Internet of Things equipment identification method and device
By extracting the abnormal characteristics and length characteristics of IoT device data packets, building multivariable timing characteristics, and using the Inception module and residual structure device recognition model, the problem of timing noise interference in the prior art is solved, and higher recognition accuracy and adaptability are achieved.
Patent Information
- Application Number
- CN202510217264.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-26
- Publication Date
- 2025-06-06
- Estimated Expiration
- 2045-02-26
AI Technical Summary
Existing IoT device identification methods do not consider timing noise interference, resulting in low recognition accuracy.
By extracting the abnormal characteristics and length characteristics of IoT device data packets, multivariate timing characteristics are constructed and inputted into a trained device recognition model, including Inception module and residual structure, to improve recognition accuracy.
It improves the accuracy of IoT device recognition, enhances the ability to adapt to complex network environments and encrypted traffic, and meets the needs of efficient and accurate identification.
Smart Images

Figure CN120105197A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of Internet of Things device identification, and in particular to an Internet of Things device identification method and device. Background Art
[0002] The Internet of Things (IoT) technology interconnects various devices to realize a network system for data exchange and intelligent control. This important technology has been widely embedded and deployed in various consumer applications. As a result, the consumer-centric Internet of Things (CIoT) came into being and covers a variety of application scenarios such as smart home, smart health, and smart grid. CIoT has become an important force in the fifth industrial revolution (Industry 5.0) and the digital transformation of smart cities. With the continuous development of technology, CIoT will further promote the intelligent upgrading of various industries and promote the sustainable development of social economy. CIoT interacts more directly with end users, so it often has a large amount of user data. Especially in smart home scenarios, the home data collected and processed by CIoT devices (such as smart speakers, smart door locks, surveillance cameras, etc.) is highly private, including users' daily behavior habits, home environment information, and security monitoring data. However, the security risks exposed by these devices may make them the target of cyber attacks, resulting in the leakage and abuse of family privacy data. For example, attackers may obtain the real-time location of family members, monitoring images, or even control smart door locks by invading devices, and then carry out theft or other illegal activities.
[0003] Therefore, effective protection and supervision measures are urgently needed to deal with the escalating network threats. IoT device identification, as a network management tool, has become an important research direction. It can help manage and configure devices, detect and isolate vulnerable devices, and take control measures to prevent security threats from expanding when devices are abnormal. For example, by identifying the type, behavior pattern, and network traffic of the device, abnormal devices can be quickly discovered and their access rights can be restricted, thereby reducing potential security risks. In recent years, researchers at home and abroad have conducted a lot of research on IoT device identification. However, the existing identification methods still have shortcomings: timing noise interference is not considered. That is, the existing timing-based methods only use the length attribute of the data packet to identify IoT devices through a single-dimensional parameter, without considering the network quality and the noise superposition of encrypted data on the timing. The fingerprint of the traffic length pattern will be affected by these factors.
[0004] Therefore, there is an urgent need to provide an IoT device identification method and apparatus that takes into account the noise superposition effect of network quality and encrypted data on timing, so as to improve the recognition accuracy when identifying IoT devices. Summary of the invention
[0005] In view of this, it is necessary to provide a method and device for identifying an IoT device to solve the technical problem in the prior art that timing noise interference is not considered, resulting in a low recognition accuracy of the IoT device.
[0006] On the one hand, in order to solve the above technical problems, the present invention provides an IoT device identification method, comprising:
[0007] Extracting abnormal features and length features of IoT device data packets, and constructing multivariate time series features based on the abnormal features and the length features;
[0008] Inputting the multivariate time series features into a fully trained device recognition model to obtain recognition results of IoT devices;
[0009] The device recognition model includes a first Inception module, a first residual structure, a second Inception module, a second residual structure, a global average pooling layer and a fully connected layer.
[0010] In a possible implementation, the abnormal features include abnormal level, abnormal grouping and top-level protocol type.
[0011] In a possible implementation, extracting the length feature of the IoT device data packet includes:
[0012] Determine outbound traffic and inbound traffic in the IoT device data packet;
[0013] The outbound flow and the inbound flow are concatenated in chronological order to obtain a target flow, and the length of the target flow is used as the length feature.
[0014] In a possible implementation, determining the outbound traffic and the inbound traffic in the IoT device data packet includes:
[0015] Obtaining an original data stream; the original data stream includes multiple data packets related to multiple IoT devices;
[0016] Based on the binary information, determine a first data packet whose source MAC address is a target IoT device and a second data packet whose destination MAC address is a target IoT address among the multiple data packets, use the first data packet as the outbound traffic, and use the second data packet as the inbound traffic.
[0017] In a possible implementation, constructing a multivariate time series feature based on the abnormal feature and the length feature includes:
[0018] Dividing the IoT device data packet based on a preset sequence length to obtain multiple time series features;
[0019] Filling the time series features of each segment according to a preset filling rule to obtain the multivariate time series features;
[0020] Among them, the number of rows of the multivariate time series feature is the same as the number of types of the abnormal features and the time series features, and the number of columns of the multivariate time series feature is the same as the number of segments of the multi-segment time series feature.
[0021] In a possible implementation, the first Inception module and the second Inception module each include a first Inception structure, a second Inception structure, and a third Inception structure connected in sequence; the first Inception structure, the second Inception structure, and the third Inception structure each include a feature compression unit, a multi-scale convolution unit, and a feature fusion unit;
[0022] The feature compression unit is used to compress the input feature to obtain a compressed feature;
[0023] The multi-scale convolution unit is used to perform multi-scale convolution on the input features to obtain multiple convolution features of different scales;
[0024] The feature fusion unit is used to fuse the compression feature and the multiple convolution features.
[0025] In a possible implementation, the feature compression unit includes a first global average pooling layer and a first convolutional layer connected in sequence.
[0026] In a possible implementation, the multi-scale convolution unit includes a bottleneck layer and a second convolution layer, a third convolution layer and a fourth convolution layer connected to the bottleneck layer, and the convolution kernel sizes of the second convolution layer, the third convolution layer and the fourth convolution layer are all different.
[0027] In a possible implementation, the feature fusion unit includes a deep convolution layer, a batch normalization layer, and an activation function layer connected in sequence.
[0028] On the other hand, the present invention also provides an IoT device identification device, comprising:
[0029] A multi-dimensional feature extraction unit, used to extract abnormal features and length features of IoT device data packets, and construct multi-variable time series features based on the abnormal features and the length features;
[0030] A device identification unit, used to input the multivariate time series features into a fully trained device identification model to obtain an identification result of the IoT device;
[0031] The device recognition model includes a first Inception module, a first residual structure, a second Inception module, a second residual structure, a global average pooling layer and a fully connected layer.
[0032] The beneficial effect of the present invention is that the IoT device identification method provided by the present invention can characterize the influence of the network environment and encrypted traffic on the flow length sequence pattern of the IoT device data packet by extracting the abnormal features other than the length features of the IoT device data packet, that is, the abnormal features can be used to characterize the interference of timing noise. In other words, the influence of timing noise can be considered in the process of identifying IoT devices, thereby improving the accuracy of identifying IoT devices. In addition, the adaptability of the IoT device identification method to complex network environments can be greatly improved to meet the identification requirements for encrypted traffic.
[0033] Furthermore, the present invention can further explore the network behaviors of IoT devices at different times and scales by setting the device recognition model to include the Inception module, thereby further improving the recognition accuracy. Furthermore, the present invention can solve the gradient vanishing problem between Inception modules by setting the device recognition model to include the residual module, thereby further improving the recognition accuracy of the device recognition model.
[0034] In summary, the present invention can achieve efficient and accurate identification of IoT devices, lay the foundation for subsequent security management and maintenance, and ensure the safe operation of devices in the IoT environment. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.
[0036] Figure 1 A schematic diagram of an embodiment of the method for identifying an Internet of Things device provided by the present invention;
[0037] Figure 2 A schematic diagram of the structure of an embodiment of the device identification model provided by the present invention;
[0038] Figure 3 For the present invention Figure 1 A schematic flow chart of an embodiment of extracting length features in step S101;
[0039] Figure 4 For the present invention Figure 3A schematic flow chart of an embodiment of S301;
[0040] Figure 5 For the present invention Figure 1 A schematic diagram of an embodiment of constructing multivariate time series features in step S101;
[0041] Figure 6 A schematic diagram of the structure of an embodiment of the first Inception module provided by the present invention;
[0042] Figure 7 A comparison chart of evaluation indicators between the embodiment of the present invention and the existing recognition method;
[0043] Figure 8 A time comparison diagram of the embodiment of the present invention and the existing method in the feature extraction and recognition stages;
[0044] Fig. 9 This is a comparison chart of recognition performance under different time windows in an embodiment of the present invention;
[0045] Fig.10 Comparison chart of accuracy indicators of unidirectional length features and bidirectional length features on the UNSW dataset;
[0046] Fig.11 This is a comparison chart of the accuracy indicators of unidirectional length features and bidirectional length features on the CICIoT dataset;
[0047] Fig.12 A schematic diagram of the structure of an embodiment of the Internet of Things device identification device provided by the present invention. DETAILED DESCRIPTION
[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0049] It should be understood that the schematic drawings are not drawn to scale. The flowchart used in the present invention shows the operations implemented according to some embodiments of the present invention. It should be understood that the operations of the flowchart can be implemented out of order, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art, under the guidance of the content of the present invention, can add one or more other operations to the flowchart, and can also remove one or more operations from the flowchart. Some of the block diagrams shown in the accompanying drawings are functional entities, which do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor systems and / or microcontroller systems.
[0050] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0051] The present invention provides an Internet of Things device identification method and device, which are described below respectively.
[0052] Figure 1 A schematic diagram of an embodiment of the method for identifying an IoT device provided by the present invention is shown in FIG. Figure 1 As shown, the IoT device identification method includes:
[0053] S101. Extract abnormal features and length features of IoT device data packets, and construct multivariate time series features based on the abnormal features and length features.
[0054] Among them, abnormal features and length features constitute the device fingerprint of IoT devices.
[0055] It should be noted that the IoT device data packet refers to the data packet corresponding to a specified IoT device, that is, the data packet corresponding to a certain IoT device, rather than the data packets of multiple IoT devices.
[0056] Since IoT device data packets are time series data, IoT device data packets refer to data packets within a specified time interval.
[0057] Among them, the method of extracting the abnormal features of the IoT device data packet is specifically: using the tshark tool to obtain the expert information features of each data packet, and through manual analysis and aggregation, finally obtaining the abnormal features of the data packet.
[0058] Expert Info is the advanced diagnostic information about potential problems or anomalies that Wireshark provides after analyzing captured network packets. This information is generated by Wireshark's built-in parser and protocol analysis engine to help users quickly identify problems or abnormal behavior in network communications.
[0059] S102: Input the multivariate time series features into a fully trained device recognition model to obtain recognition results of IoT devices.
[0060] It should be understood that before the device recognition model is used, it needs to be trained, tested and verified based on a sample set to ensure the recognition performance of the device recognition model.
[0061] In a specific embodiment of the present invention, the loss function used by the device identification model during the training process is a cross entropy loss function. Specifically, the formula of the cross entropy loss function is:
[0062]
[0063] In the formula, N represents the number of samples in the batch; C represents the number of categories of IoT devices to be identified; y i,c Indicates the true value judgment of sample i and category C. If sample i is category C, the value is 1, otherwise it is 0; p i,c It represents the probability that the model predicts that the sample i is device category C.
[0064] Among them, multivariate time series features refer to the characteristics of multivariate time series. Multivariate time series refers to time series data composed of multiple variables or dimensions. Different from univariate time series (data containing only one variable changing over time), multivariate time series simultaneously records the changes of multiple related variables over time, and there may be mutual dependence or correlation between these variables.
[0065] In some embodiments of the present invention, Figure 2 As shown, the device recognition model includes a first Inception module, a first residual structure, a second Inception module, a second residual structure, a global average pooling layer, and a fully connected layer.
[0066] The input of the first residual structure is the input of the first Inception module and the output of the first Inception module, and the input of the second residual structure is the input of the second Inception module and the output of the second Inception module.
[0067] Compared with the prior art, the IoT device identification method provided by the embodiment of the present invention extracts abnormal features of IoT device data packets in addition to length features. The abnormal features can characterize the influence of network environment and encrypted traffic on the flow length sequence pattern of IoT device data packets, that is, the abnormal features can be used to characterize the interference of timing noise. In other words, the influence of timing noise can be considered in the process of identifying IoT devices, thereby improving the accuracy of identifying IoT devices. In addition, the adaptability of the IoT device identification method to complex network environments can be greatly improved to meet the identification requirements for encrypted traffic.
[0068] Furthermore, the embodiment of the present invention can further explore the network behavior of IoT devices at different times and scales by setting the device recognition model to include the Inception module, thereby further improving the recognition accuracy. Furthermore, the embodiment of the present invention can solve the gradient vanishing problem between Inception modules by setting the device recognition model to include the residual module, thereby further improving the recognition accuracy of the device recognition model.
[0069] Furthermore, the feature extraction in the embodiment of the present invention takes less time and does not require complicated feature processing and calculation processes, which reduces the time cost of this stage and can better meet the needs of online real-time recognition.
[0070] In summary, the embodiments of the present invention can achieve efficient and accurate identification of IoT devices, lay the foundation for subsequent security management and maintenance, and ensure the safe operation of devices in the IoT environment.
[0071] In order to achieve a comprehensive characterization of IoT device data packets under noise interference, in some embodiments of the present invention, the abnormal features include abnormal level, abnormal grouping and top-level protocol type.
[0072] Among them, the abnormal levels from high to low include Error, Warning, Note and Chat (interactive information requiring user confirmation).
[0073] Among them, abnormal groups include but are not limited to Protocol, Malformed, Sequence, Comment, etc.
[0074] The top-level protocol type refers to the top-level protocol in the network protocol. In other words, the abnormal characteristics of the data packet of the ith IoT device are:
[0075] P i =(s i ,g i ,p i )
[0076] Where Pi is the data packet of the ith IoT device; s i is the abnormal level; g i is the abnormal group; p i The top-level protocol type.
[0077] In order to improve the recognition efficiency of the device recognition model, in some embodiments of the present invention, before constructing the multivariate time series features, it also includes: converting the abnormal features into available space in the numerical space to avoid embedding of high-dimensional features.
[0078] Furthermore, in order to improve the conversion efficiency, in a specific embodiment of the present invention, a database representing the conversion relationship is constructed, and the linear values corresponding to the high-dimensional abnormal features can be obtained by querying the database.
[0079] In some embodiments of the present invention, Figure 3 As shown, the step S101 of extracting the length feature of the IoT device data packet includes:
[0080] S301. Determine the outbound flow and inbound flow in the IoT device data packet.
[0081] The outbound flow and the inbound flow in the length feature are represented by different symbols. Specifically, the inbound flow is represented by -, and the outbound flow is represented by +.
[0082] S302: concatenate the outbound traffic and the inbound traffic in chronological order to obtain the target traffic, and use the length of the target traffic as a length feature.
[0083] For example, if there is outbound traffic O1 at time t1, there is inbound traffic I1 at time t2, there is outbound traffic O2 at time t3, time t1 is earlier than time t2, and time t2 is earlier than time t1, then the target traffic is (O1, I1, O2).
[0084] The length feature of the embodiment of the present invention takes into account both inbound traffic and outbound traffic, thereby improving the accuracy of the length feature, and further improving the recognition accuracy of the Internet of Things devices.
[0085] In some embodiments of the present invention, Figure 4 As shown, step S301 includes:
[0086] S401. Obtain an original data stream; the original data stream includes multiple data packets related to multiple IoT devices.
[0087] The original data stream is obtained by determining and obtaining the original data stream based on a traffic log that records the original data stream.
[0088] S402: Determine, based on the binary information, a first data packet whose source MAC address is a target IoT device and a second data packet whose destination MAC address is a target IoT address among multiple data packets, and use the first data packet as outbound traffic and the second data packet as inbound traffic.
[0089] The two-tuple information refers to the source MAC address and the destination MAC address.
[0090] Since the number of each abnormal feature of the data packet in the specified time interval along the time sequence is not the same, and the number of the abnormal features along the time sequence is also different from the length feature, the different numbers do not conform to the input of the neural network, that is, the device recognition model. Therefore, in some embodiments of the present invention, such as Figure 5 As shown, the step S101 of constructing multivariate time series features based on abnormal features and length features includes:
[0091] S501, dividing the IoT device data packet based on a preset sequence length to obtain multiple time series features;
[0092] S502, filling each segment of time series features with a preset filling rule to obtain a multivariate time series feature;
[0093] Among them, the number of rows of the multivariate time series feature is the same as the number of types of abnormal features and time series features, and the number of columns of the multivariate time series feature is the same as the number of segments of the multi-segment time series feature.
[0094] The embodiment of the present invention truncates and fills the IoT device data packet so that the number of obtained multivariate time series features is consistent and meets the input of the neural network, thereby further improving the recognition accuracy of the IoT device.
[0095] Specifically, the preset sequence length is recorded as L. Within the preset time interval t, the multivariate time series features finally generated by the IoT device D to be identified are for:
[0096]
[0097] The preset filling rule may be: filling the missing data with a preset value.
[0098] It should be understood that the preset value can be set according to experience, which will not be described in detail here.
[0099] In some embodiments of the present invention, Figure 2 As shown, the first Inception module and the second Inception module both include a first Inception structure, a second Inception structure, and a third Inception structure connected in sequence; Figure 6As shown, the first Inception structure, the second Inception structure and the third Inception structure all include a feature compression unit, a multi-scale convolution unit and a feature fusion unit;
[0100] The feature compression unit is used to compress the input features to obtain compressed features;
[0101] The multi-scale convolution unit is used to perform multi-scale convolution on the input features to obtain multiple convolution features of different scales;
[0102] The feature fusion unit is used to fuse the compressed features and multiple convolutional features.
[0103] In the embodiment of the present invention, each Inception structure includes a feature compression unit and a multi-scale convolution unit, which can respectively extract features at different time and space scales through the feature compression unit and the multi-scale convolution unit, thereby improving the characterization capability of the fused features obtained by fusion for the IoT device to be identified, thereby ensuring the recognition accuracy of the IoT device.
[0104] Furthermore, the embodiment of the present invention adds a residual structure between every three Inception structures, so that the input of the first Inception structure is quickly accumulated to the input of the next Inception structure through the residual structure, and the gradient vanishing problem is alleviated by direct gradient flow.
[0105] Specifically, the input of the i-th Inception structure is defined as d in,i , the output of each Inception structure is defined as d out,i , then the residual connection is as follows:
[0106] d i n, i =d out,i-3 +d out,i-1 , i mod 3 = 1
[0107] In the formula, mod is the remainder operator.
[0108] In some embodiments of the present invention, Figure 6 As shown, the feature compression unit includes a first global average pooling layer and a first convolutional layer connected in sequence.
[0109] Among them, the convolution kernel size and step size of the first convolution layer are both 1. The features are compressed through the first global average pooling layer, and then the time series data is expanded to a certain dimension through the first convolution layer.
[0110] In some embodiments of the present invention, Figure 6As shown, the multi-scale convolution unit includes a bottleneck layer and a second convolution layer, a third convolution layer, and a fourth convolution layer connected to the bottleneck layer, and the convolution kernel sizes of the second convolution layer, the third convolution layer, and the fourth convolution layer are all different.
[0111] In some embodiments of the present invention, Figure 6 As shown, the feature fusion unit includes a deep convolution layer, a batch normalization layer, and an activation function layer connected in sequence.
[0112] Among them, the activation function of the activation function layer is ELU.
[0113] In order to verify the effectiveness and superiority of the IoT device identification method proposed in the embodiment of the present invention, the method of the embodiment of the present invention is compared with the performance of KNN (K-Nearest Neighbor), decision tree (Decision Tree, DT), random forest (Random Forest, RF), support vector machine (Support Vector Machines, SVM) and 1D-CNN, ByteIoT, TMC and other algorithms on two public data sets (UNSW data set and CICIoT data set). The evaluation indicators include accuracy (AC), PR (Precision-Recall) value, F1 score, RC (Recall rate). The comparison results are as follows Figure 7 As shown by Figure 7 It can be seen that the embodiment of the present invention is superior to other existing models in various evaluation indicators.
[0114] Furthermore, the embodiment of the present invention also compares the time consumption of the feature engineering stage (feature extraction) and the device identification stage with the ByteIoT and TMC algorithms on two data sets. The comparison results are as follows: Figure 8 As shown by Figure 8 It can be seen that the time consumption in the two stages of the embodiment of the present invention is significantly reduced, that is, the recognition efficiency of the Internet of Things devices is improved.
[0115] The embodiment of the present invention also compares the difference between the IoT device identification method and other algorithm identification effects in different time windows. The comparison results are as follows: Fig. 9 As shown by Fig. 9 It can be seen that the embodiment of the present invention can achieve better recognition accuracy in a smaller time window, that is, it can better meet the timeliness requirements of recognition. In other words, the applicability of the IoT device recognition method is further improved.
[0116] In order to verify the superiority of selecting the inbound and outbound bidirectional lengths as the length features in the embodiments of the present invention, in some embodiments of the present invention, the accuracy indicators of the unidirectional length features and the bidirectional length features are evaluated on the UNSW dataset and the CICIoT dataset respectively. Fig.10 is the accuracy indicator on the UNSW dataset, Fig.11 is the accuracy indicator on the CICIoT dataset, Fig.10 and Fig.11 The horizontal axis is the average length, and the vertical axis is the accuracy. Fig.10 and Fig.11 It can be seen that the accuracy of the bidirectional length feature is higher, that is, the superiority of the length feature proposed in the embodiment of the present invention is verified.
[0117] In order to better implement the IoT device identification method in the embodiment of the present invention, based on the IoT device identification method, the embodiment of the present invention also provides an IoT device identification device, such as Fig.12 As shown, the IoT device identification device 1200 includes:
[0118] The multi-dimensional feature extraction unit 1201 is used to extract the abnormal features and length features of the IoT device data packet, and construct a multi-variable time series feature based on the abnormal features and length features;
[0119] The device identification unit 1202 is used to input the multivariate time series features into the fully trained device identification model to obtain the identification result of the IoT device;
[0120] Among them, the device recognition model includes a first Inception module, a first residual structure, a second Inception module, a second residual structure, a global average pooling layer and a fully connected layer.
[0121] The Internet of Things device identification apparatus 1200 provided in the above embodiment can implement the technical solution described in the above Internet of Things device identification method embodiment. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above Internet of Things device identification method embodiment, which will not be repeated here.
[0122] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.
[0123] The above is a detailed introduction to an IoT device identification method and apparatus provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.
Claims
1. A method for identifying an Internet of Things device, characterized in that: include: Extracting abnormal features and length features of IoT device data packets, and constructing multivariate time series features based on the abnormal features and the length features; Inputting the multivariate time series features into a fully trained device recognition model to obtain recognition results of IoT devices; The device recognition model includes a first Inception module, a first residual structure, a second Inception module, a second residual structure, a global average pooling layer and a fully connected layer.
2. The method for identifying an IoT device according to claim 1, characterized in that: The abnormal features include abnormal level, abnormal grouping and top-level protocol type.
3. The method for identifying an IoT device according to claim 1, characterized in that: Extract the length characteristics of IoT device data packets, including: Determine outbound traffic and inbound traffic in the IoT device data packet; The outbound flow and the inbound flow are concatenated in chronological order to obtain a target flow, and the length of the target flow is used as the length feature.
4. The method for identifying an IoT device according to claim 3, characterized in that: Determining the outbound flow and the inbound flow in the IoT device data packet includes: Obtaining an original data stream; the original data stream includes multiple data packets related to multiple IoT devices; Based on the binary information, determine a first data packet whose source MAC address is a target IoT device and a second data packet whose destination MAC address is a target IoT address among the multiple data packets, use the first data packet as the outbound traffic, and use the second data packet as the inbound traffic.
5. The method for identifying an IoT device according to claim 1, characterized in that: Constructing a multivariate time series feature based on the abnormal feature and the length feature, including: Dividing the IoT device data packet based on a preset sequence length to obtain multiple time series features; Filling the time series features of each segment according to a preset filling rule to obtain the multivariate time series features; Among them, the number of rows of the multivariate time series feature is the same as the number of types of the abnormal features and the time series features, and the number of columns of the multivariate time series feature is the same as the number of segments of the multi-segment time series feature.
6. The method for identifying an Internet of Things device according to claim 1, characterized in that: The first Inception module and the second Inception module each include a first Inception structure, a second Inception structure, and a third Inception structure connected in sequence; the first Inception structure, the second Inception structure, and the third Inception structure each include a feature compression unit, a multi-scale convolution unit, and a feature fusion unit; The feature compression unit is used to compress the input feature to obtain a compressed feature; The multi-scale convolution unit is used to perform multi-scale convolution on the input features to obtain multiple convolution features of different scales; The feature fusion unit is used to fuse the compression feature and the multiple convolution features.
7. The method for identifying an IoT device according to claim 6, characterized in that: The feature compression unit includes a first global average pooling layer and a first convolutional layer connected in sequence.
8. The method for identifying an IoT device according to claim 6, characterized in that: The multi-scale convolution unit includes a bottleneck layer and a second convolution layer, a third convolution layer and a fourth convolution layer connected to the bottleneck layer, and the convolution kernel sizes of the second convolution layer, the third convolution layer and the fourth convolution layer are all different.
9. The method for identifying an Internet of Things device according to claim 6, characterized in that: The feature fusion unit includes a deep convolution layer, a batch normalization layer and an activation function layer connected in sequence.
10. An IoT device identification device, characterized in that: include: A multi-dimensional feature extraction unit, used to extract abnormal features and length features of IoT device data packets, and construct multi-variable time series features based on the abnormal features and the length features; A device identification unit, used to input the multivariate time series features into a fully trained device identification model to obtain an identification result of the IoT device; The device recognition model includes a first Inception module, a first residual structure, a second Inception module, a second residual structure, a global average pooling layer and a fully connected layer.
Citation Information
Patent Citations
IoT device identification method and identification system under encrypted wireless network
CN111711946A
An Internet of Things equipment traffic identification method and system based on time sequence load traffic fingerprints
CN113313156A
Deep learning-based radio frequency fingerprint identification method for frequency equipment
CN114896887A
Internet of Things equipment identification model based on session spatio-temporal characteristic phase residual error
CN115331068A
Internet of Things equipment identification method and system based on fingerprint deep extraction technology
CN116886314A
Cited By
Internet of Things equipment behavior analysis method and system based on adaptive feature extraction
CN120825422A
