KR101908665B1 - Artificial intelligence system for detecting life cycle of osp site using machine learning - Google Patents

Abstract

The present invention detects a life cycle of an illegal site or an annihilation site while updating a newly generated candidate site, periodically checks whether or not the illegal site can be accessed, determines whether the UI pattern is changed or not, And detecting a site where the route or the UI pattern is changed at any time by tracking the change site where the route of the destruction site is changed by analyzing the pattern, and detecting the site where the route or the UI pattern is changed at any time.

Description

[TECHNICAL FIELD] The present invention relates to an OSP site lifecycle detection artificial intelligence system using machine learning,

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an OSP life cycle sensing artificial intelligence system using machine learning, and more particularly, to a technique for automatically detecting generation, maintenance, change, or extinction of an OSP site.

An online service provider (OSP) is divided into a provider and an operator. A provider is a person who transmits a selected work or a user's choice without modifying it, or provides a connection by designating a route for the user to transmit through the information communication network. The operator can access the information communication network to replicate and transmit the work Or provide or operate facilities for such facilities.

Internet services, e-mail, news, newspapers, music, movies, images, games, software, e-shopping, electronic finance, e-health and e-government services. For example, an OSP site can be a web hard site that provides a variety of entertainment content, such as music and movies.

OSP sites include legitimate sites, illegal sites, and destruction sites. A legitimate site is a site that distributes a variety of entertainment contents by paying royalties. Illegal sites are sites that distribute content without paying royalties. Destruction sites are sites that are not connected because the illegal sites have disappeared.

Recently, OSP site postings and contents were downloaded, and monitoring agents manually monitored illegal sites for copyright infringement.

Patent Document 1 relates to a copyright management method using a digital copyright illegal distribution site search system, collects sites by inputting copyright information search terms, compares the collected sites with a list of legitimate sites, and determines whether the sites are allegedly infringing .

However, in recent years, the OSP of the illegal site has a problem of changing the URL (Uniform Resource Locator) of the site or the UI (User Interface) pattern related to the screen displayed on the display from time to time to avoid coping with the copyright.

Korean Patent Laid-Open No. 10-2004-0076712 (September 23, 2004)

In order to solve the above problems, the present invention provides an OSP site lifecycle detection artificial intelligence system that utilizes machine learning to determine whether there is a copyright infringement site by analyzing the similarity between a UI pattern of a stored OSP site and a UI pattern of a candidate site do.

The present invention relates to an OSP site lifecycle detection artificial intelligence (UI) detection system that utilizes machine learning to periodically check whether or not an illegal site can be accessed, and to detect a life cycle of a site whose route or UI pattern has changed at any time by determining whether the UI pattern is changed. System.

The present invention provides an OSP site lifecycle detection artificial intelligence system that utilizes machine learning to track a change site whose path of a destruction site is changed by analyzing an identification pattern of a destruction site.

The OSP site lifecycle detection artificial intelligence system using the machine learning according to the present invention for solving the above problem includes lifecycle information, UI pattern information, and lifecycle information for an OSP site including a legitimate site, an illegal site, A storage unit for storing keyword information; A collection unit for collecting candidate sites newly generated using the keyword information; An inspection unit for checking whether or not the candidate site can be connected; An analysis classifying unit for classifying a candidate site by analyzing the similarity between the UI pattern of the OSP site and the UI pattern of the candidate site, and a controller for controlling information for the classified candidate site to be updated, , Change, or disappearance of the user.

The collecting unit may collect candidate sites related to the OSP site of the mobile web environment by using keyword information including keywords repeatedly or universally used and keywords substituted or derived from specific keywords.

The analysis classifier may perform a deep learning machine learning based on a neural network to compare features between the UI pattern of the OSP site and the UI pattern of the candidate site to detect the life cycle of the illegal site or the destruction site.

The checking unit checks whether or not the illegal site can be connected at a set period, and the analysis classifying unit discriminates whether the UI pattern of the illegal site is changed or not and reclassifies the illegal site where the UI pattern has changed.

And the analysis classifying unit classifies the unconnected illegal sites as annihilation sites.

The OSP site lifecycle detection artificial intelligence system utilizing machine learning further includes a tracking unit for tracking an alteration site in which the path of the destruction site is changed by analyzing an identification pattern including at least one of a path, a name, and a UI pattern of the destruction site can do.

The tracking unit may generate a candidate path by extracting a path source from a web log including a blog, a cafe, an SNS, and a bulletin board related to the identification pattern.

The analysis unit analyzes the similarity between the UI pattern of the destruction site and the UI pattern of the site connected by the candidate path. If the similarity degree value is not less than the reference value, the analysis unit classifies the changed site by the destruction site As shown in FIG.

The present invention analyzes similarity between a UI pattern of a stored OSP site and a UI pattern of a candidate site to determine whether the candidate site is a site suspected of copyright infringement, It is possible to provide an environment for efficiently monitoring an illegal site that requires copyright interception by increasing the accuracy of affinity discrimination.

The present invention can periodically check the availability of an illegal site, judge whether the UI pattern is changed or not, and enforce the copyrights so that quick response to a site whose route or UI pattern is changed at any time is possible.

According to the present invention, the identification pattern of the destruction site is analyzed, and the change site whose path of the destruction site is changed is tracked, so that the copyright enforcement can be performed so that quick response to the site whose path has changed from time to time is possible.

1 is a block diagram illustrating an OSP site lifecycle detection artificial intelligence system using machine learning of the present invention.
Fig. 2 shows the storage unit of Fig. 1 in detail.
3 illustrates an example of lifecycle information stored in the OSP base of FIG.
FIG. 4 is a block diagram showing the analysis classifying unit of FIG. 1 in detail.
FIG. 5 is a flowchart illustrating an operation method of an OSP site lifecycle sensing artificial intelligence system utilizing the machine learning of the present invention.
6 is a flowchart showing a pattern analysis and updating method of FIG.
7 is a flowchart illustrating an operation method of an OSP site lifecycle detection artificial intelligence system using machine learning according to another embodiment of the present invention.
8 is a flowchart illustrating an operation method of an OSP site lifecycle detection artificial intelligence system using machine learning according to another embodiment of the present invention.
FIG. 9 is a flowchart illustrating an operation method of an OSP site lifecycle detection artificial intelligence system using the machine learning of FIG. 8 in detail.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings and accompanying drawings, but the present invention is not limited to or limited by the embodiments.

FIG. 1 is a block diagram showing an OSP life cycle sensing artificial intelligence system using machine learning of the present invention. The OSP site life cycle sensing artificial intelligence system 100 using machine learning includes an online service provider (OSP: Online Service Provider) site is created, maintained, changed, or destroyed.

The OSP site lifecycle detection artificial intelligence system 100 using machine learning includes a storage unit 110, a collection unit 120, an inspection unit 130, an analysis classification unit 140, and a control unit 160.

The storage unit 110 stores lifecycle information, UI (UI) pattern information, and keyword information for an OSP site including a legitimate site, an illegal site, and a destruction site.

FIG. 2 is a detailed view of the storage unit of FIG. 1. FIG. 3 is a diagram illustrating life cycle information stored in the OSP base of FIG. 2. The storage unit 110 includes an OSP base 111, a UI base 112 And a keyword base 113.

The OSP base 111 may store lifecycle information in the form of a table as shown in FIG. The life cycle information is divided into a legitimate site, an illegal site, and a destruction site, and may include a name, a path, and a life cycle history for each classified OSP site.

The name may be the name of the OSP site associated with the keyword when searching by keywords on the portal search site, and the path may be an Internet address or a URL (Uniform Resource Locator).

The lifecycle history can indicate a state change such as creation, maintenance, change, or disappearance of each classified OSP site.

The UI base 112 may store UI pattern information of the OSP site. The UI pattern information may be information related to the web screen displayed on the display, and may be an HTML-based UI source.

The keyword base 113 may include keywords that are repeatedly or universally used in the name, path, or UI pattern at each OSP site, and keywords that are substituted or derived from specific keywords. For example, a keyword may include a web hard site, a movie download site, or a torrent as alternate or derived keywords if the web hard used as a repetitive or general purpose keyword is a keyword.

The collection unit 120 collects candidate sites using keyword information. The collection unit 120 may collect candidate sites related to the OSP site of the mobile web environment.

The checking unit 130 checks whether or not the candidate site can be connected. The checking unit 130 classifies the candidate site as a destruction site if the candidate site is not connected.

The analysis classifying unit 140 classifies the candidate sites by analyzing the similarity between the UI pattern of the OSP site and the UI pattern of the candidate site when the candidate site is accessible.

FIG. 4 is a detailed block diagram of the analysis classifying unit of FIG. 1. The analysis classifying unit 140 includes a first analyzing unit 141, a first classifying unit 142, a second analyzing unit 143, A classification section 144 and a third classification section 145. [

The first analysis unit 141 analyzes the UI pattern between the OSP site and the candidate site using the deep learning machine learning module and the second analysis unit 143 analyzes the identification pattern between the OSP site and the candidate site using the natural language processing module .

The UI pattern may be associated with a web screen that is displayed on the display, and may be associated with an HTML-based UI source.

The features of the UI pattern for machine learning are classified into an HTML tag and a value, and the HTML tag can include position information about a login form, an up / down / left / right frame, a button, and a banner bulletin board , The HTML value may include a click value to click on a category by genre or a click value to click on a post in each category. For example, categories on the web hard OSP site are divided into genres such as movies, dramas, videos, games, animations, utilities, music, books, education, home movies, and adults. In the drama postings, .

The first analyzer 141 may perform neural network-based deep learning machine learning to compare and analyze the features. Neural networks can be classified into two types: Deep Neural Network (CNN), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Restricted Boltzmann Machine (RBM) Deep Belief Network).

The first analysis unit 141 may be supervised learning that inputs data previously generated by the user and then outputs the data, and unsupervised learning that models the pattern only by input without output. In order to improve accuracy, Ensemble learning, Cortical learning, Incremental learning, Transfer learning or Reinforcement learning.

The first analysis unit 141 can infer whether the site is similar using logical reasoning, Bayesian inference or decision tree search.

The identification pattern includes at least one of a path, a name, and a UI pattern. For example, if the path of the destruction site is http://OOOwebhard1/ and the path of the candidate site is http://OOOwebhard2.com/, the second analysis unit 143 is identical except for the numbers between the two sites , It can be analyzed that the identification pattern between the two sites is similar.

The first classifying unit 142 first classifies the candidate sites in accordance with the UI pattern result, classifies the candidate sites in accordance with the identification pattern result of the second classifying unit 144, Classifies the candidate site into the third category in correspondence with the connection of the illegal site associated with the candidate site. 5 to 8, an operation method of the OSP site lifecycle detection artificial intelligence system 100 using machine learning will be described in detail.

FIG. 5 is a flowchart illustrating an operation method of an OSP site lifecycle sensing artificial intelligence system utilizing the machine learning of the present invention. The storage unit 110 stores a lifetime of an OSP site including a legitimate site, an illegal site, The UI pattern information, and the keyword information, and the collection unit 120 collects the candidate sites using the keyword information.

The checking unit 130 checks whether or not the candidate site can be connected. The checking unit 130 classifies the candidate site as a destruction site if the candidate site is not connected.

When the candidate site is accessible, the analysis classifying unit 140 analyzes the similarity between the UI pattern of the OSP site and the UI pattern of the candidate site, and classifies and updates the candidate site.

FIG. 6 is a flowchart illustrating a method of analyzing and updating the pattern of FIG. 5, wherein the OSP site lifecycle detection artificial intelligence system 100 using machine learning detects a life cycle of an illegal site or an annihilation site while classifying a candidate site can do.

If the candidate site can be accessed, the first analyzer 141 analyzes the similarity between the UI pattern of the OSP site and the UI pattern of the candidate site.

If the UI pattern of the candidate site is similar to the UI pattern of the illegal site or the destruction site, the first classification unit 142 temporarily classifies the candidate site as an illegal site, and classifies the candidate site as a legitimate site if not.

If the UI pattern of the candidate site is similar to the UI pattern of the illegal site, the second analysis unit 143 analyzes the identification pattern of the illegal site and the identification pattern of the candidate site. If the UI pattern of the candidate site is similar to the UI pattern of the destruction site And analyzes the identification pattern of the destruction site and the identification pattern of the candidate site.

When the identification pattern of the candidate site is similar to the identification pattern of the illegal site, the second classification unit 144 requests the checking unit 130 to check the connection of the illegal site to check whether the candidate site is changed by the illegal site, If not, the candidate site is classified as a new illegal site.

If the identification pattern of the candidate site is similar to the identification pattern of the destruction site, the second classification unit 144 classifies the candidate site as an illegal site changed by the destruction site, and classifies the candidate site as a new illegal site if not.

The third classifying unit 145 classifies the candidate site as a new illegal site when it is possible to connect an illegal site similar to the identification pattern of the candidate site. If not, we classify the illegal sites as destruction sites, and classify the candidate sites as illegal sites changed by illegal sites.

The control unit 400 controls the information about the classified candidate sites to be updated.

FIG. 7 is a flowchart illustrating an operation method of an OSP site lifecycle detection artificial intelligence system using machine learning according to another embodiment of the present invention. The OSP site life cycle detection artificial intelligence system 100 using machine learning It is possible to periodically check whether or not the illegal site can be accessed, judge whether the UI pattern is changed or not, and enforce the coping of the copyright on the site whose route or UI pattern is changed at any time.

The checking unit 130 checks whether or not an illegal site can be accessed at a set period, and classifies the illegal site as a destruction site when the illegal site can not be accessed.

The first analysis unit 141 analyzes the similarity between the UI pattern of the illegally-stored site and the UI pattern of the illegal site that is currently connected when the illegal site is accessible.

The first classification unit 142 terminates if there is no UI pattern variation, and updates the UI pattern of the connected illegal site if there is a UI pattern variation.

FIG. 8 is a flowchart illustrating an operation method of an OSP site lifecycle sensing artificial intelligence system using machine learning according to another embodiment of the present invention. Referring to FIG. 8, an OSP site life cycle sensing artificial intelligence system 100 using machine learning, And a tracking unit 150 for tracking a change site whose path has been changed by analyzing the identification pattern of the destruction site, so that the copyright enforcement of the site whose path has been changed from time to time can be performed.

The tracking unit 150 analyzes the identification pattern including at least one of the path name, the name, and the UI pattern of the destruction site, and tracks the changed site whose path of the destruction site is changed.

The tracking unit 150 extracts a path source from a UI source of a web log including a blog, a cafe, an SNS, and a bulletin board related to the identification pattern to generate a candidate path.

The checking unit 130 checks whether or not the candidate path can be connected, and if the connection is possible, the first analyzing unit 141 analyzes the similarity between the UI pattern of the destruction site and the UI pattern of the candidate path site connected by the candidate path .

If the UI pattern of the destruction site is similar to the UI pattern of the candidate route site, the first classification unit 142 classifies the candidate path site as an illegal site changed by the destruction site, and classifies the candidate path site as a tracking failure site if not.

When the web hard site disappears, the collecting unit 120 first searches the portal with the corresponding site path or name and collects related web logs. The tracking unit 150 performs morphological analysis on the name and the path of the web hard and recognizes the entity name in the collected web log to extract key individuals and extract a candidate path. For example, if you have the phrase "file-nori-address has been changed to www.filenori2.co.kr" on your blog, extract the file-nori, address, www.filenori2.co.kr and change entities, A candidate path can be generated.

The tracking unit 150 may use various algorithms such as Euclidean Distance, Cosine Distance, Jaccard Distance, Correlation Distance, and Levenshtein distance when comparing the similarities of the site paths and names.

The first analyzing unit 141 analyzes the UI pattern using a deep learning algorithm when the first analyzing unit 141 can connect to the first analyzing unit 141. The first analyzing unit 141 analyzes the similarity If the value is above the reference value, it is classified as a site changed by the destruction site.

FIG. 9 is a flowchart illustrating an operation method of an OSP site lifecycle detection artificial intelligence system using the machine learning of FIG. 8 in detail.

The tracking unit 150 analyzes the identification pattern including at least one of the path name, the name, and the UI pattern of the destruction site, and tracks the changed site whose path of the destruction site is changed.

The tracking unit 150 extracts a path source from a UI source of a weblog including a blog, a cafe, an SNS, and a bulletin board related to the identification pattern, and generates one or more candidate paths.

The checking unit 130 checks whether or not the candidate path can be connected. The checking unit 130 can check whether or not all candidate paths can be connected. The first classification unit 142 classifies the candidate path site as a tracking failure site if all candidate paths can not be connected.

The analysis classifying unit 140 analyzes the similarity between the UI pattern of the destruction site and the UI pattern of the candidate path site connected by the candidate path when the connection is possible.

If the UI pattern of the destruction site is similar to the UI pattern of the candidate route site, the analysis classifying unit 140 classifies the candidate path site as the illegal site changed by the destruction site, and if not, the UI pattern of the OSP site and the candidate path site Analyze similarities between UI patterns and classify candidate path sites. The control unit 400 controls the information about the OSP site to be updated while controlling the information about the classified candidate path site to be updated.

In other words, it is highly likely that a candidate route site that is not similar to the destruction site is similar to another OSP site, so it is necessary to analyze the similarity between the UI pattern of the OSP site and the UI pattern of the candidate route site.

A method of analyzing and updating the UI pattern for the candidate route site will be described with reference to FIG.

100: OSP site life cycle detection system 110:
120: collecting unit 130: checking unit
140: Analysis Classification Unit 150: Tracking Unit
160:

Claims (8)

A storage unit for storing lifecycle information, UI pattern information, and keyword information for an OSP site including a legitimate site, an illegal site, and a destruction site;
A collection unit for collecting candidate sites newly generated using the keyword information;
An inspection unit for checking whether or not the candidate site can be connected;
Deep learning machine learning based on neural network is performed to classify candidate sites by analyzing the similarity between the UI pattern of the OSP site and the UI pattern of the candidate site, and to compare and analyze features between the UI pattern of the OSP site and the UI pattern of the candidate site An analysis classification section for detecting the life cycle of illegal sites or annihilation sites, and
And a control unit for controlling the information about the classified candidate sites to be updated,
Wherein the generation, maintenance, change, or disappearance of the OSP site is detected and tracked. The method according to claim 1,
Wherein the collecting unit collects candidate sites related to an OSP site of a mobile web environment by using keyword information including keywords repeatedly or universally used and keywords substituted or derived from specific keywords OSP site life cycle detection artificial intelligence system. delete The method according to claim 1,
The checking unit checks whether or not an illegal site can be accessed at a set period,
Wherein the analysis classifier determines whether the UI pattern of the illegal site is changed or not, and reclassifies the illegal site in which the UI pattern has been changed. 5. The method of claim 4,
Wherein the analysis classification unit reclassifies the unconnected illegal sites to the destruction sites, and the OSP site lifecycle detection artificial intelligence system utilizing the machine learning. The method according to claim 1,
Further comprising a tracking unit for analyzing an identification pattern including at least one of a path name, a name, and a UI pattern of the destruction site, and tracking the changed site whose path of the destruction site has been changed. Sensing artificial intelligence system. The method according to claim 6,
Wherein the tracking unit generates a candidate path by extracting a path source from a web log including a blog, a cafe, a SNS, and a bulletin board related to an identification pattern, and generates a candidate path. 8. The method of claim 7,
The checking unit checks whether or not the candidate path can be connected,
Wherein the analysis classifying unit classifies the similarity between the UI pattern of the destruction site and the UI pattern of the site connected by the candidate path and classifies the site as a site changed by the destruction site if the similarity value is not less than a reference value, Site life cycle detection artificial intelligence system.

Images