI transmission Maker 9.5 is used as the simulation software. The coherent optical transmitter transmits the optical signal to the optical link with the eavesdropping point for transmission. Finally, it is filtered by an OBPF (Optical Band Pass Filter) with a center frequency of 50GHz bandwidth and a center frequency of the transmitted 16QAM signal. The coherent receiver receives the optical signal after filtering and converts the optical signal into an electrical signal using a PD (PhotoDiode). The electrical signal data is stored and input into the eavesdropping positioning model for detection and positioning. Some simulation parameters in VPI are shown in Table 2:

[0074] Table 2

[0075]

[0076] For the 16QAM optical signal transmitter, the baud rate is set to 25 GBaud, the frequency, and the transmitted optical power is 0 dBm. For the optical fiber link, the standard single-mode optical fiber length is set to 100 km, the optical fiber attenuation is set to 0.2 dB / km, the dispersion is 16 ps / (nm·km), and the dispersion slope is set to 0.08×10 3 s / m 3 , the group refractive index is set to 1.47, and the polarization mode dispersion is set to The nonlinear coefficient is set to 2.6×10 -20 m 2 / W, EDFA power is set to 20dBm.

[0077] The experiments of the present invention were carried out with TensorFlow version 2.6.0, Python version 3.8.12, CPU model Intel(R) Core(TM) i7-10700 CPU@2.90GHz, and GPU model NVIDIA GeForce GTX 1660.

[0078] Figure 5The figure shows the change of the positioning accuracy and loss of the proposed BP model on the validation set during the training process with the number of iterations. It can be seen that with the increase of the number of iterations, the detection accuracy increases until the positioning accuracy reaches 91.2%, and the loss value stabilizes after 100 iterations. This proves that the model has good convergence and is easy to implement.

[0079] In the present invention, unless otherwise clearly stipulated and limited, the terms such as "installation", "setting", "connection", "fixation" and "rotation" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral one; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. Unless otherwise clearly defined, ordinary technicians in this field can understand the specific meanings of the above terms in the present invention according to the specific circumstances.

[0080] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A physical layer eavesdropping location method for software-defined optical networks, characterized in that: Define the optical network physical layer eavesdropping location network system, build and train the eavesdropping attack location model based on BP neural network. When the Optical Network Controller detects that there is eavesdropping in the optical network physical layer eavesdropping location network system, it obtains the photoelectric conversion data of the optical network physical layer eavesdropping location network system and inputs it into the eavesdropping attack location model to locate the eavesdropping position, and sends it to the network security administrator in the form of an alarm trigger, so as to quickly deal with the network attack; The training process of the eavesdropping attack location model includes the following steps: S1. Building a coherent optical communication system, the coherent optical communication system includes a coherent transmitter, a coherent receiver and an optical fiber link with a length of 100km; S2. The coherent transmitter outputs a 200G-16QAM dual-polarization signal to the optical fiber link, which is then filtered by an OBPF with a 50 GHz bandwidth and a center frequency equal to the center frequency of the transmitted 16QAM signal. An optical splitter is used in the optical fiber link to simulate different degrees of eavesdropping attacks. S3. The coherent receiver receives the filtered 200G-16QAM dual-polarization signal and converts it into an electrical signal using a PD (PhotoDiode). The electrical signal data is then divided using a fixed time window, and the data divided from the electrical signal data is used as photoelectric conversion data. Multiple segments of photoelectric conversion data are obtained to form a data set. S4. preprocess the data set, and divide the preprocessed data set into a training set, a validation set, and a test set in a ratio of 3:1:1; Step S4 preprocesses the data set, including: S41. Extract the signal amplitude, signal frequency, signal phase, signal rise time, signal fall time, signal period, signal upper and lower limits, signal peak value and signal waveform of each signal in each section of photoelectric conversion data in the data set to form a dimensional data, and form a set of sample data by combining the dimensional data corresponding to the four signals in each section of photoelectric conversion data; perform data cleaning on all sample data to remove sample data with obvious errors; S42. Normalizing and standardizing the cleaned sample data set to obtain a standard data set; S43. According to the eavesdropping attack location, each set of standard sample data in the standard data set is labeled accordingly; S44. After labeling, all standard sample data are shuffled to obtain a preprocessed data set; S5. Use the training set to train the eavesdropping attack positioning model, calculate the loss value and back propagate until the model converges or reaches the maximum number of iterations.

2. A physical layer eavesdropping location method for software defined optical networks according to claim 1, characterized in that: The optical network physical layer eavesdropping and positioning network system includes a transmitting end, a receiving end, an optical module, an OXC and an optical network; the working process of the optical network physical layer eavesdropping and positioning network system is as follows: the transmitting end signal enters the optical network through the optical module and the OXC uplink, and then reaches the receiving end through the OXC and the optical module downlink, and the receiving end performs DSP processing on the received signal; wherein the Optical Network Controller in the SDN is used to perform signal transmission monitoring on the optical network physical layer eavesdropping and positioning network system.

3. The physical layer eavesdropping location method for software defined optical network according to claim 1, characterized in that: In the coherent optical communication system, the coherent transmitter outputs a 200G-16QAM dual-polarization signal to the optical fiber link. When the coherent receiver performs digital signal processing on the received data, the BER is set to zero to ensure the transmission effectiveness of the coherent optical communication system.

4. The physical layer eavesdropping location method for software defined optical network according to claim 1, characterized in that: The specific process of data collection includes: S21. Starting from the left end of the 100 km optical fiber link, mark an eavesdropping position at 0 km, 10 km, 20 km, 30 km, 40 km, 50 km, 60 km, 70 km, 80 km and 90 km to the right; S22. Use a 1% optical splitter to simulate an attacker at the i=1, 2, …, 10th eavesdropping position, then use a coherent transmitter to output a 200G-16QAM dual-polarization signal to the optical fiber link, and the coherent receiver collects the signal and performs photoelectric conversion. A fixed time window is used to divide the photoelectric conversion data into multiple segments and sample and store them at intervals; each segment of photoelectric conversion data includes four signals: XI, XQ, YI, and YQ; S23. Use a 10% optical splitter to simulate an attacker at the i=1, 2, …, 10th eavesdropping position, then use a coherent transmitter to output a 200G-16QAM dual-polarization signal to the optical fiber link, and the coherent receiver collects the signal and performs photoelectric conversion. A fixed time window is used to divide the photoelectric conversion data into multiple segments and sample and store them at intervals; each segment of photoelectric conversion data includes four signals: XI, XQ, YI, and YQ; S24. Combining multiple segments of photoelectric conversion data collected by the 1% splitter and the 10% splitter at different eavesdropping positions into a data set.

5. The physical layer eavesdropping location method for software defined optical network according to claim 1, characterized in that: The optical fiber link is composed of polarization multiplexing standard single-mode optical fiber and EDFA.

6. The physical layer eavesdropping location method for software defined optical network according to claim 1, characterized in that: An optical bandpass filter is provided at the coherent receiver. The signal transmitted via the optical fiber link is first processed by the optical bandpass filter, and then the processed signal is transmitted to the coherent receiver.

7. The physical layer eavesdropping location method for software defined optical network according to claim 1, characterized in that: The eavesdropping attack location detection model based on BP neural network includes an input layer, four hidden layers and an output layer; the number of nodes in the input layer is 8192, the number of nodes in the four hidden layers are 4096, 2048, 512 and 256 respectively, and the number of nodes in the output layer is 10.