KR20230060418A - Method of detecting software malfunction using Binary Image - Google Patents

Abstract

According to an embodiment of the present invention, a method for detecting a software malfunction comprises: (A) a step of starting a software malfunction detection system, and starting the software malfunction detection process; (B) a step in which a control unit of the malfunction detection system receives a large number of pieces of sensing information from a sensing unit; (C) a step in which the control unit binarizes the large number of pieces of sensing information; (D) a step in which the control unit performs comparison and learning by using a machine learning neural network; (E) a step in which the control unit determines whether there has occurred a software malfunction; and (F) a step in which the control unit outputs whether a software malfunction has occurred or not by text or sound through an output unit. Therefore, a software malfunction can be easily detected.

Description

Method of detecting software malfunction using binary image {Method of detecting software malfunction using Binary Image}

The present invention relates to a method for detecting a malfunction in software using a binary image, and more particularly, to a method for detecting a malfunction in software in which sensing information according to a run-time execution flow of software is shaped into a binary image and the malfunction is detected using a machine learning neural network. .

Conventionally, as smart electronic systems are applied to autonomous vehicles, artificial intelligence robots, and unmanned devices, etc., they are being developed to operate by sequential instrumental interlocking between detailed blocks of complex software.

In such a smart electronic system, it is very difficult to analyze the cause when a specific part of the software malfunctions. The cause may also be a code error in the block, but when an error occurs in the interaction between other connected modules, it is very difficult to analyze the cause. Patch work takes a considerable amount of time, and there is a limit to manual work.

In particular, while conventional smart electronic systems are controlled by complex software, there are various types of code error patterns caused by dynamic code execution rather than static codes. At this time, there is a problem in that defining the detection code as the code itself for each of these cases increases development complexity and increases the capacity of the code memory when inserting the check code.

Patent Document: Registered Patent Publication No. 10-2088164

The present invention has been made to solve the above problems, and an object of the present invention is to shape sensing information according to the run-time execution flow of software into a binary image and to use a machine learning neural network to detect malfunctions. Provide a method for detecting malfunctions in software is to do

A software malfunction detection method according to an embodiment of the present invention includes the steps of (A) starting a software malfunction detection system to detect a software malfunction; (B) receiving a plurality of sensing information from a sensing unit by a control unit of the malfunction detection system; (C) processing, by the controller, binarization of the plurality of sensing information; (D) comparative learning by the control unit using a machine learning neural network; (E) determining, by the control unit, whether a software malfunction has occurred; and (F) outputting, by the control unit, whether a malfunction of the software has occurred in text or sound through an output unit.

In the software malfunction detection method according to an embodiment of the present invention, the plurality of sensing information includes any one of program counters, code addresses, memory access locations, CPU registers, memory usage and variable calls. characterized by

In the software malfunction detection method according to an embodiment of the present invention, the step (C) displays a high portion of each of the plurality of sensing information as 1 based on the critical plane for the (C-1) threshold value (TH), Marking the low part as 0; (C-2) forming a binarized image matrix in which the portion indicated by 1 is represented by black pixels and the portion indicated by 0 is represented by white pixels; and (C-3) converting the binarized image matrix into a one-dimensional array image.

In the software malfunction detection method according to an embodiment of the present invention, in the step (D), the 1D array image and the corresponding 1D array reference image representing a normal operation are input together to the machine learning neural network to perform comparative learning. to be

In the software malfunction detection method according to an embodiment of the present invention, the step (E) derives an error value, determines whether the error value is higher than a preset malfunction reference value, and determines whether the malfunction occurs. characterized by

Features and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings.

Prior to this, the terms or words used in this specification and claims should not be interpreted in a conventional, dictionary sense, and the inventor properly defines the concept of the term in order to explain his/her invention in the best way. Based on the principle that it can be done, it should be interpreted as meaning and concept consistent with the technical spirit of the present invention.

A software malfunction detection method according to an embodiment of the present invention converts sensing information into a binarized image using a critical plane, thereby minimizing data loss, reducing data size, and easily detecting software malfunction. It works.

1 is a configuration diagram of a software malfunction detection system to which a software malfunction detection method according to an embodiment of the present invention is applied.
2 is a flow chart of a method for detecting an erroneous operation of software according to an embodiment of the present invention.
3 is an exemplary view of sensing information input according to a software malfunction detection method according to an embodiment of the present invention;
4 is an exemplary diagram illustrating a process of binarizing input sensing information according to a software malfunction detection method according to an embodiment of the present invention;
5 is an exemplary view showing a result of binarizing input sensing information according to a software malfunction detection method according to an embodiment of the present invention.
6 is an exemplary diagram illustrating a software malfunction detection process according to a software malfunction detection method according to an embodiment of the present invention;

Objects, specific advantages and novel features of the present invention will become more apparent from the following detailed description and preferred embodiments taken in conjunction with the accompanying drawings. In adding reference numerals to components of each drawing in this specification, it should be noted that the same components have the same numbers as much as possible, even if they are displayed on different drawings. Also, terms such as first and second may be used to describe various components, but the components should not be limited by the terms. These terms are only used for the purpose of distinguishing one component from another. In addition, in describing the present invention, if it is determined that a detailed description of related known technologies may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. 1 is a configuration diagram of a software malfunction detection system to which a software malfunction detection method according to an embodiment of the present invention is applied.

A software malfunction detection system 10 according to an embodiment of the present invention includes a control unit 11, a memory unit 12, a sensing unit 13 and an output unit 14 as shown in FIG.

The control unit 11 has at least one machine learning neural network therein, controls the overall operation of the software malfunction detection system, and stores the sensing information received through the sensing unit 13 in memory when the software is executed. A process of detecting an erroneous operation of the software is controlled by comparing with reference information stored in the unit 12.

The memory unit 12 is connected to the control unit 11 and stores a plurality of software related to the operation of the smart electronic system, standard information for detecting malfunctions, and sensing information received through the sensing unit 13.

The sensing unit 13 includes a plurality of sensors connected to the control unit 11, the memory unit 12, the output unit 14, and other components (not shown) constituting the system, respectively, and software through the plurality of sensors. Sensing information at the time of execution is received and transmitted to the control unit 11 . Here, the sensing information is an operation signal detected in the system configuration during software execution, for example, stack point change, code address change, memory usage change, variable call related to the memory access location of the memory unit 12 Include fluctuations, etc.

The output unit 14 is connected to the control unit 11, and according to the control of the control unit 11, whether or not the software malfunctions can be displayed in text or notified by sound.

In the software malfunction detection system 10 to which the software malfunction detection method according to the embodiment of the present invention configured as described above is applied, the control unit 11 binarizes the sensing information received through the sensing unit 13 when the software is executed, and , The binarized sensing image is input to one machine learning neural network, for example, LSVM (Linear Support Vector Machine) provided therein together with the reference binarized sensing image stored in the memory unit 12, and comparative learning is performed to detect malfunction of the software. there is.

Hereinafter, a software malfunction detection method according to an embodiment of the present invention will be described with reference to FIGS. 2 to 7 . 2 is a flowchart of a method for detecting a malfunction of software according to an embodiment of the present invention, FIG. 3 is an exemplary view of sensing information input according to a method for detecting a malfunction of software according to an embodiment of the present invention, and FIG. It is an exemplary view showing a process of binarizing input sensing information according to a method for detecting a malfunction of software according to an embodiment of the present invention, and FIG. 6 is an exemplary diagram showing a software malfunction detection process according to a software malfunction detection method according to an embodiment of the present invention.

In the software malfunction detection method according to an embodiment of the present invention, first, the software malfunction detection system 10 shown in FIG. 1 is started to detect software malfunction (S210).

Accordingly, the control unit 11 receives a plurality of sensing information from the sensing unit 13 (S220).

Specifically, as the smart electronic system runs and executes software, the control unit 11 starts the malfunction detection system 10 of the software shown in FIG. 1 and detects data through a plurality of sensors included in the sensing unit 13. As shown in FIG. 3, it is possible to receive sensing information represented by a change graph over time by detecting sensing information, that is, various operation signals appearing during software execution.

At this time, the operation signal includes signals of various variables such as, for example, a program counter, code address, memory access location, CPU registers, memory usage, and variable call.

After receiving a plurality of sensing information, the control unit 11 performs binarization on the received sensing information (S230).

Specifically, the controller 11 receives the sensing information shown in FIG. 4 and converts the received sensing information based on the critical plane 101 for the threshold value TH shown in FIG. 4 . That is, the control unit 11 uses the following [Equation 1] to obtain portions S11 and S12 higher than the critical plane 101 for the threshold value TH shown in FIG. 4 in the first sensing information S1. The third sensing information ( In S3), portions S31 and S32 higher than the critical plane 101 for the threshold value TH shown in FIG. 4 have a value of 1, and portions lower than the other critical plane 101 have a value of 0. let it have

Here, R(x,y) denotes a binarized image matrix, r(x,y) denotes a coordinate of sensing information, and TH denotes a threshold value.

In this case, the threshold TH may be set in advance according to the sensitivity of the smart electronic system, the number of sensed information, and the stability of the sensed information.

As shown in FIG. 5, the binarized sensing information displays sensing information s1, s2, and s3 binarized in the column direction for a row at time t, and a value of 1 as in “I” It is possible to form a binarized image matrix 200 in which pixels are displayed in black and a value of 0, such as “II”, is displayed in white.

After obtaining the binarized image matrix 200, the controller 11 converts the binarized image matrix 200 into a 1D array image 200' using [Equation 2] below.

Here, RP represents a one-dimensional array image, and R(x,y) represents a binarized image matrix.

After converting the 1D array image 200', the controller 11 inputs the 1D array image 200' to the machine learning neural network and performs comparative learning (S240).

Specifically, the controller 11 acquires the 1D array image 200' shown in FIG. 6, and transmits the acquired 1D array image 200' to a machine learning neural network, for example, a LSVM (Linear Support Vector Machine) neural network. can be entered. Here, the LSVM neural network is a simple neural network with relatively small capacity compared to other neural networks, and can improve the performance efficiency of malfunction determination. Of course, other machine learning neural networks other than LSVM neural networks may also be used.

Accordingly, as shown in FIG. 6 , the controller 11 may compare and learn the 1D array image 200 ′ with the 1D array reference image 300 representing a corresponding normal operation. Here, the 1D array reference image 300 may be selected by the control unit 11 and input to the LSVM neural network in a state in which it is included in the reference information for detecting a malfunction in the memory unit 12 and stored in advance.

In this way, by comparing and learning the 1D array image 200' and the 1D array reference image 300, the control unit 11 determines whether a software malfunction has occurred (S250).

At this time, determination of whether a malfunction has occurred may be performed using Equation 3 below in the LSVM neural network.

Here, RP represents the 1D array image 200', and td represents the 1D array reference image 300 representing a normal operation.

By determining whether a malfunction occurs using [Equation 3], an error value of 3 is derived, and the control unit 11 determines whether or not a malfunction has occurred by determining whether the derived error value is higher than a preset malfunction reference value. can do.

For example, the controller 11 may determine that a software malfunction has occurred because the error value of 3 derived by determining whether or not a malfunction has occurred is higher than a preset malfunction standard value of 1.

In accordance with determining that the software malfunction occurs, the control unit 11 may display the occurrence of the software malfunction through the output unit 14 in text or sound (S260).

On the other hand, if the error value derived from the determination of whether a malfunction has occurred is lower than the preset malfunction reference value, the control unit 11 repeatedly performs the receiving of a plurality of sensing information from the sensing unit 13 (S220) again. .

The software malfunction detection method according to an embodiment of the present invention in this process converts the sensing information into a binarized image using the critical plane 101, thereby minimizing data loss and reducing the data size, and software malfunction. can be easily detected.

Although the technical idea of the present invention has been specifically described according to the above preferred embodiments, it should be noted that the above-described embodiments are for explanation and not for limitation.

In addition, those skilled in the art will understand that various implementations are possible within the scope of the technical spirit of the present invention.

10: malfunction detection system 11: control unit
12: memory unit 13: sensing unit
14: output unit 101: critical plane
200: Binarized image matrix 200': One-dimensional array image
300: 1-dimensional array reference image

Claims (5)

(A) starting a software malfunction detection system to start software malfunction detection;
(B) receiving a plurality of sensing information from a sensing unit by a control unit of the malfunction detection system;
(C) processing, by the controller, binarization of the plurality of sensing information;
(D) comparative learning by the control unit using a machine learning neural network;
(E) determining, by the control unit, whether a software malfunction has occurred; and
(F) outputting, by the control unit, whether a malfunction of the software has occurred in text or sound through an output unit;
A method for detecting a malfunction of a software comprising a.
According to claim 1,
The plurality of sensing information includes any one of program counters, code addresses, memory access locations, CPU registers, memory usage and variable calls.
According to claim 1,
The step (C) is
(C-1) displaying a high part of each of the plurality of sensing information as 1 and a low part as 0 based on the critical plane with respect to the threshold value (TH);
(C-2) forming a binarized image matrix in which the portion indicated by 1 is represented by black pixels and the portion indicated by 0 is represented by white pixels; and
(C-3) converting the binarized image matrix into a 1D array image;
A method for detecting malfunction of software, characterized in that it further comprises.
According to claim 3,
The step (D) includes inputting the 1D array image and the 1D array reference image representing a corresponding normal operation together into the machine learning neural network for comparative learning.
According to claim 1,
The step (E) derives an error value and determines whether the error value is higher than a preset malfunction reference value to determine whether the malfunction has occurred.

Images