KR102618066B1 - Method, device and system for strengthening military security based on natural language process and image compare in soldier based community application - Google Patents

Abstract

According to one embodiment, in a method of strengthening military security based on natural language processing and image comparison in a soldier-based community application performed by a device, the method includes receiving an upload request for a first post from a first user terminal. step; When it is confirmed that the first posting contains text, extracting the text included in the first posting as the first text; Inputting the first text into a pre-trained first artificial intelligence model, and setting the probability of a military security violation of the first text as a first probability based on the output of the first artificial intelligence model; If the first probability is confirmed to be lower than a preset first reference rate, determining that there is no security problem in the first text; When it is confirmed that the first post contains an image, extracting the image included in the first post as a first image; Inputting the first image into a pre-trained second artificial intelligence model, and setting the military security violation probability of the first image to a second probability based on the output of the second artificial intelligence model; If the second probability is confirmed to be lower than a preset second reference rate, determining that there is no security problem in the first image; And if it is determined that there is no security problem with the text included in the first posting and that there is no security problem with the image included in the first posting, the status of the first posting is set to allow public. A method for enhancing military security based on natural language processing and image matching in a soldier-based community application is provided, including the steps of:

Description

METHOD, DEVICE AND SYSTEM FOR STRENGTHENING MILITARY SECURITY BASED ON NATURAL LANGUAGE PROCESS AND IMAGE COMPARE IN SOLDIER BASED COMMUNITY APPLICATION}

The embodiments below relate to technologies for enhancing military security based on natural language processing and image matching in soldier-based community applications.

As the Internet develops and the use of mobile devices increases, online communities in various fields are created, and communication between users takes place in online communities.

In particular, communities related to the military, such as soldiers, those preparing to enlist, military families, and reservists, are also increasing.

In the military-related community, stories, issues, knowledge, and information related to soldiers, the military, and military life can be uploaded and shared. However, since it is military-related, there is a problem that it may contain content related to military security.

Accordingly, conventionally, the administrator of the military-related community checks the posts and deletes the posts if it is determined that there is a military security problem. This is because a person has to check each post individually and manage the posts. , is having difficulty operating the community.

Accordingly, there is an increasing demand to determine whether a post has a military security problem and to automatically upload only the post without a security problem to the community, and the implementation of related technology is required.

Korean Patent Publication No. 10-2023-0083130 Korean Patent Publication No. 10-2022-0153753 Korean Patent No. 10-2337367 Korean Patent No. 10-1618547

According to one embodiment, the purpose is to provide a method, device, and system for strengthening military security based on natural language processing and image comparison in a soldier-based community application.

The object of the present invention is not limited to the object mentioned above, and other objects not mentioned can be clearly understood from the description below.

According to one embodiment, in a method of strengthening military security based on natural language processing and image comparison in a soldier-based community application performed by a device, the method includes receiving an upload request for a first post from a first user terminal. step; When it is confirmed that the first posting contains text, extracting the text included in the first posting as the first text; Inputting the first text into a pre-trained first artificial intelligence model, and setting the probability of a military security violation of the first text as a first probability based on the output of the first artificial intelligence model; If the first probability is confirmed to be lower than a preset first reference rate, determining that there is no security problem in the first text; When it is confirmed that the first post contains an image, extracting the image included in the first post as a first image; Inputting the first image into a pre-trained second artificial intelligence model, and setting the military security violation probability of the first image to a second probability based on the output of the second artificial intelligence model; If the second probability is confirmed to be lower than a preset second reference rate, determining that there is no security problem in the first image; And if it is determined that there is no security problem with the text included in the first posting and that there is no security problem with the image included in the first posting, the status of the first posting is set to allow public. A method for enhancing military security based on natural language processing and image matching in a soldier-based community application is provided, including the steps of:

When the first text is input, the first artificial intelligence model recognizes key keywords through natural language processing for the first text, sets the key keyword recognized in the first text as the first keyword, and Through keyword comparison between the first keyword and pre-registered military security violation keywords, the probability that there is a military security problem for the first keyword is analyzed, and the probability that there is a military security problem for each key keyword recognized in the first text is analyzed. When the probability is analyzed, the probability analyzed for each key keyword is classified into a first probability group, the highest probability among the probabilities classified into the first probability group is set as the first probability, and an output value indicating the first probability Outputs, and when the first image is input, the second artificial intelligence model recognizes an object through image analysis of the first image, sets the object recognized in the first image as the first object, and , dividing the area occupied by the first object in the first image into a first area, dividing the part where the first area is in the first image to extract a second image, and dividing the first image into a first area. The probability that there is a military security problem for the first object is analyzed through image comparison between registered military security violation images, and when the probability that there is a military security problem for each object recognized in the first image is analyzed, the object The analyzed probabilities may be classified into a second probability group, the highest probability among the probabilities classified into the second probability group may be set as the second probability, and an output value indicating the second probability may be output.

The method of strengthening military security based on natural language processing and image comparison in the soldier-based community application includes, when the first probability is confirmed to be higher than the first reference rate, the first probability from the first artificial intelligence model Obtaining information about the probability of being classified into groups; classifying a probability higher than the first reference rate among the probabilities classified into the first probability group into a third probability group; setting the number of probabilities classified into the third probability group as a first numerical value; If the first value is confirmed to be lower than a preset first standard value, determining that uploading of the first post is possible through partial modification of the first text; If the first value is confirmed to be higher than the first standard value, setting the status of the first post as not public; If the second probability is confirmed to be higher than the second reference rate, obtaining information about the probability classified into the second probability group from the second artificial intelligence model; classifying a probability higher than the second reference rate among the probabilities classified into the second probability group into a fourth probability group; setting the number of probabilities classified into the fourth probability group as a second numerical value; If the second value is confirmed to be lower than a preset second standard value, determining that uploading to the first post is possible through partial modification of the first image; And if it is confirmed that the second value is higher than the second standard value, the step of setting the status of the first post as not public may be further included.

According to one embodiment, the probability of a military security violation of a post is analyzed, and only posts with no security problems are classified through the analyzed probability and uploaded to a military-based community, thereby increasing user convenience for community operation and management, and military It has the effect of preventing posts that violate security from being uploaded.

Meanwhile, the effects according to the embodiments are not limited to those mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below.

1 is a diagram schematically showing the configuration of a system according to an embodiment.
Figure 2 is a flowchart illustrating a process for strengthening military security based on natural language processing and image comparison in a soldier-based community application according to an embodiment.
Figure 3 is a flow chart to explain the process of analyzing and outputting the probability of a military security violation of text according to an embodiment.
Figure 4 is a flow chart to explain the process of analyzing and outputting the probability of a military security violation of an image according to an embodiment.
Figure 5 is a flow chart to explain the process of determining whether to make a post public according to the number of keywords with security problems according to an embodiment.
Figure 6 is a flowchart illustrating a process for determining whether to make a post public according to the number of objects with security problems according to an embodiment.
FIG. 7 is a flowchart illustrating a process of setting the status of the first post to allow public by performing alternative keyword change or mosaic processing according to an embodiment.
FIG. 8 is a flowchart illustrating a process for setting the status of the first post to be made public by performing replacement object image change or mosaic processing according to an embodiment.
Figure 9 is a flow chart to explain the process of adding military security violation keywords.
Figure 10 is an exemplary diagram of the configuration of a device according to an embodiment.

Hereinafter, embodiments will be described in detail with reference to the attached drawings. However, various changes can be made to the embodiments, so the scope of the patent application is not limited or limited by these embodiments. It should be understood that all changes, equivalents, or substitutes for the embodiments are included in the scope of rights.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed form, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical spirit.

Terms such as first or second may be used to describe various components, but these terms should be interpreted only for the purpose of distinguishing one component from another component. For example, a first component may be named a second component, and similarly, the second component may also be named a first component.

When a component is referred to as being “connected” to another component, it should be understood that it may be directly connected or connected to the other component, but that other components may exist in between.

The terms used in the examples are for descriptive purposes only and should not be construed as limiting. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, terms such as “comprise” or “have” are intended to designate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, but are not intended to indicate the presence of one or more other features. It should be understood that this does not exclude in advance the possibility of the existence or addition of elements, numbers, steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by a person of ordinary skill in the technical field to which the embodiments belong. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and unless explicitly defined in the present application, should not be interpreted in an ideal or excessively formal sense. No.

In addition, when describing with reference to the accompanying drawings, identical components will be assigned the same reference numerals regardless of the reference numerals, and overlapping descriptions thereof will be omitted. In describing the embodiments, if it is determined that detailed descriptions of related known technologies may unnecessarily obscure the gist of the embodiments, the detailed descriptions are omitted.

Embodiments may be implemented in various types of products such as personal computers, laptop computers, tablet computers, smart phones, televisions, smart home appliances, intelligent vehicles, kiosks, and wearable devices.

In an embodiment, an artificial intelligence (AI) system is a computer system that implements human-level intelligence, and unlike existing rule-based smart systems, it is a system in which machines learn and make decisions on their own. As artificial intelligence systems are used, recognition rates improve and sellers' tastes can be more accurately understood, and existing rule-based smart systems are gradually being replaced by deep learning-based artificial intelligence systems.

Artificial intelligence technology consists of machine learning and element technologies using machine learning. Machine learning is an algorithmic technology that classifies/learns the characteristics of input data on its own, and elemental technology is a technology that mimics the functions of the human brain such as cognition and judgment by utilizing machine learning algorithms such as deep learning, including linguistic understanding and visual It consists of technical areas such as understanding, reasoning/prediction, knowledge expression, and motion control.

The various fields where artificial intelligence technology is applied are as follows. Linguistic understanding is a technology that recognizes and applies/processes human language/characters and includes natural language processing, machine translation, conversation systems, question and answer, and voice recognition/synthesis. Visual understanding is a technology that recognizes and processes objects like human vision, and includes object recognition, object tracking, image search, person recognition, scene understanding, spatial understanding, and image improvement. Inferential prediction is a technology that judges information to make logical inferences and predictions, and includes knowledge/probability-based reasoning, optimization prediction, preference-based planning, and recommendations. Knowledge expression is a technology that automatically processes human experience information into knowledge data, and includes knowledge construction (data creation/classification) and knowledge management (data utilization). Motion control is a technology that controls the autonomous driving of vehicles and the movement of robots, and includes motion control (navigation, collision, driving), operation control (behavior control), etc.

Generally, in order to apply machine learning algorithms to real life, learning is performed using a trial and error method due to the nature of the basic methodology of machine learning. In particular, deep learning requires hundreds of thousands of iterations. It is impossible to execute this in an actual physical external environment, so instead, the actual physical external environment is virtually implemented on a computer and learning is performed through simulation.

1 is a diagram schematically showing the configuration of a system according to an embodiment.

Referring to FIG. 1, a system according to an embodiment may include a plurality of user terminals 100 and devices 200 that can communicate with each other through a communication network.

First, a communication network can be configured regardless of the communication mode, such as wired or wireless, and can be implemented in various forms to enable communication between servers and between servers and terminals.

Each of the plurality of user terminals 100 may be implemented as a computing device with a communication function, and may be implemented as, for example, a mobile phone, desktop PC, laptop PC, tablet PC, smartphone, POS terminal, etc. It is not limited and may be implemented as various types of communication devices that can be connected to an external server.

The plurality of user terminals 100 are terminals used by users who wish to communicate through a military-based community application. The first user terminal 110 is used by the first user, and the second user terminal 120 is used by the second user. ), etc. may be included. Here, the soldier-based community application is an application that provides a soldier-based community, and can be provided through the device 200, which will be described later. The soldier-based community includes stories, issues, knowledge, information, etc. related to soldiers, the military, and military life. It is a community where people can freely express themselves and share empathy, and through the community, users can communicate and share information.

Each of the plurality of user terminals 100 can connect to the device 200 through a soldier-based community application provided by the device 200. To this end, a soldier-based community application is installed on each of the plurality of user terminals 100, and the application can operate in conjunction with the device 200.

Hereinafter, for convenience of explanation, the operation of the first user terminal 110 will be mainly described, but of course, other user terminals such as the second user terminal 120 can perform the operation of the first user terminal 110 instead. am.

The device 200 may be its own server owned by a person or organization that provides services using the device 200, a cloud server, or a p2p (peer-to-peer) set of distributed nodes. It may be possible. The device 200 may be configured to perform all or part of the calculation function, storage/reference function, input/output function, and control function of a typical computer. The device 200 may be equipped with at least one artificial intelligence model that performs an inference function.

The device 200 may be configured to communicate wired or wirelessly with a plurality of user terminals 100, control the operation of each of the plurality of user terminals 100, and display certain information on the screen of each of the plurality of user terminals 100. You can control whether to display it or not.

Meanwhile, for convenience of explanation, only the first user terminal 110 and the second user terminal 120 are shown in FIG. 1 among the plurality of user terminals 100, but the number of terminals may vary depending on the embodiment. . As long as the processing capacity of the device 200 allows, the number of terminals is not particularly limited.

According to one embodiment, the device 200 is implemented as a server that provides various services through a soldier-based community application, and can operate and manage a platform related to the soldier-based community.

In operating and managing a platform related to the military-based community, the device 200 checks whether there are security problems in posts requested to be uploaded to the military-based community and ensures that only posts without security problems are disclosed to the military-based community. Posts with security issues can be processed so that they are not made public. At this time, the device 200 can check whether the post has a security problem based on natural language processing and image comparison.

In other words, the device 200 can strengthen military security based on natural language processing and image comparison in a soldier-based community application, and performs natural language processing and image comparison using artificial intelligence to prevent security issues in the soldier-based community. The upload process can be automatically processed by distinguishing only posts without .

To this end, the device 200 can analyze, based on artificial intelligence, how likely it is that the text included in the post violates military security, and based on artificial intelligence, the device 200 can analyze how likely it is that the image included in the post violates military security. You can analyze what the probability is.

In the present invention, artificial intelligence (AI) refers to technology that imitates human learning ability, reasoning ability, perception ability, etc. and implements this with a computer, and includes concepts such as machine learning and symbolic logic. may include. Machine Learning (ML) is an algorithmic technology that classifies or learns the characteristics of input data on its own. Artificial intelligence technology is a machine learning algorithm that analyzes input data, learns the results of the analysis, and makes judgments or predictions based on the results of the learning. Additionally, technologies that mimic the functions of the human brain, such as cognition and judgment, using machine learning algorithms can also be understood as the category of artificial intelligence. For example, technical fields such as verbal understanding, visual understanding, reasoning/prediction, knowledge representation, and motion control may be included.

Machine learning can refer to the process of training a neural network model using experience processing data. Machine learning can mean that computer software improves its own data processing capabilities. A neural network model is built by modeling the correlation between data, and the correlation can be expressed by a plurality of parameters. A neural network model extracts and analyzes features from given data to derive correlations between data. Repeating this process to optimize the parameters of the neural network model can be called machine learning. For example, a neural network model can learn the mapping (correlation) between input and output for data given as input-output pairs. Alternatively, a neural network model may learn the relationships by deriving regularities between given data even when only input data is given.

An artificial intelligence learning model or neural network model may be designed to implement the human brain structure on a computer, and may include a plurality of network nodes with weights that simulate neurons of a human neural network. A plurality of network nodes may have a connection relationship with each other by simulating the synaptic activity of neurons in which neurons exchange signals through synapses. In an artificial intelligence learning model, multiple network nodes are located in layers of different depths and can exchange data according to convolutional connection relationships. The artificial intelligence learning model may be, for example, an artificial neural network model (Artificial Neural Network), a convolution neural network (CNN) model, etc. As an example, an artificial intelligence learning model may be machine-learned according to methods such as supervised learning, unsupervised learning, and reinforcement learning. Machine learning algorithms for performing machine learning include Decision Tree, Bayesian Network, Support Vector Machine, Artificial Neural Network, and Ada-boost. , Perceptron, Genetic Programming, Clustering, etc. can be used.

Among them, CNN is a type of multilayer perceptrons designed to use minimal preprocessing. CNN consists of one or several convolution layers and general artificial neural network layers on top of them, and additionally utilizes weights and pooling layers. Thanks to this structure, CNN can fully utilize input data with a two-dimensional structure. Compared to other deep learning structures, CNN shows good performance in both video and audio fields. CNNs can also be trained via standard back propagation. CNNs have the advantage of being easier to train and using fewer parameters than other feedforward artificial neural network techniques.

Convolutional networks are neural networks that contain sets of nodes with bound parameters. The increasing size of available training data and the availability of computational power, combined with algorithmic advances such as piecewise linear unit and dropout training, have led to significant improvements in many computer vision tasks. For extremely large data sets, such as those available for many tasks today, overfitting is not critical, and increasing the size of the network improves test accuracy. Optimal use of computing resources becomes a limiting factor. For this purpose, distributed, scalable implementations of deep neural networks can be used.

Figure 2 is a flowchart illustrating a process for strengthening military security based on natural language processing and image comparison in a soldier-based community application according to an embodiment.

Referring to FIG. 2, first, in step S201, the device 200 may receive an upload request for the first post from the first user terminal 110.

Specifically, when the upload request menu for the first post is selected while the soldier-based community application is running, the first user terminal 110 may transmit an upload request for the first post to the device 200, The device 200 may receive an upload request for the first post from the first user terminal 110. Here, the first post may be composed of one or more texts, may be composed of one or more images, may be composed of only text, may be composed of only images, or may be composed of both text and images. It may be configured as follows.

In step S202, the device 200 may check whether the first post includes text.

If it is determined in step S202 that the first posting does not contain text, the device 200 may check whether the first posting includes an image in step S208.

If it is confirmed that the first posting contains text in step S202, the device 200 may extract the text included in the first posting as the first text in step S203.

Specifically, if the device 200 determines that the first posting contains a text consisting of one sentence, the device 200 may extract one sentence included in the first posting as the first text, and If it is confirmed that a text consisting of multiple sentences is included, one of the multiple sentences included in the first post can be extracted as the first text. At this time, if the first post includes text consisting of several sentences, steps S203 to S207 may be repeatedly performed depending on the number of sentences.

In step S204, the device 200 may input the first text into a first artificial intelligence model that has been trained in advance. Here, the first artificial intelligence model may be an algorithm that receives text as input, analyzes the probability that the text violates military security, and outputs the probability of the text violating military security.

For example, if the first artificial intelligence model receives the first text as input and analyzes the probability of a military security violation of the first text as 30%, it may output an output value indicating 30%. A detailed description of the process of analyzing and outputting the military security violation probability of text in the first artificial intelligence model will be described later with reference to FIG. 3.

In step S205, the device 200 may set the probability of a military security violation of the first text as the first probability based on the output of the first artificial intelligence model.

For example, the device 200 inputs the first text into the first artificial intelligence model, obtains an output from the first artificial intelligence model, and when the output of the first artificial intelligence model is confirmed to be an output value indicating 30%, , the first probability can be set to 30%

According to one embodiment, a first artificial intelligence model may be trained to analyze how likely it is that text violates military security. Through this, the first artificial intelligence model can analyze how likely the text is to violate military security and output an output value indicating what the probability of the text is to violate military security. To this end, the first artificial intelligence model may be trained in advance to analyze the probability that there is a military security problem with the text.

The learning device in which the first artificial intelligence model is learned may be the same device as the device 200 that sets the probability of a military security violation of text using the learned first artificial intelligence model, or may be a separate device. Below, the process by which the first artificial intelligence model is learned is explained.

First, the learning device can generate input based on specific text.

Specifically, the learning device can perform a preprocessing process on a specific text. The preprocessed text can be used as an input to the first artificial intelligence model, or the input can be generated through normal processing to remove unnecessary information.

Next, the learning device can apply input to the first artificial intelligence model. The first artificial intelligence model may be an artificial neural network learned according to reinforcement learning. The first artificial intelligence model may be a Q-Network, DQN (Depp Q-Network), or relational network (RL) structure suitable for outputting abstract inference through reinforcement learning.

The first artificial intelligence model learned according to reinforcement learning can be updated and optimized by reflecting the evaluation of various rewards.

For example, the first reward is through keyword comparison between keywords recognized in a specific text and military security violation keywords. If it is confirmed that there are keywords whose meanings are more similar than a certain level, the higher the similarity of the keywords, the more the specific text If the probability of a military security violation is analyzed as high, the compensation value may be increased. Here, military security violation keywords may be composed of keywords classified as having military security problems, and may be updated periodically. At this time, military security violation keywords can be set by the administrator who operates the soldier-based community application, or can be set independently through learning of the first artificial intelligence model. Information about military security violation keywords may be stored in a database of device 200.

That is, the first artificial intelligence model obtains information about military security violation keywords from the database of the device 200, and identifies military security violation of the specific text through keyword comparison between the keywords recognized in the specific text and the military security violation keywords. Probabilities can be analyzed.

Next, the learning device may obtain an output from the first artificial intelligence model. At this time, the output of the first artificial intelligence model may be information about the probability of a military security violation of a specific text. That is, the first artificial intelligence model can analyze the probability of a military security violation of a specific text through keyword comparison between keywords recognized in the specific text and military security violation keywords, and output information about the analyzed probability.

Next, the learning device may evaluate the output of the first artificial intelligence model and pay a reward.

For example, if the learning device determines that there are keywords whose meanings are more similar than a certain level through keyword comparison between keywords recognized in a specific text and military security violation keywords, the higher the similarity of the keywords, the better the specific text. If the probability of a military security violation is analyzed as high, the first reward can be awarded in large amounts.

Next, the learning device may update the first artificial intelligence model based on the evaluation.

Specifically, the learning device operates in specific states such that the expected value of the sum of rewards is maximized in an environment where the first artificial intelligence model analyzes the probability of a military security violation of text. The first artificial intelligence model can be updated through the process of optimizing the policy that determines the actions to be taken.

For example, if the learning device has no problem with the result of analyzing the military security violation probability of the first text with the first probability, information indicating that there is no problem with the result of analyzing the military security violation probability of the first text. When generating first learning data including, applying the first learning data to the first artificial intelligence model, and analyzing the probability of a military security violation of a text having a keyword similar to the first text, a value similar to the first text The first artificial intelligence model can be updated through the process of learning the first artificial intelligence model to have .

Meanwhile, the process of optimizing the policy can be accomplished through estimating the maximum expected value of the sum of rewards or the maximum value of the Q-function, or estimating the minimum value of the loss function of the Q-function. Estimation of the minimum value of the loss function can be done through stochastic gradient descent (SGD), etc. The process of optimizing the policy is not limited to this, and various optimization algorithms used in reinforcement learning can be used.

The learning device can gradually update the first artificial intelligence model by repeating the learning process of the first artificial intelligence model described above. Through this, the learning device can train a first artificial intelligence model that outputs the probability of a military security violation of the text when the text is input.

In other words, when the learning device analyzes how likely the text is to violate military security, it can learn the first artificial intelligence model by adjusting the analysis standard by reflecting reinforcement learning through the first reward, etc.

In step S206, the device 200 may check whether the first probability is lower than the first reference rate. Here, the first reference ratio may be set differently depending on the embodiment.

If the first probability is confirmed to be higher than the first reference rate in step S206, step S501 may be performed, and a detailed description regarding this will be described later with reference to FIG. 5.

If the first probability is confirmed to be lower than the first reference rate in step S206, the device 200 may determine that there is no security problem in the first text in step S207.

According to one embodiment, when the text included in the first post consists of several sentences, steps S203 to S207 may be repeatedly performed to determine whether there is a security problem for each sentence.

For example, if the text included in the first post consists of the first sentence and the second sentence, the device 200 extracts the first sentence as the first text, and converts the first text into the first artificial intelligence Input to the model, and based on the output of the first artificial intelligence model, set the military security violation probability of the first text as the first probability, and if the first probability is confirmed to be higher than the first reference rate, the first text It can be determined that there is no security problem, and then the second sentence is extracted as the first text, the first text is input into the first artificial intelligence model, and based on the output of the first artificial intelligence model, the first text is extracted. The probability of a military security violation of the text is set as the first probability, and if the first probability is confirmed to be higher than the first standard rate, it can be determined that there is no security problem in the first text.

That is, the device 200 determines whether there is a security problem with the text included in the first posting by performing steps S203 to S207 one or more times depending on the number of sentences in the text included in the first posting. This can be done, and if it is determined that there is no security problem with the text included in the first post, it can be confirmed whether the first post contains an image in step S208.

If it is determined in step S208 that the first posting does not contain an image, the device 200 may check whether the first posting includes text in step S209.

If it is determined in step S209 that the first post does not contain text, the device 200 sends a notification message to the first user terminal 110 informing that re-upload is necessary because the first post does not contain text and images. ), and then, in step S201, the device 200 may again receive an upload request for the first post from the first user terminal 110.

That is, in step S202, it is confirmed that the first post does not contain text, in step S208, it is confirmed that the first post does not contain an image, and in step S209, it is confirmed that the first post does not contain text. If confirmed, it can be confirmed that the first post does not contain text and images, and returning to step S201, the device 200 receives the upload request for the first post again from the first user terminal 110. can do.

If it is confirmed in step S209 that the first posting contains text, in step S215, the device 200 determines that there is no security problem with the text included in the first posting, and If it is determined that there is no security problem with the image, the status of the first post can be set to allow public.

Specifically, in step S202, it is confirmed that the first post contains text, and steps S203 to S207 are performed. Then, in step S208, it is confirmed that the first post does not contain an image, and in step S209, the first post is confirmed to contain text. If it is confirmed that the post contains text, it may be confirmed that the first post contains only text, and in step S215, the device 200 performs steps S203 to S207 to display the text included in the first post. It is determined that there is no security problem, and since the first post does not contain an image, it can be determined that there is no security problem with the image included in the first post, and through this, the first post does not contain an image. Since it is determined that there is no problem, the status of the first post can be set to allow public.

Meanwhile, if it is confirmed in step S208 that the first posting contains an image, the device 200 may extract the image included in the first posting as the first image in step S210.

Specifically, if it is confirmed that the first post contains one image, the device 200 may extract one image included in the first post as the first image, and multiple images may be included in the first post. If it is confirmed that it is included, any one of several images included in the first post can be extracted as the first image. At this time, if the first post includes several images, steps S210 to S214 may be repeatedly performed depending on the number of images.

In step S211, the device 200 may input the first image into a pre-trained second artificial intelligence model. Here, the second artificial intelligence model may be an algorithm that receives an image as input, analyzes the probability that the image violates military security, and outputs the probability of the image violating military security.

For example, if the second artificial intelligence model receives the first image as input and analyzes the probability of a military security violation of the first image as 40%, it may output an output value indicating 40%. A detailed description of the process of analyzing and outputting the military security violation probability of an image in the second artificial intelligence model will be described later with reference to FIG. 4.

In step S212, the device 200 may set the military security violation probability of the first image as the second probability based on the output of the second artificial intelligence model.

For example, the device 200 inputs the first image into the second artificial intelligence model, obtains an output from the second artificial intelligence model, and when the output of the second artificial intelligence model is confirmed to be an output value indicating 40%, , the second probability can be set to 40%

According to one embodiment, a second artificial intelligence model may be trained to analyze how likely an image is to violate military security. Through this, the second artificial intelligence model can analyze how likely the image is to violate military security and output an output value indicating how likely the image is to violate military security. To this end, the second artificial intelligence model may be pre-trained to analyze how likely it is that there is a military security problem with the image.

The learning device in which the second artificial intelligence model is learned may be the same device as the device 200 that sets the probability of a military security violation of the image using the learned second artificial intelligence model, or may be a separate device. Below, the process by which the second artificial intelligence model is learned is explained.

First, the learning device can generate input based on a specific image.

Specifically, the learning device may perform a preprocessing process on a specific image. The preprocessed image can be used as an input to the second artificial intelligence model, or the input can be generated through normal processing to remove unnecessary information.

Next, the learning device can apply input to the second artificial intelligence model. The second artificial intelligence model may be an artificial neural network learned according to reinforcement learning. The second artificial intelligence model may be a Q-Network, DQN (Depp Q-Network), or relational network (RL) structure suitable for outputting abstract inference through reinforcement learning.

The second artificial intelligence model learned according to reinforcement learning can be updated and optimized by reflecting the evaluation of various rewards.

For example, the second compensation is through image comparison between object images extracted from a specific image and military security violation images. If it is confirmed that there are objects whose shapes are more similar than a certain level, the higher the similarity of the objects, the more likely it is that the specific image will be If the probability of a military security violation is analyzed as high, the compensation value may be increased. Here, military security violation images may consist of images of objects classified as having military security problems, and may be updated periodically. At this time, military security violation images can be set by the administrator who runs the soldier-based community application, or they can be set themselves through learning of the second artificial intelligence model. Information about military security violation images may be stored in a database of device 200.

That is, the second artificial intelligence model obtains information about military security violation images from the database of the device 200, and compares the object image extracted from the specific image with the military security violation images to secure the military security of the specific image. The probability of violation can be analyzed.

Next, the learning device can obtain an output from the second artificial intelligence model. At this time, the output of the second artificial intelligence model may be information about the probability of a military security violation of a specific image. That is, the second artificial intelligence model can analyze the probability of a military security violation of a specific image through image comparison between object images extracted from the specific image and military security violation images, and output information about the analyzed probability.

Next, the learning device can evaluate the output of the second artificial intelligence model and pay a reward.

For example, if it is confirmed that there is an object whose shape is more similar than a certain level through image comparison between object images extracted from a specific image and military security violation images, the learning device determines that the higher the similarity between the objects, the more likely it is that the military security of the specific image If the probability of a security violation is analyzed as high, a large secondary reward can be awarded.

Next, the learning device may update the second artificial intelligence model based on the evaluation.

Specifically, the learning device operates in specific states such that the expected value of the sum of rewards is maximized in an environment where the second artificial intelligence model analyzes the probability of a military security violation of the image. The second artificial intelligence model can be updated through the process of optimizing the policy that determines the actions to be taken.

For example, if the learning device has no problem with the result of analyzing the military security violation probability of the first image with the second probability, information indicating that there is no problem with the result of analyzing the military security violation probability of the first image. When generating second learning data including, and applying the second learning data to the second artificial intelligence model to analyze the probability of a military security violation of an image having an object similar to the object recognized in the first image, the first The second artificial intelligence model can be updated through the process of training the second artificial intelligence model to have similar values to the image.

Meanwhile, the process of optimizing the policy can be accomplished through estimating the maximum expected value of the sum of rewards or the maximum value of the Q-function, or estimating the minimum value of the loss function of the Q-function. Estimation of the minimum value of the loss function can be done through stochastic gradient descent (SGD), etc. The process of optimizing the policy is not limited to this, and various optimization algorithms used in reinforcement learning can be used.

The learning device can gradually update the second artificial intelligence model by repeating the learning process of the second artificial intelligence model described above. Through this, the learning device can train a second artificial intelligence model that outputs the probability of a military security violation of the image when the image is input.

In other words, when analyzing the probability that an image violates military security, the learning device can learn the second artificial intelligence model by adjusting the analysis standard by reflecting reinforcement learning through second compensation, etc.

In step S213, the device 200 may check whether the second probability is lower than the second reference rate. Here, the second reference ratio may be set differently depending on the embodiment.

If the second probability is confirmed to be higher than the second reference rate in step S213, step S601 may be performed, and a detailed description regarding this will be described later with reference to FIG. 6.

If the second probability is confirmed to be lower than the second reference rate in step S213, the device 200 may determine that there is no security problem in the first image in step S214.

According to one embodiment, when the first post includes several images, steps S210 to S214 may be repeatedly performed to determine whether there is a security problem for each image.

For example, if the first post includes a 1-1 image and a 1-2 image, the device 200 extracts the 1-1 image as the first image and uses the first image as the second artificial image. Input to the intelligence model, and based on the output of the second artificial intelligence model, set the military security violation probability of the first image as the second probability, and if the second probability is confirmed to be higher than the second reference rate, the first image It can be determined that there is no security problem, and then the 1-2 images are extracted as the first image, the first image is input into the second artificial intelligence model, and based on the output of the second artificial intelligence model, , the military security violation probability of the first image is set as the second probability, and if the second probability is confirmed to be higher than the second standard rate, it can be determined that there is no security problem in the first image.

That is, the device 200 performs steps S210 to S214 one or more times depending on the number of images included in the first posting to determine whether there is a security problem with the image included in the first posting. If it is determined that there is no security problem with the image included in the first posting, in step S215, the device 200 determines that there is no security problem with the text included in the first posting, and If it is determined that there is no security problem with the image included in the post, the status of the first post can be set to allow public.

Specifically, in step S202, it is confirmed that the first post contains text, and steps S203 to S207 are performed. Then, in step S208, it is confirmed that the first post contains an image, and steps S210 to S214 are performed. If so, it can be confirmed that the first post contains text and images, and in step S215, the device 200 performs steps S203 to S207 to ensure that there is no security problem with the text included in the first post. It is determined that there is no security problem with the image included in the first post by performing steps S210 to S214. Through this, it is determined that there is no security problem with the first post, and the first post is determined to have no security problem. You can set the status of the post to public.

In addition, if it is confirmed in step S202 that the first post does not contain text, and it is confirmed that the first post contains an image in step S208, and steps S210 to S214 are performed, only the image is included in the first post. It can be confirmed that there is a security problem with the text included in the first posting, and in step S215, the device 200 determines that there is no security problem with the text included in the first posting, and in steps S210 to S214. The step may be performed to determine that there is no security problem with the image included in the first post, and through this, it may be determined that there is no security problem with the first post, and the status of the first post may be set to allow public. You can.

If the status of the first post is set to allow disclosure, the device 200 may upload the first post to the soldier-based community and process the first post to be made public to the soldier-based community.

Figure 3 is a flow chart to explain the process of analyzing and outputting the probability of a military security violation of text according to an embodiment.

Referring to FIG. 3, first, in step S301, when the first text is input, the first artificial intelligence model can recognize key keywords through natural language processing on the first text.

Specifically, the first artificial intelligence model can analyze keywords by performing morphological analysis, syntactic analysis, semantic analysis, and pragmatic analysis on the first text. , Based on the analysis results, key keywords can be derived and recognized from the first text.

For example, if the text “The first unit is stationed in the first area” is entered, the first artificial intelligence model will use natural language processing on the text to read “First area” and “First unit.” ” can be derived and recognized as key keywords.

When recognizing core keywords, the first artificial intelligence model can analyze keywords frequently used in the military and generate a list of major keywords frequently used in the military. Based on the list of important keywords, core keywords can be derived and recognized. there is. Here, the important keyword list may be composed of keywords used more than a certain amount during a preset period in the military-based community, and may be updated periodically.

For example, when the first artificial intelligence model recognizes the first keyword, second keyword, and third keyword through natural language processing of the first text, the first keyword, second keyword, and third keyword are included in the important keyword list. As a result of checking whether or not it is included, if it is confirmed that only the first keyword is included in the important keyword list, the first keyword can be derived from the first text as a key keyword and recognized.

In step S302, the first artificial intelligence model may set the key keyword recognized in the first text as the first keyword.

Specifically, if only one core keyword is recognized in the first text, the first artificial intelligence model can set one core keyword recognized in the first text as the first keyword, and multiple core keywords are recognized in the first text. If so, one of several key keywords recognized in the first text can be set as the first keyword. At this time, if several key keywords are recognized in the first text, steps S302 to S303 may be repeatedly performed depending on the number of key keywords.

In step S303, the first artificial intelligence model may analyze the probability that there is a military security problem for the first keyword through keyword comparison between the first keyword and pre-registered military security violation keywords. Here, military security violation keywords may be composed of keywords classified as having military security problems, and may be updated periodically. At this time, military security violation keywords can be set by the administrator who operates the soldier-based community application, or can be set independently through learning of the first artificial intelligence model. Information about military security violation keywords may be stored in a database of device 200.

That is, the first artificial intelligence model obtains information about military security violation keywords from the database of the device 200, and compares keywords between the first keyword and military security violation keywords to determine whether there is a military security problem for the first keyword. The probability that there is can be analyzed. At this time, the first artificial intelligence model can select the keyword whose meaning is most similar to the first keyword among the military security violation keywords through keyword comparison between the first keyword and the military security violation keywords, and the selected keyword and the first keyword. Through keyword comparison between keywords, the probability that there is a military security problem for the first keyword can be analyzed.

Specifically, the first artificial intelligence model compares the first keyword and military security violation keywords, calculates the similarity with the first keyword for each military security violation keyword, and selects the keyword with the highest similarity among the military security violation keywords. The higher the similarity between the selected keyword and the first keyword, the higher the probability that there is a military security problem with the first keyword. At this time, similarity is an indicator of how similar the meanings of words are. The more similar the word meanings between keywords are, the higher the value can be calculated. For the same keywords, the similarity can be calculated as 100%. For this purpose, one or more meanings are registered for each of the military security violation keywords, and the first artificial intelligence model determines the meaning of the first keyword and then compares the meaning of the first keyword with the meaning of each of the military security violation keywords. , the probability that there is a military security problem for the first keyword can be analyzed.

According to one embodiment, when multiple core keywords are recognized in the first text, steps S302 to S303 are repeatedly performed according to the number of core keywords, and the probability of a military security problem for each core keyword can be analyzed.

In step S304, when the probability that there is a military security problem for each key keyword recognized in the first text is analyzed, the first artificial intelligence model may classify the probability analyzed for each key keyword into a first probability group.

For example, when the first artificial intelligence model recognizes only one core keyword in the first text, the probability that there is a military security problem for the one core keyword recognized in the first text is analyzed as a first ratio, The first ratio can be classified into the first probability group.

In addition, the first artificial intelligence model shows that when the 1-1 keyword and the 1-2 keyword are recognized as key keywords in the first text, the probability that there is a military security problem for the 1-1 keyword is increased to the first ratio. If the probability that there is a military security problem for the first and second keywords is analyzed as a second ratio, the first ratio and the second ratio can be classified into a first probability group.

In step S305, the first artificial intelligence model may set the highest probability among the probabilities classified into the first probability group as the first probability.

Specifically, when there is one probability classified into the first probability group, the first artificial intelligence model can set one probability classified into the first probability group as the first probability, and the probability classified into the first probability group is If there are two or more, the probabilities classified into the first probability group may be compared to select the highest probability, and the selected probability may be set as the first probability.

In step S306, the first artificial intelligence model may output an output value indicating the first probability. At this time, the device 200 may obtain the output of the first artificial intelligence model.

Figure 4 is a flow chart to explain the process of analyzing and outputting the probability of a military security violation of an image according to an embodiment.

Referring to FIG. 4, first, in step S401, when the first image is input, the second artificial intelligence model may recognize the object through image analysis of the first image.

Specifically, when there is only one object in the first image, the second artificial intelligence model can recognize one object through image analysis of the first image, and when there are multiple objects in the first image, the second artificial intelligence model can recognize one object through image analysis of the first image. 1 Each of multiple objects can be recognized through image analysis of the image.

In step S402, the second artificial intelligence model may set the object recognized in the first image as the first object.

Specifically, if only one object is recognized in the first image, the second artificial intelligence model can set one object recognized in the first image as the first object, and if multiple objects are recognized in the first image, the second artificial intelligence model can set the first object as the first object. 1 Any one of several objects recognized in the image can be set as the first object. At this time, when multiple objects are recognized in the first image, steps S402 to S405 may be repeatedly performed depending on the number of objects.

In step S403, the second artificial intelligence model may distinguish the area occupied by the first object in the first image as the first area.

Specifically, the second artificial intelligence model recognizes the outline of the first object in the first image, identifies the area inside the outline of the first object as the area occupied by the first object, and determines the area occupied by the first object. can be divided into the first area.

In step S404, the second artificial intelligence model may extract the second image by dividing the portion containing the first area in the first image.

That is, the second artificial intelligence model can segment the part where the first area is in the first image and extract the segmented image as the second image.

In step S405, the second artificial intelligence model may analyze the probability that there is a military security problem for the first object through image comparison between the second image and pre-registered military security violation images. Here, military security violation images may consist of images of objects classified as having military security problems, and may be updated periodically. At this time, military security violation images can be set by the administrator who runs the soldier-based community application, or they can be set themselves through learning of the second artificial intelligence model. Information about military security violation images may be stored in a database of device 200.

That is, the second artificial intelligence model obtains information about military security violation images from the database of the device 200, and compares the images between the second image and the military security violation images to determine whether there is a military security violation for the first object. The probability that there is can be analyzed. At this time, the second artificial intelligence model can select the image with the object most similar to the first object among the military security violation images through image comparison between the second image and the military security violation images, and match the selected image with the first object. Through comparison between the two images, the probability that there is a military security problem for the first object can be analyzed.

Specifically, the second artificial intelligence model compares the second image and the military security violation images, calculates the similarity with the first object for each military security violation image, and the object with the highest similarity among the military security violation images is selected. Images are selected, and the higher the similarity between the selected image and the first object, the higher the probability that there is a military security problem with the first object. At this time, similarity is an indicator of how similar the shapes of objects are. The more similar the shapes between objects are, the higher the value can be calculated. For identical objects, the similarity can be calculated as 100%. For this purpose, each of the military security violation images may be composed of an object image, and the second artificial intelligence model compares the second image, which is an image of the first object, with each of the military security violation images, and determines the military security violation images for the first object. The probability of a security problem can be analyzed.

According to one embodiment, when multiple objects are recognized in the first image, steps S402 to S405 are repeatedly performed depending on the number of objects to analyze the probability that each object has a military security problem.

In step S406, when the probability that there is a military security problem for each object recognized in the first image is analyzed, the second artificial intelligence model may classify the analyzed probability for each object into a second probability group.

For example, if only one object is recognized in the first image, the second artificial intelligence model analyzes the probability that there is a military security problem for one object recognized in the first image at a first rate. The ratios can be classified into a second probability group.

In addition, the second artificial intelligence model analyzes the probability that there is a military security problem for the 1-1 object as a first ratio when the 1-1 object and the 1-2 object are recognized in the first image, If the probability that there is a military security problem for the 1-2 object is analyzed as a second ratio, the first ratio and the second ratio can be classified into a second probability group.

In step S407, the second artificial intelligence model may set the highest probability among the probabilities classified into the second probability group as the second probability.

Specifically, when there is one probability classified into the second probability group, the second artificial intelligence model can set one probability classified into the second probability group as the second probability, and the probability classified into the second probability group is If there are two or more, the probabilities classified into the second probability group may be compared to select the highest probability, and the selected probability may be set as the second probability.

In step S408, the second artificial intelligence model may output an output value indicating the second probability. At this time, the device 200 may obtain the output of the second artificial intelligence model.

Figure 5 is a flow chart to explain the process of determining whether to make a post public according to the number of keywords with security problems according to an embodiment.

Referring to FIG. 5, first, in step S501, if the first probability is confirmed to be higher than the first reference rate, the device 200 obtains information about the probability classified into the first probability group from the first artificial intelligence model. can do.

Specifically, the device 200 may input a first text into a first artificial intelligence model and obtain an output value indicating a first probability from the first artificial intelligence model, where the first probability is higher than the first reference rate. If confirmed, information about the probability classified into the first probability group can be obtained from the first artificial intelligence model.

That is, the highest probability among the probabilities classified into the first probability group is set as the first probability, and if the first probability is confirmed to be higher than the first standard rate, the first standard rate among the probabilities classified into the first probability group is set as the first probability. In order to determine how many probabilities are higher than the ratio, the device 200 may obtain information about the probabilities classified into the first probability group from the first artificial intelligence model.

In step S502, the device 200 may classify a probability higher than the first reference rate among the probabilities classified into the first probability group into the third probability group.

In step S503, the device 200 may set the number of probabilities classified into the third probability group as a first value.

For example, if 25%, 35%, and 45% are classified as the first probability group, device 200 may classify 35% and 45% as the third probability group if the first reference percentage is set to 30%. You can classify them and set two as the first number.

In step S504, the device 200 may check whether the first numerical value is lower than the first reference value. Here, the first reference value may be set differently depending on the embodiment.

If the first value is confirmed to be lower than the first reference value in step S504, the device 200 may determine that uploading of the first post is possible through partial modification of the first text in step S505.

If it is determined that uploading of the first post is possible through partial modification of the first text, the device 200 may change the alternative keyword or perform mosaic processing to set the status of the first post to allow public disclosure. A detailed description regarding this will be provided later with reference to FIG. 7 .

If it is confirmed in step S504 that the first value is higher than the first reference value, in step S506, the device 200 may set the status of the first post as not public.

When the status of the first post is set to not public, the device 200 sends a notification message to the first user terminal indicating that the upload of the first post has been canceled due to a security problem with the text included in the first post. It can be sent to (110). At this time, the notification message may include a notification message indicating which part of the text has a security problem.

Figure 6 is a flowchart illustrating a process for determining whether to make a post public according to the number of objects with security problems according to an embodiment.

Referring to FIG. 6, first, in step S601, if the second probability is confirmed to be higher than the second reference rate, the device 200 obtains information about the probability classified into the second probability group from the second artificial intelligence model. can do.

Specifically, the device 200 may input the first image into a second artificial intelligence model and obtain an output value indicating a second probability from the second artificial intelligence model, where the second probability is higher than the second reference rate. If confirmed, information about the probability classified into the second probability group can be obtained from the second artificial intelligence model.

That is, the highest probability among the probabilities classified into the second probability group is set as the second probability, and if the second probability is confirmed to be higher than the second standard rate, the second standard rate among the probabilities classified into the second probability group is set as the second probability. In order to determine how many probabilities are higher than the ratio, the device 200 may obtain information about the probabilities classified into the second probability group from the second artificial intelligence model.

In step S602, the device 200 may classify a probability higher than the second reference rate among the probabilities classified into the second probability group into the fourth probability group.

In step S603, the device 200 may set the number of probabilities classified into the fourth probability group as a second value.

For example, if 25%, 35%, and 45% are classified as the second probability group, the device 200 classifies 45% as the fourth probability group if the second reference ratio is set to 40%, One can be set as the second value.

In step S604, the device 200 may check whether the second numerical value is lower than the second reference value. Here, the second reference value may be set differently depending on the embodiment.

If the second value is confirmed to be lower than the second reference value in step S604, the device 200 may determine that uploading of the first post is possible through partial modification of the first image in step S605.

If it is determined that uploading of the first post is possible through partial modification of the first image, the device 200 may change the replacement object image or perform mosaic processing to set the status of the first post to allow public access. A detailed description regarding this will be provided later with reference to FIG. 8 .

If it is confirmed in step S604 that the second value is higher than the second reference value, in step S606, the device 200 may set the status of the first post as not public.

When the status of the first post is set to not public, the device 200 sends a notification message to the first user terminal indicating that the upload of the first post has been canceled due to a security problem with the image included in the first post. It can be sent to (110). At this time, the notification message may include a notification message indicating which part of the image has a security problem.

FIG. 7 is a flowchart illustrating a process of setting the status of the first post to allow public by performing alternative keyword change or mosaic processing according to an embodiment.

Referring to FIG. 7, first, in step S701, the device 200 selects key keywords that have military security problems among key keywords included in the first text, based on the probability classified into the third probability group. You can.

For example, if keywords 1-1, 1-2, and 1-3 are recognized as core keywords in the first text, the probability that there is a military security problem for the 1-1 keyword is 1%. If the probability that there is a military security problem for the 1-2 keywords is analyzed as a second ratio, and the probability that there is a military security problem for the 1-3 keywords is analyzed as a third ratio, the device ( 200) can classify the first ratio, second ratio, and third ratio into the first probability group, and only the first ratio and second ratio among the first ratio, second ratio, and third ratio are classified into the third probability group. Once classified, based on the first ratio and the second ratio, the 1-1 keyword and 1-2 keyword can be selected as key keywords with military security problems.

In step S702, the device 200 may create a first list with selected key keywords.

For example, if the 1-1 keyword and the 1-2 keyword are selected as key keywords with military security problems, the device 200 creates a first list with the 1-1 keyword and the 1-2 keyword. can be created.

In step S703, the device 200 may set any one of the key keywords included in the first list as the second keyword.

Specifically, if it is confirmed that the first list contains only one core keyword, the device 200 may set one core keyword included in the first list as the second keyword, and may set several core keywords included in the first list. If it is confirmed that a core keyword is included, any one of the core keywords included in the first list can be set as the second keyword. At this time, if the first list includes several core keywords, steps S703 to S706 may be repeatedly performed depending on the number of core keywords.

In step S704, the device 200 may check whether a replacement keyword for the second keyword is registered. Here, the replacement keyword may mean a keyword set in advance to replace a keyword that has a problem with military security with a keyword that does not have a problem with military security. To this end, keyword information classified by keyword is stored in the database of the device 200, and the keyword information may include information about alternative keywords. At this time, the alternative keyword may be set by the administrator who operates the soldier-based community application, or may be set itself through learning of the first artificial intelligence model.

That is, the device 200 can obtain second keyword information by searching the database and, based on the second keyword information, check whether an alternative keyword is registered in the second keyword.

If it is confirmed in step S704 that an alternative keyword for the second keyword is registered and a third keyword is registered as an alternative keyword for the second keyword, in step S705, the device 200 selects the second keyword from the first text. You can change the keyword to a third keyword.

If it is confirmed in step S704 that a replacement keyword for the second keyword is not registered, in step S706, the device 200 may mosaic the portion of the first text containing the second keyword.

According to one embodiment, when the first list includes several core keywords, steps S703 to S706 may be repeatedly performed to change alternative keywords or mosaic processing for each core keyword.

In step S707, when alternative keywords are changed or mosaic processing is performed for each core keyword included in the first list, the device 200 may set the status of the first post to allow public disclosure.

FIG. 8 is a flowchart illustrating a process for setting the status of the first post to be made public by performing replacement object image change or mosaic processing according to an embodiment.

Referring to FIG. 8, first, in step S801, the device 200 may select objects that pose a military security problem among the objects included in the first image based on the probability classified into the fourth probability group. .

For example, if the 1-1 object, the 1-2 object, and the 1-3 object are recognized in the first image, the probability that there is a military security problem for the 1-1 object is analyzed as a first ratio. , if the probability that there is a military security problem for the 1-2 object is analyzed at the second ratio, and the probability that there is a military security problem for the 1-3 object is analyzed at the third ratio, the device 200 The first ratio, the second ratio, and the third ratio can be classified into the second probability group, and if only the first ratio and the second ratio among the first ratio, second ratio, and third ratio are classified into the fourth probability group, Based on the first ratio and the second ratio, the 1-1 object and the 1-2 object can be selected as objects that pose a military security problem.

In step S802, the device 200 may create a second list with the selected objects. At this time, the device 200 may create a second list with object images for the selected objects. Each object image can be extracted through the area occupied by the object.

For example, when the 1-1 object and the 1-2 object are selected as objects with military security problems, the device 200 creates a second list with the 1-1 object and the 1-2 object. can do.

In step S803, the device 200 may set one of the objects included in the second list as the second object.

Specifically, if it is confirmed that the second list includes only one object, the device 200 may set one object included in the second list as the second object, and if multiple objects are included in the second list. If it is confirmed that it is included, any one of the objects included in the second list can be set as the second object. At this time, if the second list includes several objects, steps S803 to S806 may be repeatedly performed depending on the number of objects.

In step S804, the device 200 may check whether a replacement object for the second object is registered. Here, the replacement object may mean an object set in advance to replace an object with a military security problem with an object without a military security problem. To this end, object information classified by object is stored in the database of the device 200, and the object information may include information about alternative object images. At this time, the replacement object image can be set by the administrator who operates the soldier-based community application, or it can be set itself through learning of the first artificial intelligence model.

That is, the device 200 can query the database to obtain second object information and, based on the second object information, check whether a replacement object is registered in the second object.

If it is confirmed in step S804 that a replacement object for the second object is registered and a third object is registered as a replacement object for the second object, in step S805, the device 200 selects the second object in the first image. If the area occupied by the object is divided into a second area, the part of the first image containing the second area can be changed to the image of the third object.

That is, when it is confirmed that the third object is registered as a replacement object for the second object, the device 200 deletes the second object in the first image and adds the third object in the deleted place, thereby creating the first image. The image can be processed to display a third object by replacing the second object.

If it is confirmed in step S804 that a replacement object for the second object is not registered, the device 200 may mosaic the second area in the first image in step S806.

According to one embodiment, when the second list includes several objects, steps S803 to S806 may be repeatedly performed to change or mosaic a replacement object image for each object.

In step S807, when replacement object image change or mosaic processing is performed for each object included in the second list, the device 200 may set the status of the first post to be made public.

Figure 9 is a flow chart to explain the process of adding military security violation keywords.

Referring to FIG. 9 , first, in step S901, the device 200 may classify posts whose status is set as not public among posts requested to be uploaded during a preset period into a first post group.

For example, the device 200 may classify posts whose status is set as not public among posts requested to be uploaded in the past one week into the first post group.

In step S902, the device 200 may set the number of posts classified into the first post group to a third value.

In step S903, the device 200 may confirm that the first post is included in the first post group.

In step S904, if it is confirmed that the first post is included in the first post group, the device 200 selects the third list of the key keywords included in the first text, excluding the key keywords included in the first list. can be created.

That is, the device 200 selects core keywords that have military security problems among the core keywords included in the first text, creates a first list with the selected core keywords, and selects core keywords that have military security problems among the core keywords included in the first text. You can select key keywords that do not have military security issues and create a third list using the selected key keywords.

In step S905, the device 200 may set any one of the key keywords included in the third list as the third keyword.

Specifically, if it is confirmed that the third list contains only one core keyword, the device 200 may set one core keyword included in the third list as the third keyword, and may set several core keywords included in the third list. If it is confirmed that a core keyword is included, any one of the core keywords included in the third list can be set as the third keyword. At this time, if the third list includes several core keywords, steps S905 to S914 may be repeatedly performed depending on the number of core keywords.

In step S906, the device 200 may classify a post containing the third keyword among the first post group into a second post group.

That is, the device 200 checks whether the third keyword is included based on the text included in each post classified into the first post group, and selects the post confirmed to contain the third keyword. 2 Posts can be classified into groups.

In step S907, the device 200 may set the number of posts classified into the second post group to a fourth value.

In step S908, the device 200 checks how many military security violation keywords are included in each post classified into the second post group, and calculates the number of military security violation keywords included in each post classified into the second post group. can be set as the first quantity.

For example, if the first post, the second post, and the third post are classified into the second post group, the device 200 determines that the first post includes several military security violation keywords based on the text included in the first post. By checking whether the first post contains five military security violation keywords, and based on the text included in the second post, check how many military security violation keywords it contains. So, it was confirmed that the 2nd post contains 4 military security violation keywords, and based on the text included in the 3rd post, it was checked how many military security violation keywords were included, and the 3rd post was After confirming that 6 military security violation keywords are included, the first quantity of the first post is set to 5, the first quantity of the second post is set to 4, and the first quantity of the third post is set to 6. It can be set to dog.

That is, the device 200 can check how many military security violation keywords are included in the text included in each post and set the number of military security violation keywords included in each post as the first quantity. there is.

In step S909, the device 200 may set the average of the first quantity set for each post classified into the second post group as the fifth value.

For example, when the device 200 sets the first quantity of the first post to 5, sets the first quantity of the second post to 4, and sets the first quantity of the third post to 6, 5 The average of 5, 4, and 6 can be calculated, and the fifth number can be set to 5.

In step S910, the device 200 may assign a higher first score as the value obtained by dividing the fourth numerical value by the third numerical value is larger.

For example, if the third numerical value is 10 and the fourth numerical value is 3, the device 200 may calculate the fourth numerical value divided by the third numerical value as 0.3 and give the first score as 3 points. , if the third numerical value is 10 and the fourth numerical value is 4, the fourth numerical value divided by the third numerical value is calculated as 0.4, and the first score can be given as 5 points.

In step S911, the device 200 may assign a higher second score as the value of the fifth numerical value increases.

For example, if the value of the fifth numerical value is 5, the device 200 may grant a second score of 5 points, and if the value of the fifth numerical value is 6, the device 200 may grant a second score of 6 points. .

In step S912, the device 200 may calculate a third score, which is the probability of a military security violation of the third keyword, as a value obtained by adding the first score and the second score. At this time, the device 200 applies the first weight to the first score, applies the second weight to the second score, and then calculates the value by adding the first score to which the first weight is applied and the second score to which the second weight is applied. , a third score can be calculated. Here, the first weight and the second weight may be set differently depending on the embodiment, and the first weight may be set to a higher value than the second weight.

In step S913, the device 200 may check whether the third score is higher than the reference score. Here, the reference score may be set differently depending on the embodiment.

If it is confirmed in step S913 that the third score is lower than the reference score, the device 200 may check whether the first list includes key keywords other than the keyword that calculated the military security violation score.

If the device 200 determines that the first list includes key keywords other than the keywords for which the military security violation score was calculated, the device 200 returns to step S905, and the device 200 selects one of the key keywords included in the third list. You can repeat the process of setting any one as the third keyword.

If the device 200 determines that the first list does not contain any key keywords other than the keywords for which the military security violation score was calculated, after a certain period of time has passed, the device 200 returns to step S901 and uploads the posts requested for upload during the preset period. You can repeat the process of classifying posts whose status is set to not public into the first post group.

If the third score is confirmed to be higher than the reference score in step S913, the device 200 may add the third keyword to the military security violation keywords in step S914.

According to one embodiment, if the third keyword includes several core keywords, steps S905 to S914 are repeatedly performed to calculate a military security violation score for each keyword, and the keywords with the calculated score above a certain level are selected as military security violations. It can be added to security violation keywords.

Figure 10 is an exemplary diagram of the configuration of a device according to an embodiment.

Device 200 according to one embodiment includes a processor 210 and memory 220. The processor 210 may include at least one device described above with reference to FIGS. 1 to 9 or may perform at least one method described above with reference to FIGS. 1 to 9 . A person or organization using the device 200 may provide services related to some or all of the methods described above with reference to FIGS. 1 to 9 .

The memory 220 may store information related to the methods described above or store a program in which methods described later are implemented. Memory 220 may be volatile memory or non-volatile memory.

The processor 210 can execute programs and control the device 200. The code of the program executed by the processor 210 may be stored in the memory 220. The device 200 is connected to an external device (eg, a personal computer or a network) through an input/output device (not shown) and can exchange data through wired or wireless communication.

The device 200 may be used to learn an artificial intelligence model or use a learned artificial intelligence model. Memory 220 may include a learning or learned artificial intelligence model. The processor 210 may learn or execute the algorithm of the artificial intelligence model stored in the memory 220. The learning device that trains the artificial intelligence model and the device 200 that uses the learned artificial intelligence model may be the same or may be separate.

The embodiments described above may be implemented with hardware components, software components, and/or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, and a field programmable gate (FPGA). It may be implemented using one or more general-purpose or special-purpose computers, such as an array, programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. A processing device may execute an operating system (OS) and one or more software applications that run on the operating system. Additionally, a processing device may access, store, manipulate, process, and generate data in response to the execution of software. For ease of understanding, a single processing device may be described as being used; however, those skilled in the art will understand that a processing device includes multiple processing elements and/or multiple types of processing elements. It can be seen that it may include. For example, a processing device may include a plurality of processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are possible.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., singly or in combination. Program instructions recorded on the medium may be specially designed and configured for the embodiment or may be known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -Includes optical media (magneto-optical media) and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions include machine language code, such as that produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing unit to operate as desired, or may be processed independently or collectively. You can command the device. Software and/or data may be used on any type of machine, component, physical device, virtual equipment, computer storage medium or device to be interpreted by or to provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

Although the embodiments have been described with limited drawings as described above, those skilled in the art can apply various technical modifications and variations based on the above. For example, the described techniques are performed in a different order than the described method, and/or components of the described system, structure, device, circuit, etc. are combined or combined in a different form than the described method, or other components are used. Alternatively, appropriate results may be achieved even if substituted or substituted by an equivalent.

Therefore, other implementations, other embodiments, and equivalents of the claims also fall within the scope of the following claims.

Claims (5)

A method for enhancing military security based on natural language processing and image matching in a soldier-based community application that provides a soldier-based community, a community related to military life, performed by a device, comprising:
Receiving an upload request for a first post from a first user terminal;
When it is confirmed that the first posting contains text, extracting the text included in the first posting as the first text;
Inputting the first text into a pre-trained first artificial intelligence model, and setting the probability of a military security violation of the first text as a first probability based on the output of the first artificial intelligence model;
If the first probability is confirmed to be lower than a preset first reference rate, determining that there is no security problem in the first text;
When it is confirmed that the first post contains an image, extracting the image included in the first post as a first image;
Inputting the first image into a pre-trained second artificial intelligence model, and setting the military security violation probability of the first image to a second probability based on the output of the second artificial intelligence model;
If the second probability is confirmed to be lower than a preset second reference rate, determining that there is no security problem in the first image; and
If it is determined that there is no security problem with the text included in the first posting and that there is no security problem with the image included in the first posting, the status of the first posting is set to allow public. Includes steps,
The first artificial intelligence model is,
When the first text is input, key keywords are recognized through natural language processing of the first text,
Setting the core keyword recognized in the first text as the first keyword,
Analyzing the probability that there is a military security problem for the first keyword through keyword comparison between the first keyword and pre-registered military security violation keywords,
When the probability that there is a military security problem for each key keyword recognized in the first text is analyzed, the probability analyzed for each key keyword is classified into a first probability group,
Set the highest probability among the probabilities classified in the first probability group as the first probability,
Outputs an output value indicating the first probability,
The second artificial intelligence model is,
When the first image is input, an object is recognized through image analysis of the first image,
Setting the object recognized in the first image as the first object,
Divide the area occupied by the first object in the first image into a first area,
Extracting a second image by dividing the portion containing the first area in the first image,
Analyzing the probability that there is a military security problem for the first object through image comparison between the second image and pre-registered military security violation images,
When the probability that there is a military security problem for each object recognized in the first image is analyzed, the probability analyzed for each object is classified into a second probability group,
Set the highest probability among the probabilities classified in the second probability group as the second probability,
Output an output value indicating the second probability,
If the first probability is confirmed to be higher than the first reference rate, obtaining information about the probability classified into the first probability group from the first artificial intelligence model;
classifying a probability higher than the first reference rate among the probabilities classified into the first probability group into a third probability group;
setting the number of probabilities classified into the third probability group as a first numerical value;
If the first value is confirmed to be lower than a preset first standard value, determining that uploading of the first post is possible through partial modification of the first text;
If the first value is confirmed to be higher than the first standard value, setting the status of the first post as not public;
If the second probability is confirmed to be higher than the second reference rate, obtaining information about the probability classified into the second probability group from the second artificial intelligence model;
classifying a probability higher than the second reference rate among the probabilities classified into the second probability group into a fourth probability group;
setting the number of probabilities classified into the fourth probability group as a second numerical value;
If the second value is confirmed to be lower than a preset second standard value, determining that uploading to the first post is possible through partial modification of the first image; and
If the second value is confirmed to be higher than the second reference value, further comprising setting the status of the first post as not public,
How to enhance military security based on natural language processing and image matching in soldier-based community applications. delete delete According to paragraph 1,
When it is determined that uploading of the first post is possible through partial modification of the first text, based on the probability classified into the third probability group, military security issues among key keywords included in the first text Selecting core keywords and generating a first list using the selected core keywords;
setting one of the key keywords included in the first list as a second keyword;
When it is confirmed that a third keyword is registered as a replacement keyword for the second keyword, changing the second keyword to the third keyword in the first text;
When it is confirmed that a replacement keyword for the second keyword is not registered, mosaicizing a portion of the first text containing the second keyword; and
When alternative keyword change or mosaic processing is performed for each key keyword included in the first list, setting the status of the first post to allow public, further comprising setting the status of the first post to allow public,
If it is determined that uploading to the first post is possible through partial modification of the first image,
Based on the probability classified into the fourth probability group, selecting objects that pose military security problems among the objects included in the first image and generating a second list with the selected objects;
setting one of the objects included in the second list as a second object;
If it is confirmed that a third object is registered as a replacement object for the second object, and if the area occupied by the second object in the first image is divided into a second area, the second area in the first image changing the part with an image of the third object;
When it is confirmed that a replacement object for the second object is not registered, mosaicing the second area in the first image; and
Further comprising setting the status of the first post to allow public when replacement object image change or mosaic processing is performed for each object included in the second list,
How to enhance military security based on natural language processing and image matching in soldier-based community applications. According to paragraph 4,
Classifying posts whose status is set as not public among posts requested to be uploaded during a preset period into a first post group;
setting the number of posts classified into the first post group to a third number;
When it is confirmed that the first post is included in the first post group, generating a third list using the core keywords included in the first text excluding the core keywords included in the first list;
setting one of the key keywords included in the third list as a third keyword;
classifying a post containing the third keyword among the first post group into a second post group;
setting the number of posts classified into the second post group to a fourth number;
By checking how many of the military security violation keywords are included in each post classified into the second post group, if the number of military security violation keywords included in each post classified into the second post group is confirmed, Setting the first quantity of each post classified into the second post group to the number of confirmed military security violation keywords included in each post classified into the second post group;
When the first quantity is set for each post classified into the second post group, setting an average of the first quantity set for each post classified into the second post group as a fifth value;
assigning a higher first score as the value obtained by dividing the fourth value by the third value is larger;
assigning a higher second score as the value of the fifth numerical value increases;
Calculating a third score, which is a military security violation score of the third keyword, by adding the first score and the second score; and
If the third score is confirmed to be higher than a preset standard score, further comprising adding the third keyword to the military security violation keywords,
How to enhance military security based on natural language processing and image matching in soldier-based community applications.