Estimation device, estimation method, and estimation program

The estimation device uses dynamic programming to compare abnormal and normal data, effectively identifying anomalies in variable-length communication data by extracting non-matching portions, improving detection accuracy and explainability.

JP7586306B2Active Publication Date: 2024-11-19NIPPON TELEGRAPH & TELEPHONE CORP
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2023522024
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-05-17
Publication Date
2024-11-19
Estimated Expiration
2041-05-17

AI Technical Summary

Technical Problem

Conventional anomaly detection technologies struggle to identify the cause of abnormalities in variable-length communication data using deep learning, as the decision-making process is a black box and lacks explainability.

Method used

An estimation device that calculates similarity between abnormal and normal data using dynamic programming to identify the cause of anomalies by extracting portions of abnormal data that do not correspond to normal data.

Benefits of technology

Enables accurate identification of the cause of anomalies in variable-length communication data by comparing abnormal data with the most similar normal data, thereby enhancing explainability and detection accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007586306000003
    Figure 0007586306000003
  • Figure 0007586306000004
    Figure 0007586306000004
  • Figure 0007586306000005
    Figure 0007586306000005
Patent Text Reader

Abstract

An estimation unit (15b) estimates a set of normal data among a plurality of sets of normal data determined to be normal that has the greatest similarity to abnormal data determined to be abnormal. An extraction unit (15c) compares the abnormal data with the set of normal data estimated to have the greatest similarity, and extracts, as a portion representing the cause of an abnormality, a portion of the abnormal data that does not include a portion corresponding to the set of normal data.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an estimation device, an estimation method, and an estimation program. [Background technology]

[0002] In recent years, efforts have been made to optimize factories by utilizing communication data from control devices in the factory, known as smart factories. However, there are concerns about the emergence of new types of cyber attacks targeting smart factories, and security measures are urgently needed. Defending against new types of cyber attacks is difficult, so detection technology is in demand. In particular, anomaly-based anomaly detection using deep learning, which automatically learns normal characteristics from input data, is expected to be able to detect new types of cyber attacks with a high degree of accuracy.

[0003] However, the decision-making process of deep learning is a black box, and in practical use, it is impossible for security personnel to identify the cause of anomalies because the cause of the detected anomaly cannot be explained.In recent years, a technology called explainable AI that explains the cause of anomalies from deep learning has been disclosed (see Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] William Briguglio and Sherif Saad, “Interpreting Machine Learning Malware Detectors Which Leverage N-gram Analysis”, [online], [Retrieved April 14, 2021], Internet<URL: https: / / arxiv.org / abs / 2001.10916.pdf> [Non-Patent Document 2] Xiao Zhang, Manish Marwah, I-ta Lee, Martin Arlitt, Dan Goldwasser, “ACE-An Anomaly Contribution Explainer for Cyber-Security Applications”, [online], [Retrieved April 14, 2021], Internet<URL: https: / / arxiv.org / pdf / 1912.00314.pdf> Summary of the Invention [Problem to be solved by the invention]

[0005] However, conventional techniques only detect anomalies in fixed-length communication data and identify the cause of the anomaly, and cannot detect anomalies in variable-length communication data and identify the cause of the anomaly.

[0006] The present invention has been made in view of the above, and has an object to make it possible to identify the cause of an anomaly by detecting an anomaly in variable-length communication data. [Means for solving the problem]

[0007] In order to solve the above-mentioned problems and achieve the object, the estimation device according to the present invention is characterized by having an estimation unit that uses abnormal data determined to be abnormal and a plurality of normal data determined to be normal to estimate normal data that has the greatest similarity to the abnormal data, and an extraction unit that compares the abnormal data with the estimated normal data and extracts a portion of the abnormal data that has no corresponding portion in the normal data as a cause of the abnormality. Effect of the Invention

[0008] According to the present invention, it is possible to identify the cause of an anomaly by detecting an anomaly in variable-length communication data. [Brief description of the drawings]

[0009] [Figure 1]FIG. 1 is a diagram for explaining an outline of the estimation device. [Diagram 2] FIG. 2 is a schematic diagram illustrating a schematic configuration of the estimation device. [Diagram 3] FIG. 3 is a diagram illustrating the extraction processing result. [Figure 4] FIG. 4 is a flowchart showing the procedure of the estimation process. [Diagram 5] FIG. 5 is a diagram for explaining an embodiment. [Figure 6] FIG. 6 is a diagram illustrating a computer that executes the estimation program. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to this embodiment. In addition, in the description of the drawings, the same parts are indicated by the same reference numerals.

[0011] [Outline of the estimation device] FIG. 1 is a diagram for explaining an outline of the estimation device. As shown in FIG. 1, the estimation device of this embodiment detects anomalies in variable-length communication data and identifies byte locations that are estimated to be the cause of the anomaly in the communication data. Specifically, as shown in FIG. 1(a), for a pair of communication data determined to be anomalous (abnormal judgment data) and a plurality of communication data determined to be normal (normal judgment data), the estimation device identifies normal judgment data that is most similar to the abnormal judgment data, as shown in FIG. 1(b). At this time, the estimation device applies dynamic programming, which obtains the overall calculation result while recording the calculation results of a plurality of divided parts, to calculate the similarity between the abnormal judgment data and each normal judgment data. Among them, as shown in FIG. 1(c), a location with a gap is identified, and this is estimated to be the abnormal byte that caused the abnormal judgment.

[0012] [Configuration of the estimation device] 2 is a schematic diagram illustrating a schematic configuration of an estimation device 10. As illustrated in FIG. 2, the estimation device 10 is realized by a general-purpose computer such as a personal computer, and includes an input unit 11, an output unit 12, a communication control unit 13, and a control unit 15.

[0013] The input unit 11 is realized by input devices such as a keyboard and a mouse, and inputs various instruction information such as a command to start processing to the control unit 15 in response to an input operation by an operator. The output unit 12 is realized by a display device such as a liquid crystal display, a printing device such as a printer, or the like.

[0014] The communication control unit 13 is realized by a NIC (Network Interface Card) or the like, and controls communication between an external device such as a server via a network and the control unit 15. For example, the communication control unit 13 controls communication between the control unit 15 and a management device or the like that manages data sets and parameters that are targets of the estimation process described below.

[0015] The control unit 15 is realized using a CPU (Central Processing Unit) or the like, and executes a processing program stored in a memory. As a result, the control unit 15 functions as an acquisition unit 15a, an estimation unit 15b, and an extraction unit 15c, as exemplified in FIG. 2. Note that each or a part of these functional units may be implemented in different hardware. For example, the estimation unit 15b and the extraction unit 15c may be implemented in different hardware. The control unit 15 may also include other functional units.

[0016] The acquisition unit 15a acquires data determined to be abnormal and a plurality of data determined to be normal. For example, the acquisition unit 15a acquires abnormal data to be subjected to estimation processing described later, a data set of normal data to be used in the estimation processing, parameters to be used in the estimation processing, and the like, via the input unit 11 or from a management device that manages the results of deep learning, or the like, via the communication control unit 13.

[0017] The acquiring unit 15a may store the acquired data in a storage unit (not shown) realized by a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk. Alternatively, the acquiring unit 15a may transfer the information to the estimating unit 15b described below without storing the information in the storage unit.

[0018] The estimation unit 15b estimates normal data that has the maximum similarity between the abnormal data determined to be abnormal and each of the multiple normal data determined to be normal. Specifically, the estimation unit 15b uses the acquired data to evaluate the similarity between the payload of a packet determined to be abnormal and a set of payloads of packets determined to be normal, and estimates the payload that is most similar to the payload of the packet determined to be abnormal.

[0019] For example, the estimation unit 15b calculates the similarity between the abnormal data and the normal data by applying dynamic programming. Here, the estimation device 10 processes the payload of a packet of one abnormal data and, for example, 100 or more normal data among the data determined to be abnormal / normal by anomaly detection using deep learning. The payload is, for example, a hexadecimal character string represented as [00 00 10 FF 10], and is a variable-length data string of length 1 or more.

[0020] The estimation unit 15b calculates the similarity between the abnormal data and the normal data by assuming that characters appearing in the abnormal data and the normal data are looped continuously. That is, the estimation unit 15b assumes that byte characters (00 to FF) appearing in the data are looped, and assumes that the similarity between the byte characters "00" and "FF" and the similarity between "00" and "01" are the same value, for example.

[0021] In addition, the estimation unit 15b estimates one abnormal data string X=[x1, x2, ..., x n ] and multiple normal data columns Y 1 ,Y 2 ,…,Y k ,…,Y lOne normal data sequence Y=[y1,y2,…,y m In this case, the estimation unit 15b uses dynamic programming to calculate the similarity S between X and Y, target Calculate.

[0022] Dynamic programming can be expressed as follows:

[0023]

number

number

[0024] The above formula (1) is a single character x in the abnormal data string X. i and one character y in the normal data string Y j Similarity s(x i ,y j ) is used to quantify the similarity S(i,j) between X and Y.

[0025] In addition, in the first line of the above formula (1), as shown in the above formula (2), x i andy j If the two match, s(x i ,y j ) as +10, if no match, x i andy j The distance is subtracted.

[0026] Alternatively, the estimation unit 15b may preset, for example, when comparing with the byte character “00” of abnormal data, the character similarity s with the byte character “00” of the normal data to be compared to is +10, the character similarity s with “01”, “02”, “03”, “FF”, “FE”, and “FD” is +5, and the similarity s with other byte characters is −5.

[0027] In this manner, the estimation unit 15b calculates the similarity between the abnormal data and the normal data by setting the similarity s between the characters of the abnormal data and the normal data to a value within a predetermined range.

[0028] The estimation unit 15b also estimates the abnormal data sequence X and the k-th normal data sequence Y k After evaluating the similarity between the abnormal data sequence X and the k+1th normal data sequence Y k+1 The process of evaluating the similarity between the l-th normal data sequence Y l Repeat until.

[0029] In this way, the estimation unit 15b can compare variable-length data using dynamic programming.

[0030] Thereafter, the estimation unit 15b calculates the similarity between the abnormal data sequence X and each normal data sequence Y by performing a correction using the anomaly score. That is, the estimation unit 15b calculates the similarity between the abnormal data sequence X and each normal data sequence Y by multiplying the above S(i,j) by the AI ​​anomaly score (anomaly score) resulting from anomaly detection by deep learning for each normal data sequence Y as a bias value. For example, the estimation unit 15b applies a bias so that the similarity score between the abnormal data and the normal data that has a lower anomaly score among the normal data is higher. Then, the estimation unit 15b determines the normal data sequence Y with the highest calculated similarity as Y target Identify as:

[0031] The extraction unit 15c compares the abnormal data with the normal data estimated to have the highest similarity, and extracts a portion of the abnormal data that does not correspond to the normal data as a cause of the abnormality. Specifically, the extraction unit 15c extracts a portion of the payload of the packet determined to be abnormal that has a gap with the estimated payload as an abnormal portion.

[0032] For example, if the payload of the abnormal data of the comparison source is "00 11 FF FD" and the payload of the normal data of the comparison destination is "00 FF FD", the extraction unit 15c determines that the part corresponding to the second byte "11" of the comparison source is an abnormal part and extracts it because it does not exist in the normal data. In this case, the extraction unit 15c replaces the second byte of the normal data with a blank character.

[0033] Incidentally, even if the second byte of the normal data is not "11", the second byte of the abnormal data and the normal data may be extracted as the location of the gap.

[0034] Furthermore, the extraction unit 15c outputs the extracted abnormal portion via the output unit 12. For example, the extraction unit 15c outputs a payload of a pair of abnormal data and normal data to the output unit 12 such as a display, and displays a portion having a gap determined to be an abnormal portion in a highlighted manner. The extraction unit 15c may output information on the extracted abnormal portion to another information processing device via the communication control unit 13.

[0035] Here, Fig. 3 is a diagram illustrating an example of the extraction process result. Fig. 3 illustrates normal data with the highest similarity to the abnormal data, and a portion of the abnormal data where there is a gap with the normal data. As shown in bold in Fig. 3, "0X86 0X8C 0X5F" is displayed as an abnormal portion with a gap.

[0036] [Estimation process] Next, the estimation process by the estimation device 10 according to the present embodiment will be described with reference to Fig. 4. Fig. 4 is a flowchart showing the procedure of the estimation process. The flowchart in Fig. 4 is started, for example, when an operation input is made to instruct the start of the estimation process.

[0037] First, the acquiring unit 15a acquires abnormal data determined to be abnormal and a plurality of normal data determined to be normal (step S1).

[0038] Next, the estimation unit 15b creates pairs of abnormal data and normal data (step S2), and estimates a pair having the highest similarity between the abnormal data and the normal data from among the created pairs (step S3).

[0039] For example, the estimation unit 15b calculates the similarity of each pair using dynamic programming. At that time, the estimation unit 15b calculates the similarity of each pair assuming that characters appearing in the abnormal data and normal data are continuously looped. In addition, the estimation unit 15b calculates the similarity of each pair by setting the similarity of characters in the abnormal data and normal data to a value within a predetermined range. In addition, the estimation unit 15b calculates the similarity of each pair by performing correction using the anomaly score.

[0040] The extraction unit 15c also compares the abnormal data with the normal data estimated to have the highest similarity, and extracts a portion of the abnormal data that has no corresponding portion in the normal data and has a gap as the cause of the abnormality (step S4).

[0041] Then, the extraction unit 15c outputs the portion having the gap determined to be the abnormal portion to the output unit 12 by highlighting and displaying it (step S5). This ends a series of estimation processes.

[0042] As described above, in the estimation device 10, the estimation unit 15b estimates normal data that has the maximum similarity between abnormal data determined to be abnormal and each of a plurality of normal data determined to be normal. The extraction unit 15c compares the abnormal data with the normal data estimated to have the maximum similarity, and extracts a portion of the abnormal data that does not correspond to the normal data as the cause of the abnormality.

[0043] In this way, the estimation device 10 can identify the part of the abnormal data that is most similar to the normal data and differs from the normal data as the cause of the abnormality, regardless of the length of the normal data and the abnormal data. This enables the estimation device 10 to identify the cause of the abnormality by detecting anomalies in communication data of variable length.

[0044] In addition, the estimation unit 15b calculates the similarity between the abnormal data and the normal data by applying dynamic programming. In this way, the estimation device 10 can accurately identify the cause of the abnormality by calculating the similarity between the abnormal data and the normal data, specifically targeting the variable-length communication data as the target of the abnormality detection.

[0045] In addition, the estimation unit 15b calculates the similarity between the abnormal data and the normal data by assuming that characters appearing in the abnormal data and the normal data are looped continuously. In this way, the estimation device 10 can specifically and efficiently calculate the similarity, thereby identifying the cause of the abnormality.

[0046] In addition, the estimation unit 15b calculates the similarity between the abnormal data and the normal data by setting the similarity between the characters of the abnormal data and the normal data to a value within a predetermined range. In this way, the estimation device 10 can specifically and efficiently calculate the similarity, thereby enabling the cause of the abnormality to be identified with high accuracy.

[0047] In addition, the estimation unit 15b calculates the similarity between the abnormal data and the normal data by performing correction using the anomaly score. In this way, the estimation device 10 can specifically calculate the similarity with high accuracy, thereby identifying the cause of the abnormality.

[0048] [Example] 5 is a diagram for explaining an embodiment. In this embodiment, the estimation accuracy of the byte location of abnormal communication data occurring in a cyber attack scenario assumed as follows was measured.

[0049] As a cyber attack scenario, we assumed a malicious code insertion attack in which a malicious person inserts code to execute a malicious program into a part of a normally operating communication, enabling unauthorized operation.The byte location where the malicious code is inserted is estimated as the abnormal location.

[0050] For example, in the case of normal communication "00 01 00 00 00 41 6b" and abnormal communication "00 01 00 00 54 1b FF FF 31 00 41 6b," the inserted malicious code "54 1b FF FF 31" is identified as the abnormal part.

[0051] Among the packets that had been assigned normal or abnormal labels through deep learning, 2,133 pieces of normal data and 100 pieces of abnormal data for each length of malicious code to be inserted (three types: 1 to 3) were used to insert a byte sequence of malicious code into one random location in the normal data.

[0052] Anomalies were identified for each packet. At that time, each byte was converted from hexadecimal (0x00 to 0xff) to a numeric value (0 to 255) to create a data format that can be used with dynamic programming.

[0053] The evaluation results of the embodiment are shown in Fig. 5. In the evaluation, for each packet, it was confirmed whether the byte sequence of the anomaly part was perfectly matched, and if it was perfectly matched, it was judged to be correct, and if it was not, it was judged to be incorrect.

[0054] As shown in Figure 5, when there was one anomalous byte sequence inserted, the anomaly could be inferred in all packets. When there were two anomalous byte sequences inserted, the anomaly could not be correctly inferred in two packets, and when there were three anomalous byte sequences inserted, the anomaly could not be correctly inferred in six packets. However, overall, we confirmed that the estimation could be performed with an accuracy of over 90%.

[0055] [program] A program in which the process executed by the estimation device 10 according to the above embodiment is written in a language executable by a computer can also be created. As an embodiment, the estimation device 10 can be implemented by installing an estimation program that executes the above estimation process as package software or online software on a desired computer. For example, the information processing device can function as the estimation device 10 by executing the above estimation program on the information processing device. In addition, the information processing device also includes mobile communication terminals such as smartphones, mobile phones, and PHS (Personal Handyphone System), and further slate terminals such as PDA (Personal Digital Assistant). The function of the estimation device 10 may be implemented on a cloud server.

[0056] 6 is a diagram showing an example of a computer that executes an estimation program. The computer 1000 includes, for example, a memory 1010, a CPU 1020, a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.

[0057] The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012. The ROM 1011 stores a boot program such as a BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to a hard disk drive 1031. The disk drive interface 1040 is connected to a disk drive 1041. A removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1041. The serial port interface 1050 is connected to a mouse 1051 and a keyboard 1052, for example. The video adapter 1060 is connected to a display 1061, for example.

[0058] Here, the hard disk drive 1031 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. Each piece of information described in the above embodiments is stored in the hard disk drive 1031 or memory 1010, for example.

[0059] The estimation program is stored in the hard disk drive 1031 as, for example, a program module 1093 in which instructions to be executed by the computer 1000 are written. Specifically, the hard disk drive 1031 stores the program module 1093 in which each process executed by the estimation device 10 described in the above embodiment is written.

[0060] Data used for information processing by the estimation program is stored as program data 1094, for example, in the hard disk drive 1031. Then, the CPU 1020 reads out the program module 1093 and the program data 1094 stored in the hard disk drive 1031 into the RAM 1012 as necessary, and executes each of the above-mentioned procedures.

[0061] The program module 1093 and program data 1094 related to the estimation program are not limited to being stored in the hard disk drive 1031, and may be stored in, for example, a removable storage medium and read by the CPU 1020 via the disk drive 1041 or the like. Alternatively, the program module 1093 and program data 1094 related to the estimation program may be stored in another computer connected via a network such as a LAN (Local Area Network) or a WAN (Wide Area Network), and may be read by the CPU 1020 via the network interface 1070.

[0062] Although the embodiment to which the invention made by the present inventor is applied has been described above, the present invention is not limited by the description and drawings which form a part of the disclosure of the present invention according to the present embodiment. In other words, other embodiments, examples, operation techniques, etc. made by those skilled in the art based on the present embodiment are all included in the scope of the present invention. [Explanation of symbols]

[0063] 10 Estimation device 11 Input section 12 Output section 13 Communication control section 15 Control section 15a Acquisition part 15b Estimation part 15c Extraction part

Claims

1. an estimation unit that estimates normal data having a maximum similarity between the abnormal data determined to be abnormal and each of a plurality of normal data determined to be normal; an extraction unit that compares the abnormal data with normal data estimated to have the highest similarity, and extracts a portion of the abnormal data that does not correspond to the normal data as a cause of the abnormality; having The estimation device, wherein the estimation unit calculates the similarity by applying dynamic programming, assuming that characters appearing in the abnormal data and the normal data form a continuous loop.

2. An estimation unit that estimates normal data that has a maximum similarity between abnormal data determined to be abnormal and each of a plurality of normal data determined to be normal; an extraction unit that compares the abnormal data with normal data estimated to have the highest similarity, and extracts a portion of the abnormal data that does not correspond to the normal data as a cause of the abnormality; having The estimation device, wherein the estimation unit calculates the similarity by applying dynamic programming through correction using an anomaly score.

3. 3. The estimation device according to claim 1, wherein the estimation unit calculates the similarity between characters in the abnormal data and the normal data by setting the similarity to a value within a predetermined range.

4. An estimation method executed by an estimation device, an estimation step of estimating normal data having a maximum similarity between the abnormal data determined to be abnormal and each of a plurality of normal data determined to be normal; an extraction step of comparing the abnormal data with normal data estimated to have the highest similarity, and extracting a portion of the abnormal data that does not correspond to the normal data as a cause of the abnormality; Including, The estimation method is characterized in that the estimation step calculates the similarity by applying dynamic programming, assuming that characters appearing in the abnormal data and the normal data form a continuous loop.

5. An estimation method executed by an estimation device, comprising: an estimation step of estimating normal data having a maximum similarity between the abnormal data determined to be abnormal and each of a plurality of normal data determined to be normal; an extraction step of comparing the abnormal data with normal data estimated to have the highest similarity, and extracting a portion of the abnormal data that does not correspond to the normal data as a cause of the abnormality; Including, The estimation method, wherein the estimation step calculates the similarity by applying dynamic programming through correction using an anomaly score.

6. On the computer, an estimation step of estimating normal data having a maximum similarity between the abnormal data determined to be abnormal and each of a plurality of normal data determined to be normal; an extraction step of comparing the abnormal data with normal data estimated to have the highest similarity, and extracting a portion of the abnormal data that does not correspond to the normal data as a cause of the abnormality; Run the command, The estimation program is characterized in that the estimation step calculates the similarity by applying dynamic programming, assuming that characters appearing in the abnormal data and the normal data are in a continuous loop.

7. A computer comprising: an estimation step of estimating normal data having a maximum similarity between the abnormal data determined to be abnormal and each of a plurality of normal data determined to be normal; an extraction step of comparing the abnormal data with normal data estimated to have the highest similarity, and extracting a portion of the abnormal data that does not correspond to the normal data as a cause of the abnormality; Run the command, The estimation program, wherein the estimation step calculates the similarity by applying dynamic programming through correction using an anomaly score.

Citation Information

Patent Citations

  • Apparatus, method, and medium for detecting payload anomalies using n-gram distributions of normal data

    JP2007515867A