Estimation system, estimation program, and estimation method

Abstract

To make it possible to estimate the internal state of a terminal.SOLUTION: An estimation system includes at least one processor. The at least one processor acquires target CSI data indicating time variations in a channel state of wireless communication transmitted and received between a target terminal, and a wireless communication device connected to the target terminal via a wireless communication network, and inputs the target CSI data to a trained model that accepts the CSI data and outputs the internal state of the terminal that transmits and receives the CSI data, thereby estimating the internal state of the target terminal.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] One aspect of the present disclosure relates to an estimation system, an estimation program, and an estimation method. [Background technology]

[0002] There are known methods for estimating the physical motion of an object using CSI data. Non-Patent Document 1 describes a method for recognizing human behavior using CSI data that indicates time-varying wireless communication channel conditions in a commercial Wi-Fi system. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] S. Yousefi, H. Narui, S. Dayal, S. Ermon, and S. Valaee, “A survey on behavior recognition using Wi-Fi channel state information,” IEEE Communications Magazine, vol.55, no.10, pp.98-104, Oct.2017. Summary of the Invention [Problem to be solved by the invention]

[0004] A method for estimating the internal state of a device is desired. [Means for solving the problem]

[0005] An estimation system according to an aspect of the present disclosure includes at least one processor that acquires target CSI data indicating time variations in channel conditions of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network, and inputs the target CSI data to a trained model that receives the CSI data and outputs an internal state of the terminal that transmits and receives the CSI data, thereby estimating the internal state of the target terminal.

[0006] An estimation program according to one aspect of the present disclosure causes a computer to execute the following steps: acquiring target CSI data indicating changes over time in channel conditions of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network; and inputting the target CSI data into a trained model that receives the CSI data and outputs the internal state of a terminal that transmits and receives the CSI data, thereby estimating the internal state of the target terminal.

[0007] An estimation method according to an aspect of the present disclosure is executed by an estimation system including at least one processor. The estimation method includes the steps of acquiring target CSI data indicating time variations in channel conditions of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network, and inputting the target CSI data into a trained model that receives the CSI data and outputs an internal state of the terminal that transmits and receives the CSI data, thereby estimating the internal state of the target terminal.

[0008] In this aspect, CSI data is used to estimate the internal state of the target terminal. As in the conventional method described above, CSI data has been used to estimate the physical motion of an object located between the target terminal and a wireless communication device. However, the present inventors have newly discovered that CSI data correlates with the internal state of the terminal. By processing the CSI data using a trained model that infers this correlation, it becomes possible to estimate the internal state of the target terminal. [Effects of the Invention]

[0009] According to one aspect of the present disclosure, the internal state of a terminal can be estimated. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 2 is a diagram illustrating an example of a functional configuration of the estimation system. [Figure 2] FIG. 10 is a diagram illustrating an example of CSI data. [Figure 3] FIG. 10 is a diagram illustrating another example of CSI data. [Figure 4] FIG. 2 is a diagram illustrating an example of a hardware configuration of a computer that constitutes the estimation system. [Figure 5] 1 is a flowchart illustrating an example of a process for generating a trained model. [Figure 6] 10 is a flowchart illustrating an example of a process for estimating an internal state of a target terminal. DETAILED DESCRIPTION OF THE INVENTION

[0011] Various examples of the present disclosure will be described in detail below with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are designated by the same reference numerals, and redundant description will be omitted.

[0012] [System Overview] The estimation system according to the present disclosure is a system for estimating an internal state of a terminal. In one example, the estimation system estimates the internal state of a target terminal using CSI (Channel State Information) data transmitted and received between the target terminal and a wireless communication device.

[0013] In this disclosure, a terminal refers to a computer equipped with a wireless communication function. The terminal may be, for example, a smartphone, a tablet terminal, a wearable terminal, or an IoT (Internet of Things) device. In this disclosure, a target terminal refers to a terminal whose internal state is to be estimated by an estimation system, and a wireless communication device refers to a communication device connected to the target terminal via a wireless communication network. The wireless communication device may be, for example, a wireless router.

[0014] The CSI data is information indicating a time change in the channel state of the wireless communication transmitted and received between the target terminal and the wireless communication device. In one example, the time change in the channel state of the wireless communication transmitted and received between the target terminal and the wireless communication device is indicated by a time change in the amplitude and phase information of the wireless communication. That is, in one example, the CSI data indicates a displacement of the amplitude and phase information of the wireless communication. The amplitude and phase information are represented by a matrix of values expressed by the absolute value and argument of a complex number. The wireless communication is, for example, Wi-Fi (registered trademark), a wireless network protocol based on IEEE802.11.

[0015] In the present disclosure, the internal state of a terminal is a concept that indicates how the terminal is operating. In one example, the internal state of a terminal includes at least one of which applications are running on the terminal, whether an abnormality has occurred in the operation of the terminal, and which operating system the terminal is running on. That is, the estimation system may estimate, as the internal state of the target terminal, the applications running on the target terminal, whether an abnormality has occurred on the target terminal, or the operating system that is running the target terminal.

[0016] When an application running on a target terminal is estimated, the estimation result may be the name of the application or the type of the application. That is, the estimation system may estimate the name of the application running on the target terminal or the type of the application.

[0017] An application name refers to an identifier that uniquely identifies an individual application, and an application type refers to a category classified according to the genre of the application. Examples of application names include "TikTok," "LINE (registered trademark)," "Puzzle and Dragons (registered trademark)," "Mahjong Soul (registered trademark)," "Youtube (registered trademark)," and "Comic C'moa (registered trademark)." Examples of application types include "SNS," "games," "video viewing," and "manga."

[0018] An abnormality in the target terminal indicates, for example, a state in which unauthorized wireless communication is occurring between the target terminal and a wireless communication device. Examples of unauthorized wireless communication include communications related to cyber attacks from external terminals, such as DoS attacks. Examples of operating systems that run the terminal include "Android (registered trademark)," "iOS," "Windows (registered trademark)," and "MacOS (registered trademark)."

[0019] Conventionally, CSI data has been used to estimate the physical motion of an object. One example is a method for estimating the physical motion of an object located between a target terminal and a wireless communication device. In contrast, the present inventors have newly discovered that CSI data correlates with the internal state of a terminal, and have found a new way to utilize CSI data by using it to estimate the internal state of the terminal. Using such CSI data, an estimation system can estimate the internal state of the target terminal. In one example, the estimation system estimates the internal state of the target terminal using a trained model generated by performing machine learning. Machine learning is a method for autonomously finding laws or rules by iteratively learning based on given information.

[0020] [System Configuration] An example of application of an estimation system 10 according to an embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of the functional configuration of the estimation system 10. In this example, the estimation system 10 estimates the internal state of the target terminal 20 using CSI data transmitted and received between the target terminal 20 and a wireless communication device 30 connected via a wireless communication network N. The estimation system 10 accesses a sample database 40 to perform its processing.

[0021] In one example, the estimation system 10 includes functional modules including a learning unit 11, an acquisition unit 12, an estimation unit 13, and an output unit 14. The learning unit 11 is a functional module that generates a trained model M. The acquisition unit 12 is a functional module that acquires target CSI data. In the present disclosure, the target CSI data refers to CSI data used to estimate the internal state of the target terminal 20. The estimation unit 13 is a functional module that inputs the target CSI data into the trained model and estimates the internal state of the target terminal 20. The output unit 14 is a functional module that outputs the processing result.

[0022] The sample database 40 stores sample data used in machine learning to generate a trained model. In one example, the sample database 40 is provided in a computer system separate from the estimation system 10. Alternatively, the sample database 40 may be a component of the estimation system 10.

[0023] The sample data includes multiple data records indicating combinations of CSI data and the internal states of the target terminal 20 corresponding to the CSI data. In other words, each of the multiple data records can be considered as teacher data in machine learning, and the internal states of the target terminal 20 can be considered as labels, which are information treated as ground truth in the machine learning.

[0024] In one example, the CSI data in the sample data indicates changes over time in amplitude and phase information of wireless communication between the target terminal 20 and the wireless communication device 30 for each packet. Alternatively, the CSI data in the sample database may indicate changes over time in amplitude and phase information of wireless communication between the wireless communication device 30 and another terminal of the same model as the target terminal 20 for each packet. In these examples, in each data record, the internal state of the target terminal 20 is set as a label for CSI data having a reference packet width. The reference packet width is set to correspond to changes in the internal state of the target terminal 20, for example.

[0025] The packet width is represented by the number of consecutive packets. The packet width may be set taking into consideration the internal state to be estimated. For example, one packet width may correspond to 100 packets or 200 packets. In one example, pre-acquired CSI data is divided into pieces of a reference packet width, and then labels are assigned to each piece of divided CSI data by a user operation, thereby generating individual data records.

[0026] An example of the data structure of the CSI data stored in the sample database 40 is described in more detail below. FIG. 2 is a diagram showing an example of CSI data, specifically raw CSI data transmitted and received between the target terminal 20 and the wireless communication device 30. In the graph shown in FIG. 2, the vertical axis represents the displacement of amplitude and phase information of wireless communication between the target terminal 20 and the wireless communication device 30, and the horizontal axis represents packets. In this disclosure, "raw CSI data" refers to unprocessed CSI data. Such raw CSI data is obtained, for example, by decoding a pcap file in which network traffic transmitted and received between the target terminal 20 and the wireless communication device 30 is saved on a packet-by-packet basis.

[0027] As shown in Fig. 2, raw CSI data may contain noise in the displacement of amplitude and phase information. Therefore, in each data record of sample data, the internal state of the corresponding target terminal 20 is set as a label for CSI data obtained by removing the noise from the raw CSI data. In the following description, "CSI data obtained by removing noise from raw CSI data" may be referred to as post-noise-removal CSI data.

[0028] In one example, noise contained in raw CSI data is removed by calculating a moving average of the displacement for each of multiple intervals set along the horizontal axis (packets) of the graph in Figure 2. The interval is set to, for example, 10 packets, and the average value of the displacement at 10 points within the interval is calculated for each packet.

[0029] FIG. 3 shows the results of applying such noise removal processing to the raw CSI data shown in FIG. 2. FIG. 3 is a diagram illustrating another example of CSI data, specifically, CSI data after noise removal. The vertical and horizontal axes of the graph shown in FIG. 3 are the same as those of the graph shown in FIG. 2. The CSI data after noise removal shown in FIG. 3 is obtained by calculating a moving average of the raw CSI data shown in FIG. 2, with the above-mentioned interval set to 10 packets. Comparing the raw CSI data shown in FIG. 2 and the CSI data after noise removal shown in FIG. 3, fine fluctuations in displacement can be confirmed in the CSI data after noise removal. In this way, by performing noise removal processing on the raw CSI data, fine fluctuations in displacement can be captured, and by using such CSI data after noise removal for estimation, it becomes possible to accurately estimate the internal state of the target terminal 20.

[0030] FIG. 4 is a diagram showing an example of the hardware configuration of a computer 100 constituting the estimation system 10. For example, the computer 100 includes a processor 101, a main memory 102, an auxiliary memory 103, a communication control unit 104, an input device 105, and an output device 106. The processor 101 executes an operating system and application programs. The main memory 102 is composed of, for example, ROM and RAM. The auxiliary memory 103 is composed of, for example, a hard disk or flash memory, and generally stores larger amounts of data than the main memory 102. The communication control unit 104 is composed of, for example, a network card or a wireless communication module. The input device 105 is composed of, for example, a keyboard, a mouse, a touch panel, etc. The output device 106 is composed of, for example, a monitor and speakers.

[0031] Each functional module of the estimation system 10 is realized by an estimation program 110 pre-stored in the auxiliary storage unit 103. Each functional module is realized by loading the estimation program 110 onto the processor 101 or the main storage unit 102 and having the processor 101 execute the estimation program 110. The processor 101 operates the communication control unit 104, the input device 105, or the output device 106 in accordance with the estimation program 110, and reads and writes data from and to the main storage unit 102 or the auxiliary storage unit 103.

[0032] The estimation program 110 may be provided in a state of being recorded on a non-transitory recording medium such as a CD-ROM, a DVD-ROM, a semiconductor memory, etc. Alternatively, the estimation program 110 may be provided via a communication network as a data signal superimposed on a carrier wave.

[0033] The estimation system 10 may be configured with one computer 100 or may be configured with multiple computers 100. When multiple computers 100 are used, these computers 100 are connected via a communication network such as the Internet or an intranet to logically construct one estimation system 10. The estimation system 10 may also be constructed by combining multiple types of computers.

[0034] [System Operation] The operation of the estimation system 10 will now be described, along with an estimation method according to the present disclosure.

[0035] (Generating a trained model) First, a process for generating a trained model M will be described with reference to Fig. 5. Fig. 5 is a flowchart showing an example of a process for generating a trained model M as a processing flow S1.

[0036] In step S11, the learning unit 11 acquires one data record from the sample database 40. In process flow S1, the learning unit 11 acquires a data record from the sample database 40, in which the internal state of the device corresponding to the CSI data is set as a label. In one example, the label included in each data record for generating a certain trained model M is unified to one of an application running on the device, information indicating whether an abnormality has occurred in the device, and the operating system that is running the device.

[0037] In step S12, the learning unit 11 performs machine learning based on the data records. In process flow S1, the learning unit 11 performs machine learning using one-dimensional convolution. The learning unit 11 may perform machine learning using LSTM (Long Short-Term Memory), TCN (Temporal Convolutional Network), LSTM-FCN (Long Short Term Memory-Fully Convolutional Network), or the like.

[0038] In one example, the learning unit 11 inputs the acquired data records into a machine learning model and obtains an inference result output from the machine learning model. The learning unit 11 updates parameters in the machine learning model using a technique such as backpropagation based on the error between the inference result and the label. The learning unit 11 updates, for example, the weights of a neural network as parameters in the machine learning model.

[0039] In step S13, the learning unit 11 determines whether to terminate the machine learning based on a predetermined termination condition. The termination condition may be set based on an error or the number of data records to be processed, i.e., the number of times machine learning has been performed. Alternatively, the learning unit 11 may evaluate the performance of the machine learning model using given validation data, and terminate the machine learning if the evaluation satisfies a given criterion.

[0040] If the learning unit 11 determines that the predetermined termination condition is not satisfied (NO in step S13), the process returns to step S11. In the repeated process, the learning unit 11 acquires the next data record in step S11, and performs machine learning based on that data record in step S12.

[0041] If the learning unit 11 determines that the predetermined termination condition is satisfied (YES in step S13), the process proceeds to step S14. In step S14, the learning unit 11 outputs the trained model M. The learning unit 11 outputs the machine learning model for which machine learning through the series of processes in steps S11 to S13 has been completed as the trained model M. In one example, the learning unit 11 stores the trained model M in the auxiliary storage unit 103.

[0042] As described above, in the process flow S1, the learning unit 11 performs machine learning using one-dimensional convolution in step S12. Therefore, the trained model M can be said to be a trained model generated by performing machine learning using one-dimensional convolution.

[0043] (Estimation of the internal state of the target device) Next, an example of processing for estimating the internal state of the target terminal 20 will be described with reference to Fig. 6. Fig. 6 is a flowchart showing an example of processing for estimating the internal state of the target terminal 20 as processing flow S2.

[0044] In step S21, the acquisition unit 12 acquires target CSI data. The target CSI data has the same data structure as the CSI data in the sample data. That is, in one example, the target CSI data indicates time changes in amplitude and phase information of wireless communication between the target terminal 20 and the wireless communication device 30 per packet. The target CSI data may have a packet width larger than the reference packet width or the same as the reference packet width.

[0045] In one example, the acquisition unit 12 acquires, as the target CSI data, raw CSI data transmitted and received between the target terminal 20 and the wireless communication device 30. In this example, the acquisition unit 12 acquires a pcap file in which network traffic exchanged between the target terminal 20 and the wireless communication device 30 is saved on a packet-by-packet basis, and decodes the pcap file to acquire the target CSI data.

[0046] In step S22, the estimation unit 13 removes noise from the target CSI data. In one example, the estimation unit 13 removes noise from the target CSI data by calculating a moving average of the displacement of amplitude and phase information for each of a plurality of sections set along the packet axis of the target CSI data. The noise removal process for the target CSI data is performed in the same manner as the process for the CSI data in the sample data, for example. In step S22, the estimation unit 13 can also be said to generate noise-removed CSI data for the target CSI data.

[0047] In step S23, the estimation unit 13 inputs the target CSI data into the trained model M to estimate the internal state of the target terminal 20. In one example, the estimation unit 13 inputs the target CSI data from which noise has been removed into the trained model M to estimate the internal state of the target terminal 20. The output from the trained model M depends on the labels of the sample data used in the machine learning. The estimation unit 13 may estimate, as the internal state, an application running on the target terminal 20, whether or not an abnormality has occurred in the target terminal 20, or the operating system that operates the target terminal 20.

[0048] If the target CSI data has a packet width greater than the reference packet width, the estimation unit 13 may estimate the internal state of the target terminal 20, for example, as follows. First, the estimation unit 13 divides the target CSI data, from which noise has been removed, into units of the reference packet width. Then, the estimation unit 13 inputs each divided CSI data into the trained model M to estimate the internal state of the target terminal 20 for each reference packet width. If the target CSI data has the same packet width as the reference packet width, the estimation unit 13 inputs the target CSI data into the trained model M without performing the above-described division process, and estimates the internal state of the target terminal 20.

[0049] In step S24, the output unit 14 outputs the processing result. In one example, the output unit 14 outputs a graph of the target CSI data together with the estimated internal state of the target terminal 20 as the processing result. In this example, the output unit 14 may clearly show the relationship between the target CSI data and the internal state of the target terminal 20 by, for example, changing the line color of the graph for each reference packet width in accordance with the internal state of the target terminal 20. This allows the user to quickly grasp the correspondence between the two. Alternatively, the output unit 14 may output only the estimated internal state of the target terminal 20 as the processing result.

[0050] If it is estimated that an abnormality has occurred in the target terminal 20, the output unit 14 may output an abnormality notification as at least a part of the processing result. Alternatively, in this case, the output unit 14 may output a command to disconnect the connection with the wireless communication network N to the target terminal 20 together with the abnormality notification.

[0051] The output unit 14 may display the processing results on a display device, may store the processing results in a given storage device such as a memory, or may transmit the processing results to another computer system.

[0052] [Variations] Various examples of the present disclosure have been described above in detail. However, the present disclosure is not limited to the above examples. Various modifications can be made to the present disclosure without departing from the spirit and scope of the present disclosure.

[0053] 1 shows one target terminal 20, the number of target terminals 20 may be two or more. In this example, the estimation system 10 may estimate the internal state of each of the multiple target terminals 20.

[0054] The CSI data and target CSI data in the sample data may indicate time changes in amplitude and phase information of wireless communication between the target terminal 20 and the wireless communication device 30 per unit time. In this example, in each data record of the sample data, the internal state of the target terminal 20 may be set as a label for the CSI data of a reference time interval. The reference time interval may be set to correspond to changes in the internal state of the target terminal 20, for example.

[0055] In step S21, the acquisition unit 12 may acquire, as the target CSI data, CSI data from which noise has already been removed. That is, the acquisition unit 12 may acquire CSI data after noise removal from the target CSI data. In this example, the process flow S2 may not include step S22. In this example, a computer system different from the estimation system 10 may perform the noise removal process on the raw CSI data corresponding to step S22.

[0056] In step S23, the estimation unit 13 may input the raw CSI data as target CSI data to another trained model different from the trained model M to estimate the internal state of the target terminal 20. The other trained model may be generated by performing machine learning using a plurality of other data records that indicate combinations of the raw CSI data and the internal state of the target terminal 20. In this example, the process flow S2 may not include step S22.

[0057] In this disclosure, the expression "at least one processor executes a first process, executes a second process, ... executes an nth process" or a corresponding expression is a concept that includes cases where the entity executing the n processes from the first process to the nth process (i.e., the processor) changes midway through. In other words, this expression is a concept that includes both cases where all n processes are executed by the same processor and cases where the processor changes among the n processes according to an arbitrary policy.

[0058] The information processing method executed by at least one processor is not limited to the above examples. For example, some of the steps or processes described above may be omitted, or the steps may be executed in a different order. Furthermore, any two or more of the steps described above may be combined, or some of the steps may be modified or deleted. Alternatively, other steps may be executed in addition to the steps described above.

[0059] [Note] As can be seen from the various examples above, the present disclosure includes the following aspects. <Item 1> at least one processor; the at least one processor: acquire target CSI data indicating a time change in a channel state of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network; inputting the target CSI data into a trained model that receives CSI data and outputs an internal state of a terminal that transmits and receives the CSI data, and estimating the internal state of the target terminal; Estimation system. <Item 2> the at least one processor estimates, as the internal state, an application running on the target terminal; Item 1. The estimation system according to item 1. <Item 3> the at least one processor: As the internal state, it is estimated whether or not an abnormality has occurred in the target terminal; If it is estimated that an abnormality has occurred in the target terminal, an abnormality notification is output. Item 3. The estimation system according to item 1 or 2. <Item 4> the at least one processor acquires the target CSI data, which indicates a time change in amplitude and phase information of the wireless communication between the target terminal and the wireless communication device per packet as the time change in the channel state of the wireless communication; Item 3. The estimation system according to any one of items 1 to 3. <Item 5> The at least one processor performs machine learning based on the CSI data and the internal state using one-dimensional convolution to generate the trained model. 5. The estimation system according to any one of items 1 to 4. <Item 6> acquiring target CSI data indicating a time change in a channel state of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network; inputting the target CSI data into a trained model that receives CSI data and outputs an internal state of a terminal that transmits and receives the CSI data, and estimating the internal state of the target terminal; An estimation program that causes a computer to execute the above. <Item 7> 1. An estimation method performed by an estimation system comprising at least one processor, comprising: acquiring target CSI data indicating a time change in a channel state of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network; inputting the target CSI data into a trained model that receives CSI data and outputs an internal state of a terminal that transmits and receives the CSI data, and estimating the internal state of the target terminal; Estimation methods including:

[0060] In items 1, 6, and 7, CSI data is used to estimate the internal state of the target terminal 20. Conventionally, CSI data has been used to estimate the physical motion of an object between the target terminal 20 and the wireless communication device 30. However, the present inventors have newly discovered that CSI data correlates with the internal state of the terminal. By processing the CSI data using a trained model that infers this correlation, it becomes possible to estimate the internal state of the target terminal 20.

[0061] A conventional method for estimating active applications is, for example, a method based on information about the target terminal 20 itself. However, there is a risk that the information about the target terminal 20 itself may be tampered with by the user of the target terminal 20, and in this case, it may be impossible to properly estimate the applications active on the target terminal 20. In contrast, the CSI data indicates time variations in amplitude and phase information transmitted and received between the target terminal 20 and the wireless communication device 30, and therefore is less likely to be tampered with by the user of the target terminal 20. Therefore, according to Item 2, it is possible to properly estimate the applications active on the target terminal 20.

[0062] According to item 3, it is possible to notify, for example, a user of the target terminal 20 that an abnormality has occurred in the target terminal 20.

[0063] In wireless communication between the target terminal 20 and the wireless communication device 30, the time interval between transmitted packets may vary depending on the internal state of the target terminal 20 or the communication quality of the wireless communication network N. By acquiring the time variation of the amplitude and phase information transmitted and received between the target terminal 20 and the wireless communication device 30 per packet as the target CSI data, the influence of the variation in the time interval between packets can be avoided. According to Item 4, by inputting such target CSI data into the trained model, the internal state of the target terminal 20 can be estimated more accurately.

[0064] According to item 5, when performing machine learning based on CSI data and internal states, one-dimensional convolution, which is good at handling time changes, can be used to generate a trained model M with high estimation accuracy. [Explanation of symbols]

[0065] 10...estimation system, 11...learning unit, 12...acquisition unit, 13...estimation unit, 14...output unit, 20...target terminal, 30...wireless communication device, M...trained model, N...wireless communication network.

Claims

1. at least one processor; the at least one processor: Acquire target CSI data indicating a time change in a channel state of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network; inputting the target CSI data into a trained model that receives CSI data and outputs an internal state of a terminal that transmits and receives the CSI data, and estimating the internal state of the target terminal; Estimation system.

2. the at least one processor estimates, as the internal state, an application running on the target terminal; The estimation system of claim 1 .

3. the at least one processor: As the internal state, it is estimated whether or not an abnormality has occurred in the target terminal; If it is estimated that an abnormality has occurred in the target terminal, an abnormality notification is output. The estimation system according to claim 1 or 2.

4. The at least one processor acquires the target CSI data indicating a time change in amplitude and phase information of the wireless communication between the target terminal and the wireless communication device per packet as the time change in the channel state of the wireless communication. The estimation system according to claim 1 or 2.

5. The at least one processor performs machine learning based on the CSI data and the internal state using one-dimensional convolution to generate the trained model. The estimation system according to claim 1 or 2.

6. acquiring target CSI data indicating a time change in a channel state of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network; inputting the target CSI data into a trained model that receives CSI data and outputs an internal state of a terminal that transmits and receives the CSI data, and estimating the internal state of the target terminal; An estimation program that causes a computer to execute the above.

7. 1. An estimation method performed by an estimation system comprising at least one processor, comprising: acquiring target CSI data indicating a time change in a channel state of wireless communication transmitted and received between a target terminal and a wireless communication device connected to the target terminal via a wireless communication network; inputting the target CSI data into a trained model that receives CSI data and outputs an internal state of a terminal that transmits and receives the CSI data, and estimating the internal state of the target terminal; Estimation methods including:

Images