Device fingerprint extraction method and system based on device attributes and passive traffic characteristics

By using a lightweight network model on power IoT devices combined with device attributes and passive traffic characteristics, the problem of high resource requirements of traditional methods is solved, and equipment recognition is achieved with high accuracy and security. The generated fingerprint is unique. and non-forgery, suitable for identity authentication of power IoT devices.

CN118503685BActive Publication Date: 2024-11-12WUXI POWER SUPPLY BRANCH OF STATE GRID JIANGSU ELECTRIC POWER CO LTD +1
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202410962888.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-12
Estimated Expiration
2044-07-18

AI Technical Summary

Technical Problem

Traditional device identification methods require high computing and storage capabilities of power IoT devices, making it difficult to apply on edge devices with limited resources, and the high memory footprint and expensive operating costs of deep learning models limit their application scenarios.

Method used

A lightweight network model based on device attributes and passive traffic characteristics is adopted. Through device attribute feature selection and traffic data preprocessing, multiple MBConv modules and coordinate attention mechanisms are combined to generate a fused feature vector for device identification.

Benefits of technology

It improves the accuracy and security of device identification, and the generated device fingerprint is unique and non-forgery, suitable for security certification of power IoT devices, reducing computing and storage requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118503685B_ABST
    Figure CN118503685B_ABST
Patent Text Reader

Abstract

A device fingerprint extraction method and system based on device attributes and passive traffic characteristics. The method includes collecting device attribute information and passive traffic data of power Internet of Things devices; performing feature selection on the device attribute information to obtain an attribute feature subset, and performing preprocessing operations on the passive traffic data; performing normalized statistical analysis on the information value of the attribute feature subset to generate an attribute feature vector, and using a lightweight network model to perform feature extraction on the preprocessed passive traffic data to generate a traffic feature vector; performing vector splicing on the attribute feature vector and the traffic feature vector to obtain a fused feature vector; performing full-connection layer classification on the fused feature vector, and outputting the device category of the power Internet of Things device. The solution of the present invention improves the uniqueness and unforgeability of device fingerprint information, and meets the requirements of power Internet of Things devices for security and lightweight calculation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of identity authentication, and in particular relates to a method and system for extracting device fingerprints based on device attributes and passive flow characteristics. Background Art

[0002] With the construction of energy Internet, a large number of power Internet of Things devices have been applied to the equipment control, data collection, environmental monitoring and other work of the power grid. On the one hand, this has improved the information and intelligence level of the power grid, but at the same time it has also introduced a large number of information security vulnerabilities, posing a great threat to the safe and stable operation of the power system, such as physical safety, operation safety, and data security. The perception layer of the power Internet of Things has the characteristics of complex node equipment types, large differences in equipment capabilities and resources, and the perception layer devices have a large amount of original data. If they are accessed by illegal nodes and servers, it will lead to data leakage, threatening the information security of users and industries, and increasing management risks. Therefore, the identity recognition of the perception layer devices of the Internet of Things is particularly important.

[0003] Traditional identity recognition technologies include password-based identity recognition and electronic identification card-based identity recognition. These methods are not suitable for IoT business systems due to security defects and high costs. At present, biometric recognition based on fingerprints, voiceprints, irises, etc. has been widely used, with the characteristics of high stability and difficulty in imitation and theft. Device identification is to achieve identity authentication without relying on traditional usernames and passwords by providing identification based on device features. In actual operation, the specific output of the system may vary depending on the actual security policy and configuration. For example, some systems may use machine learning algorithms to improve the accuracy of device identification and automatically adjust the response measures to legal and illegal devices. The device fingerprint-based device identification method has low requirements on the computing and storage capabilities of the device itself and is more suitable for the power IoT environment.

[0004] The traditional device fingerprint identification method mainly uses the traffic generated during the communication between the terminal device and the server as the device fingerprint, and uses the uniqueness of passive traffic characteristics to identify the device. However, power IoT devices also have attribute characteristics, such as device serial number, device general parameters, device communication parameters, device hardware parameters, device software version number, embedded module identification number, etc. In order to fully extract the characteristics of device fingerprints, the device attribute characteristics and the traffic generated during the communication process are used as device fingerprints for device identification. This can obtain a more comprehensive feature set, improve the accuracy of device identification, and improve the identity authentication mechanism.

[0005] Traditional device identification methods mainly use deep models such as Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) to extract features from traffic data. Most deep learning models have a memory footprint of up to gigabytes and high operating costs, making them difficult to apply to edge devices with limited resources, which severely limits the application scenarios of deep learning technology. Summary of the invention

[0006] In order to solve the deficiencies in the prior art, the present invention provides a device fingerprint extraction method and system based on device attributes and passive traffic characteristics to solve the technical problem that power devices in the power Internet of Things system have limited computing and storage capabilities.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions.

[0008] The present invention first discloses a device fingerprint extraction method based on device attributes and passive traffic characteristics, the method comprising the following steps:

[0009] Step 1: Collect device attribute information and passive flow data of power Internet of Things devices;

[0010] Step 2, performing feature selection on the device attribute information to obtain an attribute feature subset, and performing a preprocessing operation on the passive traffic data;

[0011] Step 3, performing normalized statistical analysis on the information value of the attribute feature subset to generate an attribute feature vector, and using a lightweight network model to extract features from the preprocessed passive traffic data to generate a traffic feature vector;

[0012] Step 4, concatenating the attribute feature vector and the flow feature vector to obtain a fused feature vector;

[0013] Step 5: Perform full-connection layer classification on the fused feature vector and output the device category of the power Internet of Things device.

[0014] The present invention further includes the following preferred embodiments:

[0015] The collecting of device attribute information of the power Internet of Things device further includes:

[0016] The attribute information of each power Internet of Things terminal device is obtained by device scanning, and the attribute information includes device serial number, device general parameters, device communication parameters and device hardware parameters.

[0017] The collecting of passive flow data of electric power Internet of Things devices further includes:

[0018] Use network monitoring tools installed on network devices or servers to capture network traffic data generated by power IoT devices during communication.

[0019] The performing feature selection on the device attribute information to obtain an attribute feature subset further includes:

[0020] Step 2.1.1: Get the device attribute feature set ,in Indicates the number of features to initialize the label set: ,and ,in Indicates the number of label values ​​in the label set;

[0021] set up for Any feature in the set of labels, with uniqueness, unforgeability, and stability as the label set , then calculate the feature and tag collection The mutual information is:

[0022] ,

[0023] in Representation characteristics The number of eigenvalues ​​of is a discrete probability density function, which means The probability of occurrence in the data set; is the eigenvalue The probability of appearing in the data set, that is, the probability density function ; Yes Tags The probability of appearing in the data set; is the eigenvalue and tags The probability of simultaneous occurrence; i, j are feature numbers, k is the label number;

[0024] Mutual Information Arrange in descending order and select multiple features with the largest mutual information value as expert features ;

[0025] Step 2.1.2: Combine the expert feature set with the remaining feature set as a condition and calculate the mutual information of the two feature sets And arrange them in ascending order, select multiple features with the smallest mutual information value as the auxiliary feature set, is any feature in the expert feature set, is any feature in the remaining feature set;

[0026] The expert feature set is merged with the auxiliary feature set to form the final attribute feature subset.

[0027] The passive traffic data is preprocessed, further comprising:

[0028] Segment the original traffic based on the session;

[0029] Anonymize IP address and MAC address information;

[0030] The overlong part of the session data is truncated, and the missing part is filled with 0 to unify the session data into a fixed size, and the traffic data is converted into a grayscale image.

[0031] The lightweight network model is formed by stacking multiple MBConv modules, and the MBConv modules sequentially include:

[0032] The first convolutional layer contains the batch normalization module and Activation function layer;

[0033] Depthwise separable convolutional layers, including BN modules and Activation function layer, where k is the depth of the convolutional layer, Includes two structures: 5×5 and 3×3;

[0034] Coordinate attention mechanism CA module;

[0035] The second convolutional layer includes a BN module and a linear activation function layer. ;

[0036] Dropout layer.

[0037] The performing full connection layer classification on the fused feature vector further includes:

[0038] use Activation function for classification:

[0039] ,

[0040] in, is the first elements, is the dimension of the input vector.

[0041] The present invention also discloses a device fingerprint extraction system based on device attributes and passive traffic characteristics using the device fingerprint extraction method based on device attributes and passive traffic characteristics, comprising:

[0042] Equipment information collection module, used to collect equipment attribute information and passive flow data of power Internet of Things equipment;

[0043] A feature screening preprocessing module, used to perform feature selection on the device attribute information, obtain an attribute feature subset, and perform preprocessing operations on the passive traffic data;

[0044] A feature vector generation module is used to perform normalized statistical analysis on the information value of the attribute feature subset to generate an attribute feature vector, and to extract features from the pre-processed passive traffic data using a lightweight network model to generate a traffic feature vector;

[0045] A feature fusion module, used for concatenating the attribute feature vector and the flow feature vector to obtain a fused feature vector;

[0046] The device category identification module is used to perform full-connection layer classification on the fused feature vector and output the device category of the power Internet of Things device.

[0047] Accordingly, the present application also discloses a terminal, including a processor and a storage medium;

[0048] The storage medium is used to store instructions;

[0049] The processor is used to operate according to the instructions to execute the steps of the device fingerprint extraction method based on device attributes and passive traffic characteristics.

[0050] Accordingly, the present application also discloses a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the aforementioned device fingerprint extraction method based on device attributes and passive traffic characteristics are implemented.

[0051] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a device fingerprint extraction method and system based on device attributes and passive traffic characteristics, utilizes the device attribute characteristics of the power Internet of Things device and the passive traffic characteristics for feature fusion, and combines the attribute characteristics of the power equipment to obtain a more comprehensive feature set, generates a unique device fingerprint, and thus improves the device recognition rate, and improves the security of the power Internet of Things edge devices. The accuracy of device identification is improved, and the identity authentication mechanism is improved. In view of the problem of limited computing power and storage capacity of the power Internet of Things edge devices, a lightweight network model is used to extract features from traffic data. The fingerprint information generated by the present invention has the characteristics of uniqueness and non-forgeability, which can meet the requirements of power Internet of Things devices for security and lightweight computing, is conducive to data privacy protection, and provides a data basis for identity identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1It is a flow chart of the device fingerprint extraction method based on device attributes and passive traffic characteristics in the present invention.

[0053] Figure 2 It is a system framework diagram of the method of the present invention.

[0054] Figure 3 This is a diagram of a lightweight network model described in the method of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention more clear, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0056] The embodiments described in this application are only some embodiments of the present invention, not all embodiments. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the protection scope of the present invention.

[0057] In view of the deficiencies of the prior art, the present invention proposes a device fingerprint extraction method and system based on device attributes and passive traffic characteristics, which utilizes the attribute characteristics of the power Internet of Things device and the passive traffic characteristics for feature fusion to generate a unique device fingerprint, fully utilizes the device information in the power Internet of Things system, obtains a more comprehensive feature set, and thus improves the device recognition rate and improves the security of the power Internet of Things system. In view of the problem of limited computing power and storage capacity of the edge devices of the power Internet of Things, the present invention also lightweights the deep learning model part, making the method more suitable for the power Internet of Things system. The fingerprint information generated by the present invention has the characteristics of uniqueness and non-forgeability, which can meet the requirements of the power Internet of Things devices for security and lightweight computing, is conducive to data privacy protection, and improves the accuracy of device identification.

[0058] See also Figure 1 As shown, the device fingerprint extraction method based on device attributes and passive traffic characteristics disclosed in the present invention includes the following steps:

[0059] Step 1: Collect device attribute information and passive flow data of power IoT devices.

[0060] The step 1 further comprises:

[0061] Step 1.1: Collecting power equipment attribute information:

[0062] Obtain device attribute information through device scanning. Each power IoT terminal device has unique attribute information, such as device serial number / ID, device general parameters, device communication parameters, device hardware parameters and other attribute information.

[0063] Step 1.2: Passive traffic information collection:

[0064] Power IoT devices generate network traffic data during communication. Use network monitoring tools to capture data packets on the network. These tools can be installed on network devices (such as routers, switches) or servers to monitor and record network traffic. When a new device communicates with the gateway, the gateway records the device's setup phase. Communication data flow packets.

[0065] Step 2: Perform feature selection on the device attribute information to obtain an attribute feature subset, and perform preprocessing operations on the passive traffic data.

[0066] The step 2 further comprises:

[0067] Step 2.1: Perform feature selection on the device attribute information to obtain an attribute feature subset.

[0068] The feature selection method based on the expert information weighting mechanism deletes redundant or irrelevant features in the device attribute information set, and selects device attributes with uniqueness, unforgeability and stability as the attribute feature subset.

[0069] The feature selection method of the expert information weighting mechanism further comprises:

[0070] Step 2.1.1: Get the device attribute feature set ,in Indicates the number of features to initialize the label set: ,and ,in Indicates the number of label values ​​in the label set. for Any feature in Representation characteristics The number of eigenvalues ​​of is a discrete probability density function, which means The probability of occurrence in the dataset. is the eigenvalue The probability of appearing in a data set (i.e., the probability density function ). Yes Tags The probability of appearing in the data set. is the eigenvalue and tags The probability of co-occurrence. i, j are feature numbers, and k is the label number.

[0071] set up for Any feature in the set of labels, with uniqueness, unforgeability, and stability as the label set , then the feature and tag collection The mutual information is defined as:

[0072] ,

[0073] The mutual information , is the amount of information contained in one random variable about another random variable. The larger the mutual information, the more important the feature is, and vice versa.

[0074] Mutual Information Arrange in descending order and select multiple features with the largest mutual information value as expert features .

[0075] Step 2.1.2: Use the expert feature set as condition and remaining features Combined, calculate the mutual information of these two feature sets , is any feature in the expert feature set, For any feature in the remaining feature set, arrange them in ascending order, select multiple features with the smallest mutual information value, reduce the redundancy of the features, and use them as auxiliary feature sets to avoid repeated calculations and improve the speed of feature selection.

[0076] The expert feature set is merged with the auxiliary feature set to form the final attribute feature subset for subsequent model training or other analysis tasks.

[0077] The attribute feature subset obtained by this method is: device / module unique serial number, common parameters, product test serial number and embedded module operation status. This subset covers as much category information as possible while reducing redundancy.

[0078] Step 2.2: Preprocess the passive traffic data, including:

[0079] Step 2.2.1, when dividing the original traffic, data segmentation is performed according to the flow or session method. In a preferred embodiment, the traffic is segmented based on the session method;

[0080] Step 2.2.2, in the data cleaning stage, anonymize the IP address, MAC address and other information to prevent the classification network from overfitting on this part of information;

[0081] Step 2.2.3: The session data is unified into a fixed size by truncating the overlong part and filling the missing part with 0. After being fixed to a uniform size (784 bytes), the traffic data is converted into a 28×28 grayscale image.

[0082] Step 3: Perform normalized statistical analysis on the information value of the attribute feature subset to generate an attribute feature vector, and use a lightweight network model to extract features from the preprocessed passive traffic data to generate a traffic feature vector.

[0083] The step 3 further comprises:

[0084] Perform statistical numerical analysis on the information values ​​in the feature subset, use the Max-Min normalization method to normalize the statistical information, distribute all statistical values ​​in the range of [0, 1], and generate attribute feature vectors. Use the lightweight network model to extract traffic features from the preprocessed traffic data and generate traffic feature vectors.

[0085] The lightweight network model is implemented by an improved EfficientNet model. The EfficientNet model is composed of multiple MBConv modules stacked together. The compression and excitation (SE) module in the MBConv module enhances the model's expressiveness by modeling the relationship between channels. However, there is no strong correlation between traffic data channels, and reliable channel feature relationships cannot be established. At the same time, the characteristics of traffic determine that the payload information between traffic data has a strong correlation with the field position. Position information is crucial for traffic data. However, the global pooling operation of the SE module causes the loss of spatial position information when modeling channel relationships. In response to the above problems, the MBConv module of the EfficientNet model is optimized, and the compression and excitation module (SE) in the MBConv module is improved in spatial scale to obtain an attention mechanism that integrates precise spatial information: the coordinate attention (CA) module, which can more fully extract traffic features. The SE module is an attention mechanism module in the MBConv module of the EfficientNet model, which enhances the model's expressiveness by modeling the relationship between channels.

[0086] The improved MBConv module is as follows Figure 3 As shown, it includes:

[0087] The first convolutional layer includes the BN (Batch Normalization) module and Activation function layer, used for feature dimension increase;

[0088] Depthwise separable convolutional layers, including BN modules and Activation function layer, where k is the depth of the convolutional layer, It includes two structures, 5×5 and 3×3, for extracting features;

[0089] CA module, used to enhance feature representation capabilities;

[0090] The second convolutional layer includes a BN module and a linear activation function layer. , used for dimensionality reduction;

[0091] Dropout layer is used to prevent overfitting.

[0092] Wherein, the BN module further comprises:

[0093] The normalization layer is used to calculate the mean and variance of each mini-batch of data, and then use these statistics to normalize the data so that the normalized data has zero mean and unit variance;

[0094] The scaling and shifting layers are used to scale and shift the normalized data. This is done by two learnable parameters (expressed as and ) to ensure that the network can recover the original distribution features that may be useful for specific tasks.

[0095] The forward propagation layer is used to apply the above normalization, scaling and offset operations during the forward propagation process, and then pass the processed data to the next layer.

[0096] The swish activation function is expressed as follows:

[0097] ,

[0098] in, is the input of the function, yes Function, that is:

[0099] ,

[0100] The depth-wise separable convolution layer is used to decompose the standard convolution into two steps: depth-wise convolution and point-wise convolution (also called 1×1 convolution). In EfficientNet, the depth-wise separable convolution includes the following process to further reduce the amount of computation:

[0101] Step 3.1, perform convolution operation on each input channel separately. If the input has channels, the depth convolution generates Different feature maps, each corresponding to one input channel. This operation reduces the number of parameters because the weight of each channel is only related to itself, not to all channels.

[0102] Step 3.2: generate the deep convolution The feature maps are combined together and point-by-point convolution is performed based on a 1×1 convolution kernel through point-by-point multiplication to increase the depth of the model. 1×1 convolution allows the model to increase the capacity of the network without increasing the spatial dimension (i.e., without changing the width and height of the feature map).

[0103] The CA module can consider the attention in both the channel dimension and the spatial dimension at the same time, and can make the model pay more attention to useful channel information by learning adaptive channel weights. The specific steps are:

[0104] Step 3.3, assuming The number of channels, height and width of the input feature map are respectively. Decompose the global average pooling of the SE module to generate a one-dimensional direction-aware attention feature map on X and Y:

[0105] ,

[0106] ,

[0107] ,

[0108] in, is the output, It is The two-dimensional feature set of the channel, and represents the coordinates on the output feature, and The The convolution kernel weights of the channel.

[0109] Step 3.4: Concatenate the one-dimensional feature maps in the X and Y directions and input them into the convolution transformation function F1 to generate an intermediate feature map containing spatial information in the horizontal and vertical directions. :

[0110] ,

[0111] in, is a one-dimensional feature map in the horizontal direction, is a one-dimensional feature map in the vertical direction, is the spatial dimension concatenation operation, , is a nonlinear activation function, is a hyperparameter that controls the module size.

[0112] Step 3.5, along the spatial dimension Decomposed into 2 separate tensors and . Then use the convolution transformation function and Respectively and Transform to a tensor with the same number of channels as the input features:

[0113] ,

[0114] ,

[0115] in, for function.

[0116] Step 3.6, expand the input feature map to obtain the output features of the coordinate attention module:

[0117] ,

[0118] in, and It is divided into two parts, which represent the and .

[0119] The Dropout layer is used to randomly discard or temporarily remove some neurons and their connections in the network during the training process. For each training sample, a random number in the interval [0, 1] is generated for each neuron. If this random number is less than a preset drop probability (usually expressed as p), the neuron is discarded in this training iteration, that is, its output is set to 0.

[0120] Step 4: Concatenate the attribute feature vector and the traffic feature vector to obtain a fused feature vector.

[0121] The attribute feature vector and the flow feature vector are expanded, and one-dimensional feature vector splicing is performed to obtain a fused feature vector with more comprehensive feature information, which serves as the device fingerprint required in the identification process of the power Internet of Things device.

[0122] Step 5: Perform full-connection layer classification on the fused feature vector and output the device category.

[0123] According to the fused feature vector, classification is performed using a fully connected layer to output the category of the power equipment.

[0124] The output of the fully connected layer is passed through Activation function for classification:

[0125] ,

[0126] in, is the first elements, is the dimension of the input vector. The function is usually used in the output layer of multi-classification problems, which can output the predicted probability of each category.

[0127] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a device fingerprint extraction method and system based on device attributes and passive traffic characteristics, utilizes the device attribute characteristics of the power Internet of Things device and the passive traffic characteristics for feature fusion, and combines the attribute characteristics of the power equipment to obtain a more comprehensive feature set, generates a unique device fingerprint, and thus improves the device recognition rate, and improves the security of the power Internet of Things edge devices. The accuracy of device identification is improved, and the identity authentication mechanism is improved. In view of the problem of limited computing power and storage capacity of the power Internet of Things edge devices, a lightweight network model is used to extract features from traffic data. The fingerprint information generated by the present invention has the characteristics of uniqueness and non-forgeability, which can meet the requirements of power Internet of Things devices for security and lightweight computing, is conducive to data privacy protection, and provides a data basis for identity identification.

[0128] The present invention may be a system, a method and / or a computer program product. The present invention also discloses a device fingerprint extraction system based on device attributes and passive traffic characteristics based on the aforementioned device fingerprint extraction method based on device attributes and passive traffic characteristics, comprising:

[0129] Equipment information collection module, used to collect equipment attribute information and passive flow data of power Internet of Things equipment;

[0130] A feature screening preprocessing module, used to perform feature selection on the device attribute information, obtain an attribute feature subset, and perform preprocessing operations on the passive traffic data;

[0131] A feature vector generation module is used to perform normalized statistical analysis on the information value of the attribute feature subset to generate an attribute feature vector, and to extract features from the pre-processed passive traffic data using a lightweight network model to generate a traffic feature vector;

[0132] A feature fusion module, used for concatenating the attribute feature vector and the flow feature vector to obtain a fused feature vector;

[0133] The device category identification module is used to perform full-connection layer classification on the fused feature vector and output the device category of the power Internet of Things device.

[0134] Based on the spirit of the present invention, those skilled in the art can easily think that a computer program product can be obtained based on the aforementioned device fingerprint extraction method based on device attributes and passive traffic characteristics. The computer program product may include a computer-readable storage medium, which carries computer-readable program instructions for enabling a processor to implement various aspects of the present disclosure. That is, the present application also includes a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the aforementioned device fingerprint extraction method based on device attributes and passive traffic characteristics.

[0135] Computer readable storage medium can be a tangible device that can keep and store the instructions used by the instruction execution device. Computer readable storage medium can be, for example, - but not limited to - electrical storage device, magnetic storage device, optical storage device, electromagnetic storage device, semiconductor storage device or any suitable combination of the above. More specific examples (non-exhaustive list) of computer readable storage medium include: portable computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, such as punch card or groove protrusion structure with instructions stored thereon and any suitable combination of the above. Computer readable storage medium used here is not interpreted as instantaneous signal itself, such as radio wave or other free propagating electromagnetic wave, electromagnetic wave propagated by waveguide or other transmission medium (for example, light pulse by optical fiber cable) or electrical signal transmitted by wire.

[0136] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0137] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA) or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions, thereby implementing various aspects of the present disclosure.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A device fingerprint extraction method based on device attributes and passive traffic characteristics, characterized in that: The following steps are involved: Step 1: Collect device attribute information and passive flow data of power Internet of Things devices; Step 2, performing feature selection on the device attribute information to obtain an attribute feature subset, and performing a preprocessing operation on the passive traffic data; Step 3, performing normalized statistical analysis on the information value of the attribute feature subset to generate an attribute feature vector, and using a lightweight network model to extract features from the preprocessed passive traffic data to generate a traffic feature vector; Step 4, concatenating the attribute feature vector and the flow feature vector to obtain a fused feature vector; Step 5, performing full connection layer classification on the fused feature vector, and outputting the device category of the power Internet of Things device; The collecting of device attribute information of the power Internet of Things device further includes: Obtain attribute information of each power Internet of Things terminal device by device scanning, wherein the attribute information includes device serial number, device general parameters, device communication parameters and device hardware parameters; The collecting of passive flow data of electric power Internet of Things devices further includes: Use network monitoring tools installed on network devices or servers to capture network traffic data generated by power IoT devices during communication; The performing feature selection on the device attribute information to obtain an attribute feature subset further includes: Step 2.1.1: Get the device attribute feature set ,in Represents the number of features; the initial label set is ,and ,in Indicates the number of label values ​​in the label set; set up for Any feature in the set of labels, with uniqueness, unforgeability, and stability as the label set , then calculate the feature and tag collection The mutual information is: in Representation characteristics The number of eigenvalues ​​of is a discrete probability density function, which means The probability of occurrence in the data set; is the eigenvalue The probability of appearing in the data set, that is, the probability density function ; Yes Tags The probability of appearing in the data set; is the eigenvalue and tags The probability of simultaneous occurrence; i, j are feature numbers, k is the label number; Mutual Information Arrange in descending order and select multiple features with the largest mutual information value as expert features ; Step 2.1.2: Combine the expert feature set with the remaining feature set as a condition and calculate the mutual information of the two feature sets And arrange them in ascending order, select multiple features with the smallest mutual information value as the auxiliary feature set, is any feature in the expert feature set, is any feature in the remaining feature set; Merge the expert feature set and the auxiliary feature set to form the final attribute feature subset; The passive traffic data is preprocessed, further comprising: Segment the original traffic based on the session; Anonymize IP address and MAC address information; The overlong part of the session data is truncated, and the missing part is filled with 0 to unify the session data into a fixed size, and the traffic data is converted into a grayscale image; The lightweight network model is formed by stacking multiple MBConv modules, and the MBConv modules sequentially include: The first convolutional layer contains the batch normalization module and Activation function layer; Depthwise separable convolutional layers, including BN modules and Activation function layer, where k is the depth of the convolutional layer, Includes two structures: 5×5 and 3×3; Coordinate attention mechanism CA module; The second convolutional layer includes a BN module and a linear activation function layer. ; Dropout layer.

2. The device fingerprint extraction method based on device attributes and passive traffic characteristics according to claim 1 is characterized in that: The performing full connection layer classification on the fused feature vector further includes: use Activation function for classification: in, is the first elements, is the dimension of the input vector.

3. A device fingerprint extraction system based on device attributes and passive traffic characteristics, characterized in that: include: Equipment information collection module, used to collect equipment attribute information and passive flow data of power Internet of Things equipment; A feature screening preprocessing module, used to perform feature selection on the device attribute information, obtain an attribute feature subset, and perform preprocessing operations on the passive traffic data; A feature vector generation module is used to perform normalized statistical analysis on the information value of the attribute feature subset to generate an attribute feature vector, and to extract features from the preprocessed passive traffic data using a lightweight network model to generate a traffic feature vector; A feature fusion module, used for concatenating the attribute feature vector and the flow feature vector to obtain a fused feature vector; A device category identification module, used for performing full-connection layer classification on the fused feature vector and outputting the device category of the power Internet of Things device; The collecting of device attribute information of the power Internet of Things device further includes: Obtain attribute information of each power Internet of Things terminal device by device scanning, wherein the attribute information includes device serial number, device general parameters, device communication parameters and device hardware parameters; The collecting of passive flow data of electric power Internet of Things devices further includes: Use network monitoring tools installed on network devices or servers to capture network traffic data generated by power IoT devices during communication; The performing feature selection on the device attribute information to obtain an attribute feature subset further includes: Step 2.1.1: Get the device attribute feature set ,in Represents the number of features, and the initial label set is ,and ,in Indicates the number of label values ​​in the label set; set up for Any feature in the set of labels, with uniqueness, unforgeability, and stability as the label set , then calculate the feature and tag collection The mutual information is: in Representation characteristics The number of eigenvalues ​​of is a discrete probability density function, which means The probability of occurrence in the data set; is the eigenvalue The probability of appearing in the data set, that is, the probability density function ; Yes Tags The probability of appearing in the data set; is the eigenvalue and tags The probability of simultaneous occurrence; i, j are feature numbers, k is the label number; Mutual Information Arrange in descending order and select multiple features with the largest mutual information value as expert features ; Step 2.1.2: Combine the expert feature set with the remaining feature set as a condition and calculate the mutual information of the two feature sets And arrange them in ascending order, select multiple features with the smallest mutual information value as the auxiliary feature set, is any feature in the expert feature set, is any feature in the remaining feature set; Merge the expert feature set and the auxiliary feature set to form the final attribute feature subset; The passive traffic data is preprocessed, further comprising: Segment the original traffic based on the session; Anonymize IP address and MAC address information; The overlong part of the session data is truncated, and the missing part is filled with 0 to unify the session data into a fixed size, and the traffic data is converted into a grayscale image; The lightweight network model is formed by stacking multiple MBConv modules, and the MBConv modules sequentially include: The first convolutional layer contains the batch normalization module and Activation function layer; Depthwise separable convolutional layers, including BN modules and Activation function layer, where k is the depth of the convolutional layer, Includes two structures: 5×5 and 3×3; Coordinate attention mechanism CA module; The second convolutional layer includes a BN module and a linear activation function layer. ; Dropout layer.

4. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is used to operate according to the instruction to execute the steps of the device fingerprint extraction method based on device attributes and passive traffic characteristics according to any one of claims 1-2.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the device fingerprint extraction method based on device attributes and passive traffic characteristics described in any one of claims 1-2 are implemented.

Citation Information

Patent Citations

  • Internet of Things safety monitoring method and system based on double engines

    CN111163115A

  • Wireless equipment fingerprint identification method and system, equipment and readable storage medium

    CN111385297A

  • Internet-of-Things equipment fingerprint identification method based on deep learning

    CN112564974A

  • Precise identification method and system for intelligent internet-of-things terminal operating system and application software

    CN116074056A

  • Internet of Things equipment identification method and system based on multi-scale attention capsule network

    CN116958586A