A method and system for intelligent cognition in wireless ad hoc networks

Abstract

This invention relates to an intelligent cognitive method and system for wireless ad hoc networks, belonging to the field of wireless ad hoc network technology. The invention employs machine learning algorithms to automatically identify the MAC protocol of the target network's data link layer using signals intercepted by reconnaissance equipment. It also identifies the source of transmitted and received signals through radiation source identification, thereby reconstructing the network topology and identifying key network nodes. The designed cognitive method and system can intelligently analyze the target network's topology and key nodes using physical layer signals. This provides the decision-making system with targeted interference and attack targets, enabling the implementation of various attack methods to achieve the operational objective of "disconnecting, breaking, and paralyzing" the target ad hoc network, thus improving our operational effectiveness.

Description

Technical Field

[0001] This invention belongs to the field of wireless ad hoc network technology, specifically relating to an intelligent cognitive method and system for wireless ad hoc networks. Background Technology

[0002] With the development and application of wireless communication and network technologies in the military field, communication network confrontation has become an important form of modern warfare. In the modern battlefield environment, aerial drones, fighter jets, reconnaissance aircraft, and ground combat platforms and command centers transmit information through wireless ad hoc networks. The entire operational network possesses advantages such as rapid network deployment, high mobility and flexibility, multi-hop transmission, dynamic topology, and strong anti-jamming capabilities. In communication network confrontation, understanding and analyzing the enemy's ad hoc network is crucial for disrupting its information flow and gaining information control and initiative on the battlefield. Therefore, how to intelligently analyze the topology and key combat nodes of a target ad hoc network, using signal processing, network analysis, and artificial intelligence technologies, without knowing the target's ad hoc network information, is a key issue that needs to be addressed for targeted interference and attacks on the enemy's ad hoc network.

[0003] There are many studies on the cognition and analysis of wireless ad hoc networks. For example, Yang Yi et al. in the paper "A dynamic topology discovery method for smart substation networks" (2014 China International conference on electricity distribution (CICED 2014), Sept. 2014, pp. 1070-1074) used address information contained in signals to deduce the communication relationship between nodes. However, in the battlefield environment, it is usually difficult to crack the enemy's communication data packets. Therefore, this method is difficult to extend to the cognitive analysis problem of tactical ad hoc networks. Yang Hongwa et al. in the paper "Battlefield backbone network topology inference technology" (Communication Countermeasures, 2009(3): pp. 14-17) proposed a battlefield backbone network topology inference method based on wireless signal reconnaissance. This method includes key technologies such as signal screening, source clustering, node identification, backbone link identification and topology confirmation. However, signal screening requires prior knowledge of the enemy's network information, including node type, link rate, modulation style, etc. The specific judgment rules for source clustering and node identification processes need to be determined according to the specific combat equipment type of the enemy. In his paper "Research on Communication Network Cognition and Interference Strategies" (Xi'an University of Electronic Science and Technology, 2021), Yang Hao used multifractal theory to extract device fingerprint features, then density clustering algorithm to identify communication nodes, and finally used node information to identify communication links, ultimately obtaining the network topology and key nodes. However, in the process of radiation source identification through density clustering, in order to achieve feature dimensionality reduction, it is necessary to manually analyze and determine distance and density thresholds. Fan Min, in his paper "Research on Key Node Identification Technology in Tactical Networks" (University of Electronic Science and Technology of China, 2020), proposed a method for network topology acquisition and key node identification. However, the signal used is relatively ideal and a supervised learning method is used to achieve radiation source identification, making it difficult to apply to battlefield environments where our side does not know the enemy's information.

[0004] Based on the above analysis, the problems and shortcomings of the existing technology are as follows:

[0005] (1) In terms of network topology reconstruction, the existing method of cracking the communication data packets of the target self-organizing network and obtaining the address information of the communication nodes to reconstruct the network topology is difficult to apply to non-cooperative communication scenarios in the battlefield environment.

[0006] (2) In terms of key node identification, existing technologies such as degree value method, mediation method, node deletion method, and multi-attribute fusion method have problems such as one-sided perception of topological features, destruction of topological structure, or high complexity, which are difficult to meet the needs of complex network cognition and confrontation in battlefield environment.

[0007] (3) In communication network confrontation, the introduction of artificial intelligence technology to achieve intelligent network topology cognition is an important development trend. However, in terms of radiation source identification, existing solutions either ignore the premise that the opponent's communication network and equipment information is unknown, or require human parameter analysis and setting when using machine learning algorithms, which cannot overcome the dependence on command and decision-making, and the identification efficiency and accuracy need to be improved. Summary of the Invention

[0008] (a) Technical problems to be solved

[0009] The technical problem to be solved by this invention is: how to intelligently analyze the topology and key combat nodes of a target ad hoc network by using signal processing, network analysis and other methods, combined with artificial intelligence technology, without knowing the information of the target ad hoc network.

[0010] (II) Technical Solution

[0011] To address the aforementioned technical problems, this invention provides an intelligent cognitive method for wireless ad hoc networks, comprising the following steps:

[0012] (1) Obtain the communication signals of the target ad hoc network;

[0013] (2) Preprocess the communication signal samples, including energy detection, blind signal separation, and data fragmentation;

[0014] (3) Based on the differences in the start time interval of data packets in different types of MAC protocols, the K-means algorithm and LSTM network are used to identify the MAC protocol categories;

[0015] (4) Based on step (3), for contention-type MAC protocols, the SVM algorithm is used to identify the specific MAC protocol category of the target ad hoc network based on the theoretical values ​​of frame analogy and duty cycle.

[0016] (5) Based on the identification of MAC protocol categories, the preprocessed communication signal samples are matched according to the data interaction mode of the MAC protocol;

[0017] (6) Identify the radiation source for the successfully matched signal segments to obtain the identifiers of the connected node pairs;

[0018] (7) Perform steps (2) to (6) on all intercepted communication signal samples to obtain the identifiers of all mutually communicating node pairs in the target ad hoc network, and then reconstruct the network topology of the target ad hoc network.

[0019] (8) Using complex network theory and the reconstructed target self-organizing network topology, calculate the importance of edges and the contribution rate of nodes to edges, thereby obtaining the importance index of each node.

[0020] (9) Sort the importance of each node in the network to obtain the key nodes in the target self-organizing network.

[0021] Preferably, in step (2), firstly, the energy detection method is used to filter out blank and noise segments in the signal and extract the effective signal portion; secondly, independent component analysis is used to perform blind signal separation on the effective signals from multiple radiation sources; finally, the signal sample obtained after blind signal separation is divided into data segments of equal length, with each segment having a length of N = 2. n .

[0022] Preferably, the identification of the MAC protocol category in step (3) specifically includes the following steps:

[0023] 3a) First, based on the length difference between the data group and the control group and management group, the K-means algorithm is used to divide the group into two categories: the first category is the data group, and the second category is the control group and management group.

[0024] 3b) By utilizing the differences in the start time interval of data packets in the three MAC protocols of allocation, contention and hybrid, a pre-trained LSTM network is used to identify the category of the target ad hoc network's MAC protocol.

[0025] Preferably, the specific steps for identifying the target ad hoc network's MAC protocol category in step (4) include:

[0026] 4a) Using the grouping and clustering results in step 3a), the frame ratio and duty cycle of samples using the competitive MAC protocol at the link layer are statistically analyzed and used as feature vectors;

[0027] 4b) Based on the theoretical values ​​of frame analogy and duty cycle of different competing MAC protocols such as MACA, CSMA / CA, and MACA-BI, the feature vector is input into a pre-trained SVM to obtain the MAC protocol category of the target ad hoc network link layer.

[0028] Preferably, the specific steps for radiation source identification in step (6) include:

[0029] 6a) Based on multifractal theory, calculate the corresponding singularity index under different powers and dimensions, and construct the feature vector of the signal sample;

[0030] 6b) After standardizing the feature vectors, input them into the autoencoder to reduce the dimensionality of the feature vectors;

[0031] 6c) Using low-dimensional feature vectors, a density clustering algorithm based on kernel density estimation is used to classify and identify radiation sources.

[0032] Preferably, step 6a) specifically includes:

[0033] 6a.1 Initialize the exponent q to (2q0+1) values ​​in the interval [-q0, q0];

[0034] 6a.2 Under different values ​​of q, the signal segment is divided n times using different line dimensions θ, where the line dimension θ = 2 in the i-th division. i ;

[0035] 6a.3 Calculate the singularity index for different values ​​of q and θ:

[0036]

[0037] Where T = θ / 2 is the size of each signal interval obtained by dividing the signal segment s using the linear dimension θ, and f i q (θ) is the density distribution function:

[0038]

[0039] s represents the signal sequence, and s(k) represents the k-th sample point;

[0040] 6a.4 Under different values ​​of q, the signal segment is divided n times. The calculated singularity index is spliced ​​together, and the resulting multifractal spectrum vector with a length of (2q0+1)·n is the feature vector of the signal.

[0041] Preferably, step 6c) involves using a density clustering algorithm based on kernel density estimation to classify and identify radiation sources, specifically including: using a kernel density estimation method and an integral mean square error criterion, based on the variance of the signal samples... Determine the required Eps neighborhood in the density clustering algorithm DBSCAN Minimum number of points Let the signal samples be x1, x2, ... x i ,...,x n ;

[0042] minPts is the number of sample points existing in a space centered at x and with radius h, where the kernel function K is defined as follows: Where d is the dimension of the sample points, the DBSCA algorithm is then used to identify different signal radiation sources.

[0043] Preferably, steps (8) and (9) specifically obtain the key nodes in the target ad hoc network through the following steps:

[0044] 8a) In the reconstructed target ad hoc network topology, edge e uvThe network is divided into two regions, U and V. The former is the region closer to node u, and the latter is the region closer to node v. First, the edge e in the network is calculated. uv Importance:

[0045]

[0046] in, and They are respectively edge e uv Connectivity and substitutability, where p and k are integers, connectivity The following formula is used to calculate:

[0047]

[0048] in, and Let p be the number of p-order branches in regions U and V, respectively. The following conditions must be met:

[0049]

[0050] in, Let p be the set of p-order neighbor nodes of node u. Let Γ′ be the set of p-1 order neighbor nodes of node v. (u) Let u be the set of neighbors with degree 1; substitutability The following formula is used to calculate:

[0051]

[0052] Among them, L j Therefore, e uv The number of polygons with sides;

[0053] 8b) Calculate the number of p-order branches of node u on edge e using the number of p-order branches of regions U and V. uv Contribution rate:

[0054]

[0055] 8c) According to edge e uv The importance of node u is determined by its importance and contribution rate to the edge.

[0056]

[0057] Among them, Γ (u) Let be the set of neighbors of node u; based on the ranking of the importance of each node in the network, the key nodes of the target ad hoc network are obtained.

[0058] The present invention also provides an intelligent cognitive system for implementing the method in a wireless ad hoc network, comprising:

[0059] The signal receiving module is used to collect communication signals of the target ad hoc network through a radio monitoring receiver and to provide information on the start time and duration of the signal.

[0060] The signal preprocessing module is used to perform energy detection on communication signals, extract the effective signal portion, perform blind signal separation on signals from multiple radiation sources, and divide signal samples into data segments of equal length.

[0061] The MAC protocol identification module is used to identify MAC protocol types based on the differences in the start time interval of data packets of different protocols, as well as the differences in frame analogy and duty cycle, using machine learning algorithms.

[0062] The radiation source matching module is used to match the physical layer signal segments of the preprocessed communication signal based on the identification of the MAC protocol type and the data interaction mode of the MAC protocol, and to identify the transmitting and receiving nodes of the signal through radiation source identification.

[0063] The network topology analysis module, based on the MAC protocol identification module and radiation source matching module, reconstructs the network topology by analyzing the communication relationships between nodes in the target network and identifies key nodes in the network, enabling cognitive analysis of the target ad hoc network under non-cooperative communication confrontation.

[0064] The present invention also provides a method for effectively interfering with and attacking a target network based on the aforementioned method.

[0065] (III) Beneficial Effects

[0066] This invention addresses the communication network adversarial problem in non-cooperative communication scenarios by proposing an intelligent cognitive method and system for wireless ad hoc networks. Through signal reconnaissance and processing, combined with machine learning algorithms and complex network theory, K-means, LSTM, and SVM algorithms are used to automatically identify the MAC protocol of the target ad hoc network. An autoencoder and a density clustering method based on kernel density estimation are used to reduce the dimensionality of the extracted multifractal spectral vector of the signal, thereby enabling automatic identification and labeling of signal radiation sources. Edge-based and local partitioning methods are used to effectively identify key network nodes.

[0067] This invention fully considers the characteristics of communication network warfare, enabling rapid identification of key nodes through signal reconnaissance and network analysis even when the target ad hoc network and communication equipment information are unknown, thereby providing strategic support for the decision-making system. The radiation source identification method proposed in this invention employs an unsupervised learning algorithm and does not rely on manual parameter analysis and setting, avoiding the influence of human intervention on the results while improving identification efficiency and accuracy. The proposed key node identification method can accurately identify key nodes in the network with low computational complexity by evaluating the importance of edges and nodes. Furthermore, based on the identification and analysis of the opponent's communication patterns, protocol types, and formats, further network attacks can be launched against the enemy ad hoc network, achieving integrated network and electronic warfare attacks by combining electronic warfare techniques. Attached Figure Description

[0068] Figure 1 This is a flowchart of the intelligent cognitive method for wireless ad hoc networks provided in an embodiment of the present invention;

[0069] Figure 2 This is a schematic diagram of the structure of the intelligent cognitive system for wireless self-organizing networks provided in an embodiment of the present invention;

[0070] Figure 3 This is a schematic diagram of a cognitive adversarial scenario in a wireless ad hoc network provided in an embodiment of the present invention;

[0071] Figure 4 This is a flowchart illustrating the implementation of the intelligent cognitive method for wireless ad hoc networks provided in this embodiment of the invention.

[0072] Figure 5 This is a schematic diagram of network key node identification provided in an embodiment of the present invention. Detailed Implementation

[0073] To make the objectives, contents, and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0074] This invention provides an intelligent cognitive method and system for wireless ad hoc networks. In a non-cooperative communication confrontation environment, it utilizes reconnaissance signals to analyze the topology and key nodes of a target ad hoc network, assisting in command and decision-making. The intelligent cognitive method for wireless ad hoc networks employs machine learning algorithms. In a non-cooperative communication confrontation scenario, it uses signal reconnaissance equipment to intercept target ad hoc network communication signals. Based on signal preprocessing, it reconstructs the target network's topology and identifies its key nodes through MAC protocol identification and signal radiation source matching. This provides intelligent network analysis results to the decision-making system, assisting in effective interference and attacks against the target network.

[0075] like Figure 1As shown, the intelligent cognitive method for wireless ad hoc networks provided by this invention includes the following steps:

[0076] (1) Acquire the communication signals of the target ad hoc network provided by the signal reconnaissance equipment;

[0077] (2) Preprocessing of communication signals, including energy detection, blind signal separation, and data fragmentation;

[0078] (3) Based on the differences in the start time interval of data packets in different types of MAC protocols, the K-means and LSTM algorithms are used to identify the MAC protocol categories;

[0079] (4) Based on step (3), for contention-type MAC protocols, the specific MAC protocol of the target ad hoc network is identified using the SVM algorithm based on the theoretical values ​​of frame analogy and duty cycle.

[0080] (5) Based on the identification of the MAC protocol type, the preprocessed communication signals (physical layer signals) are matched according to the data interaction mode of the MAC protocol;

[0081] (6) Input the successfully matched signal segments into the radiation source identification model to obtain the identification of connected node pairs and reconstruct the network topology;

[0082] (7) Calculate the importance of each node using complex network theory and the network topology of the target self-organizing network, and thus determine the key nodes in the network.

[0083] The intelligent cognitive method for wireless ad hoc networks provided by this invention can also be implemented using other steps by those skilled in the art. Figure 1 The intelligent cognitive method for wireless ad hoc networks provided by this invention is merely a specific embodiment.

[0084] like Figure 2 As shown, the intelligent cognitive system for wireless ad hoc networks provided by this invention includes:

[0085] (1) Signal receiving module, used to collect communication signals of the target self-organizing network through radio monitoring receiver, and provide information such as the start time and duration of the signal;

[0086] (2) Signal preprocessing module, used to perform energy detection on the received signal, extract the effective signal part, perform blind signal separation on signals from multiple radiation sources, and divide the signal sample into data segments of equal length to provide effective usable data for subsequent processing and analysis;

[0087] (3) MAC protocol identification module, which is used to identify MAC protocol type by using machine learning algorithm based on the differences in the start time interval of data packets of different protocols and the differences in frame analogy and duty cycle;

[0088] (4) Radiation source matching module, used for specific MAC protocol data interaction mode, to match physical layer signal segments and identify signal transmission and reception nodes through radiation source identification;

[0089] (5) Network topology analysis module, which is used to reconstruct the network topology by analyzing the communication relationships between nodes in the target network and to identify key nodes in the network, so as to realize the cognitive analysis of the target ad hoc network under non-cooperative communication confrontation.

[0090] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0091] Example 1

[0092] This invention provides an intelligent cognitive method for wireless ad hoc networks, which can be applied to, but is not limited to, various applications. Figure 3 The combat scenario shown. Figure 3 In this context, the enemy's battlefield communication network is a wireless ad hoc network composed of drones, fighter jets, and ground combat platforms. Our reconnaissance aircraft are equipped with signal reconnaissance equipment, which detects communication signals within the target network and transmits this information to our command and decision-making center. Utilizing the intelligent cognitive method and system for wireless ad hoc networks provided by this invention, we can achieve intelligent perception of the target ad hoc network's topology and key nodes. This allows us to implement electronic jamming, network attacks, and other attack methods to disrupt its normal operational processes and network function, achieving the operational objective of "breaking the link, destroying the network, and paralyzing the entire system." (Refer to...) Figure 4 The specific implementation steps of the present invention include:

[0093] (1) Acquire the communication signals of the target ad hoc network provided by the signal reconnaissance equipment;

[0094] (2) Preprocessing the communication signal, the specific steps of which include:

[0095] 2a) Using energy detection, filter out blank and noise segments in the signal to extract the effective signal portion;

[0096] 2b) The independent component analysis method is used to perform blind signal separation on the effective signals from multiple radiation sources;

[0097] 2c) Divide the signal samples obtained after blind signal separation into equal-length data segments, each segment having a length of N = 2. n This is to facilitate subsequent processing and analysis;

[0098] (3) Based on the differences in the start time interval of data packets in different categories of MAC protocols, the K-means algorithm and LSTM network are used to identify the MAC protocol categories, i.e., to perform MAC protocol inter-class identification. The specific steps include:

[0099] 3a) First, based on the length difference between the data group and the control group and management group, the K-means algorithm is used to divide the group into two categories: the first category is the data group, and the second category is the control group and management group.

[0100] 3b) By utilizing the differences in the start time interval of data packets in the three MAC protocols of allocation, contention and hybrid, a pre-trained LSTM network is used to identify the category of the target ad hoc network's MAC protocol.

[0101] (4) Based on step (3), for contention-based MAC protocols, according to the theoretical values ​​of frame analogy and duty cycle, the SVM algorithm is used to identify the specific MAC protocol of the target ad hoc network, that is, to perform intra-class identification of contention-based MAC protocols. The specific steps include:

[0102] 4a) Based on the grouping and clustering results in step 3a), calculate the frame analogy and duty cycle of samples using the competitive MAC protocol at the link layer, and use them as feature vectors;

[0103] 4b) Based on the theoretical values ​​of frame analogy and duty cycle of different competing MAC protocols such as MACA, CSMA / CA, and MACA-BI, the feature vector is input into a pre-trained SVM to obtain the MAC protocol category of the target ad hoc network link layer.

[0104] (5) Based on the identified MAC protocol category, match the preprocessed physical layer signal with the corresponding data interaction mode of the MAC protocol;

[0105] (6) Identify the radiation source for the successfully matched signal segments to obtain the identifiers of the connected node pairs;

[0106] (7) Perform steps (2) to (6) on all intercepted communication signal samples to obtain the identifiers of all mutually communicating node pairs in the target ad hoc network, and then reconstruct the network topology of the target ad hoc network.

[0107] (8) Using complex network theory and the reconstructed target self-organizing network topology, calculate the importance of the edges and the contribution rate of the nodes to the edges, thereby obtaining the importance of the nodes;

[0108] (9) Sort the importance of each node in the network to obtain the key nodes in the target self-organizing network.

[0109] Example 2

[0110] The intelligent cognitive method for wireless ad hoc networks of the present invention is the same as that in Embodiment 1. The training process of the pre-trained LSTM network in step (3b) is as follows: taking typical allocation-type MAC protocols such as TDMA, typical contention-type MAC protocols such as CSMA / CA, and typical hybrid-type MAC protocols such as HN as objects, a network topology is constructed using multiple network communication devices using the above-mentioned MAC protocols. In a non-ideal channel containing noise, propagation delay, multipath fading, etc., the communication transmission between devices is simulated and the signal is collected by a signal acquisition device. To construct the dataset, step (2) first preprocesses the communication signals intercepted by the signal reconnaissance equipment, including energy detection and signal segmentation. In step (3), the effective packets in the signal are input into the K-means classifier. Based on the length difference between data packets and control / management packets, the packets are divided into two categories: the first category is data packets, and the second category is control / management packets. Using the difference in the start time interval of data packets in the three MAC protocols (allocation, competition, and hybrid), the start time interval of data packets in the signal sequence is extracted as a feature. Every 10 start time intervals are used as a feature vector, and the label is the MAC protocol category corresponding to the signal sample. This yields the dataset for MAC protocol recognition, which is then divided into a training set, a validation set, and a test set. The LSTM algorithm is used to train the MAC protocol inter-class recognition model. After training the LSTM network, the training model is saved, which is the trained LSTM network.

[0111] Example 3

[0112] The intelligent cognitive method for wireless ad hoc networks of the present invention is the same as in Embodiments 1 and 2. The training process of the pre-trained SVM in step (4b) is as follows: the generation, acquisition, and classification of effective packets of communication signals are the same as in Embodiment 2. Based on the unsupervised classification of packets, the number of data packets, control packets, and management packets of the signal samples are counted respectively. The frame analogy and duty cycle are calculated as feature vectors of the signal samples. The ratio of the sum of control packets and management packets to the number of data packets is the frame analogy, the ratio of the length of all effective packets to the total length of the signal is the duty cycle, and the label is the MAC protocol category corresponding to the signal sample. In this way, a dataset for competitive MAC protocol identification is obtained, and it is divided into training set, validation set, and test set. The MAC protocol identification model is trained using the SVM model, that is, the SVM algorithm is used to realize intra-class identification of MAC protocols, including the use of grid search and cross-validation to select penalty coefficients, kernel function parameters, etc. After the SVM training is completed, the training model is saved, which is the trained SVM.

[0113] Example 4

[0114] The intelligent cognitive method for wireless ad hoc networks of the present invention is the same as that in Embodiments 1 to 3. The radiation source identification in step (6) specifically includes the following steps:

[0115] 6a) Based on multifractal theory, calculate the corresponding singularity exponents under different powers and dimensions to construct the feature vectors of signal samples, specifically including:

[0116] 6a.1 Initialize the exponent q to (2q0+1) values ​​in the interval [-q0, q0];

[0117] 6a.2 Under different values ​​of q, the signal segment is divided n times using different line dimensions θ, where the line dimension θ = 2 in the i-th division. i ;

[0118] 6a.3 Calculate the singularity index for different values ​​of q and θ:

[0119]

[0120] Where T = θ / 2 is the size of each signal interval obtained by dividing the signal segment s using the linear dimension θ, and f i q (θ) is the density distribution function:

[0121]

[0122] s represents the signal sequence, and s(k) represents the k-th sample point;

[0123] 6a.4 Under different values ​​of q, the signal segment is divided n times, and the calculated singularity index is spliced ​​together. The resulting multifractal spectrum vector with a length of (2q0+1)·n is the feature vector of the signal.

[0124] 6b) After standardizing the feature vectors, input them into the autoencoder to reduce the dimensionality of the feature vectors;

[0125] 6c) Using low-dimensional feature vectors, a density clustering algorithm based on kernel density estimation is employed to classify and identify radiation sources. Specifically, this includes using kernel density estimation methods and the integral mean square error criterion, based on the variance of signal samples... Determine the required Eps neighborhood size in the density clustering algorithm DBSCAN Minimum number of points Let the signal samples be x1, x2, ... x i ,...,x n ,

[0126] minPts is the number of sample points existing in a space centered at x and with radius h, where the kernel function K is defined as follows: Where d is the dimension of the sample points; the density clustering algorithm DBSCAN is used to identify different signal radiation sources, where the Eps neighborhood and the minimum number of points are the values ​​determined by the kernel density estimation mentioned above, and finally the identifiers of interconnected node pairs can be obtained.

[0127] Example 5

[0128] The intelligent cognitive method for wireless ad hoc networks of the present invention is the same as that in Examples 1 to 4, except that steps (8) and (9) specifically obtain the key nodes in the target ad hoc network through the following steps:

[0129] 8a) In the reconstructed target ad hoc network topology, such as Figure 5 As shown, edge e uv The network is divided into two regions, U and V. The former is the region closer to node u, and the latter is the region closer to node v. First, calculate the edge e in the network. uv Importance:

[0130]

[0131] in, and They are respectively edge e uv Connectivity and substitutability, where p and k are integers, connectivity The following formula is used to calculate:

[0132]

[0133] in, and Let p be the number of p-order branches in regions U and V, respectively. The following conditions must be met:

[0134]

[0135] in, Let p be the set of p-order neighbor nodes of node u. Let Γ′ be the set of p-1 order neighbor nodes of node v. (u) Let u be the set of neighbors with degree 1. Substitutability The following formula is used to calculate:

[0136]

[0137] Among them, L j Therefore, e uv The number of polygons with sides.

[0138] 8b) Calculate the number of p-order branches of node u on edge e using the number of p-order branches of regions U and V. uv Contribution rate:

[0139]

[0140] 8c) According to edge e uv The importance of node u is determined by its importance and contribution rate to the edge.

[0141]

[0142] Among them, Γ (u) Let be the set of neighbors of node u. Based on the importance ranking of nodes in the network, the key nodes of the target ad hoc network are obtained.

[0143] As can be seen, this invention employs machine learning algorithms to automatically identify the MAC protocol of the target network's data link layer using signals intercepted by reconnaissance equipment. It also identifies the source of transmitted and received signals through radiation source identification, thereby reconstructing the network topology and determining key network nodes. The designed cognitive method and system can intelligently analyze the target network's topology and key nodes using physical layer signals. This can provide the decision-making system with targeted interference and attack targets. Furthermore, through the implementation of various attack methods, it can achieve the operational objective of "breaking the link, destroying the network, and paralyzing the system" of the target ad hoc network, thereby improving our operational effectiveness.

[0144] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent cognition in wireless ad hoc networks, characterized in that, Includes the following steps: (1) Obtain the communication signals of the target ad hoc network; (2) Preprocess the communication signal samples, including energy detection, blind signal separation, and data fragmentation; (3) Based on the differences in the start time interval of data packets in different types of MAC protocols, the K-means algorithm and LSTM network are used to identify the MAC protocol categories; (4) Based on step (3), for contention-type MAC protocols, the SVM algorithm is used to identify the specific MAC protocol category of the target ad hoc network based on the theoretical values ​​of frame analogy and duty cycle. (5) Based on the identification of MAC protocol categories, the preprocessed communication signal samples are matched according to the data interaction mode of the MAC protocol; (6) Identify the radiation source for the successfully matched signal segments to obtain the identifiers of the connected node pairs; (7) Perform steps (2) to (6) on all intercepted communication signal samples to obtain the identifiers of all mutually communicating node pairs in the target ad hoc network, and then reconstruct the network topology of the target ad hoc network. (8) Using complex network theory and the reconstructed target self-organizing network topology, calculate the importance of edges and the contribution rate of nodes to edges, thereby obtaining the importance index of each node. (9) Sort the importance of each node in the network to obtain the key nodes in the target self-organizing network.

2. The method as described in claim 1, characterized in that, In step (2), firstly, the energy detection method is used to filter out blank and noise segments in the signal and extract the effective signal portion; secondly, independent component analysis is used to perform blind signal separation on the effective signals from multiple radiation sources; finally, the signal samples obtained after blind signal separation are divided into data segments of equal length, with each segment having a length of N = 2. n .

3. The method as described in claim 2, characterized in that, The specific steps for identifying the MAC protocol type in step (3) include: 3a) First, based on the length difference between the data group and the control group and management group, the K-means algorithm is used to divide the group into two categories: the first category is the data group, and the second category is the control group and management group. 3b) By utilizing the differences in the start time interval of data packets in the three MAC protocols of allocation, contention and hybrid, a pre-trained LSTM network is used to identify the category of the target ad hoc network's MAC protocol.

4. The method as described in claim 3, characterized in that, The specific steps for identifying the target ad hoc network's MAC protocol category as described in step (4) include: 4a) Using the grouping and clustering results in step 3a), the frame ratio and duty cycle of samples using the competitive MAC protocol at the link layer are statistically analyzed and used as feature vectors; 4b) Based on the theoretical values ​​of frame analogy and duty cycle of different competing MAC protocols such as MACA, CSMA / CA, and MACA-BI, the feature vector is input into a pre-trained SVM to obtain the MAC protocol category of the target ad hoc network link layer.

5. The method as described in claim 4, characterized in that, The specific steps for radiation source identification in step (6) include: 6a) Based on multifractal theory, calculate the corresponding singularity index under different powers and dimensions, and construct the feature vector of the signal sample; 6b) After standardizing the feature vectors, input them into the autoencoder to reduce the dimensionality of the feature vectors; 6c) Using low-dimensional feature vectors, a density clustering algorithm based on kernel density estimation is used to classify and identify radiation sources.

6. The method as described in claim 5, characterized in that, Step 6a) specifically refers to: 6a.1 Initialize the exponent q to (2q0+1) values ​​in the interval [-q0, q0]; 6a.2 Under different values ​​of q, the signal segment is divided n times using different line dimensions θ, where the line dimension θ = 2 in the i-th division. i ; 6a.3 Calculate the singularity index for different values ​​of q and θ: Where T = θ / 2 is the size of each signal interval obtained by dividing the signal segment s using the linear dimension θ, and f i q (θ) is the density distribution function: s represents the signal sequence, and s(k) represents the k-th sample point; 6a.4 Under different values ​​of q, the signal segment is divided n times. The calculated singularity index is spliced ​​together, and the resulting multifractal spectrum vector with a length of (2q0+1)·n is the feature vector of the signal.

7. The method as described in claim 5, characterized in that, Step 6c) describes using a density clustering algorithm based on kernel density estimation to classify and identify radiation sources. Specifically, this includes: using the kernel density estimation method and the integral mean square error criterion, based on the variance of the signal samples... Determine the required Eps neighborhood in the density clustering algorithm DBSCAN Minimum number of points Let the signal samples be x1, x2, ... x i ,...,x n ; minPts is the number of sample points existing in a space centered at x and with radius h, where the kernel function K is defined as follows: Where d is the dimension of the sample points, the DBSCA algorithm is then used to identify different signal radiation sources.

8. The method as described in claim 6, characterized in that, Steps (8) and (9) specifically involve obtaining the key nodes in the target ad hoc network through the following steps: 8a) In the reconstructed target ad hoc network topology, edge e uv The network is divided into two regions, U and V. The former is the region closer to node u, and the latter is the region closer to node v. First, the edge e in the network is calculated. uv Importance: in, and They are respectively edge e uv Connectivity and substitutability, where p and k are integers, connectivity The following formula is used to calculate: in, and Let p be the number of p-order branches in regions U and V, respectively. The following conditions must be met: in, Let p be the set of p-order neighbor nodes of node u. Let Γ′ be the set of p-1 order neighbor nodes of node v. (u) Let u be the set of neighbors with degree 1; substitutability The following formula is used to calculate: Among them, L j Therefore, e uv The number of polygons with sides; 8b) Calculate the number of p-order branches of node u on edge e using the number of p-order branches of regions U and V. uv Contribution rate: 8c) According to edge e uv The importance of node u is determined by its importance and contribution rate to the edge. Among them, Γ (u) Let be the set of neighbors of node u; based on the ranking of the importance of each node in the network, the key nodes of the target ad hoc network are obtained.

9. A wireless ad hoc network intelligent cognitive system for implementing the method as described in any one of claims 1 to 8, characterized in that, include: The signal receiving module is used to collect communication signals of the target ad hoc network through a radio monitoring receiver and to provide information on the start time and duration of the signal. The signal preprocessing module is used to perform energy detection on communication signals, extract the effective signal portion, perform blind signal separation on signals from multiple radiation sources, and divide signal samples into data segments of equal length. The MAC protocol identification module is used to identify MAC protocol types based on the differences in the start time interval of data packets of different protocols, as well as the differences in frame analogy and duty cycle, using machine learning algorithms. The radiation source matching module is used to match the physical layer signal segments of the preprocessed communication signal based on the identification of the MAC protocol type and the data interaction mode of the MAC protocol, and to identify the transmitting and receiving nodes of the signal through radiation source identification. The network topology analysis module, based on the MAC protocol identification module and radiation source matching module, reconstructs the network topology by analyzing the communication relationships between nodes in the target network and identifies key nodes in the network, enabling cognitive analysis of the target ad hoc network under non-cooperative communication confrontation.

10. A method for effectively interfering with and attacking a target network based on the method described in any one of claims 1 to 8.

Images