CN110442771A - A kind of method and device that the detection website based on deep learning is distorted - Google Patents
A kind of method and device that the detection website based on deep learning is distorted Download PDFInfo
- Publication number
- CN110442771A CN110442771A CN201910741015.6A CN201910741015A CN110442771A CN 110442771 A CN110442771 A CN 110442771A CN 201910741015 A CN201910741015 A CN 201910741015A CN 110442771 A CN110442771 A CN 110442771A
- Authority
- CN
- China
- Prior art keywords
- word
- text
- sensitive
- information
- picture
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/12—Applying verification of the received information
- H04L63/123—Applying verification of the received information received data contents, e.g. message integrity
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1441—Countermeasures against malicious traffic
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/02—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Security & Cryptography (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Computing Systems (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Storage Device Security (AREA)
- Image Analysis (AREA)
Abstract
The application is the method and device that the detection website about a kind of based on deep learning is distorted, and belongs to the communications field.The described method includes: crawling the content of pages in website to be detected, when the content of pages is picture, the picture is inputted into sensitive image detection model, the sensitive image detection model is for detecting whether the picture includes sensitive image, obtain the testing result of the sensitive image detection model output, when it includes sensitive image that the testing result, which is the picture, determine that the website to be detected is tampered.When the content of pages is text information, the text information is input to text detection model, whether the text detection model is for detecting in the text information including sensitive information, obtain the testing result of the text detection model output, when the testing result is that the text information includes sensitive information, determine that the website to be detected is tampered.The application, which can be improved, improves the precision whether website is tampered.
Description
Technical field
This application involves the communications field, in particular to method and dress that a kind of detection website based on deep learning is distorted
It sets.
Background technique
The website of website often has the contents such as a large amount of webpage, can access for user and browse these webpages.But at present
Have criminal that can distort the webpage in website, can be embedded in the station to pornographic, gambling, it is sudden and violent probably, politics etc. it is relevant
Sensitive information passes through site propagation sensitive information.It include several for example, with reference to website homepage shown in FIG. 1, in the website homepage
The link of the page, news, medicine common sense, hospital transparency, on-line consulting and life relief expansion funds etc. respectively in institute.Referring to
Fig. 2, but after the website is distorted by undesirable, the link of medicine common sense is revised as the rich chess and card of gold, for publicizing gambling.
It propagates, website can be detected in the station in order to prevent sensitive information from being embedded into, detect to be tampered
Website, to remind the administrator of website to handle in time.The method of detection website at present are as follows: crawl each webpage in website, obtain
With MD5 (Message-Digest Algorithm5, Message Digest 5 5) value of each webpage.The MD5 value of each webpage is distinguished
It is compared with the MD5 value of each webpage of history acquisition.The webpage that the MD5 value and history for comparing some webpage obtain
When MD5 value difference, determine that the content of the webpage is tampered, and remind the webpage to the person of running affairs of website.
Webpage in website may be dynamic web page, and the variation of the content of dynamic web page may not be because by criminal
Sensitive content has been distorted, but can also have been come out by above method error detection.In addition, criminal can individually do sensitive content
At a webpage and the webpage is put on website, due to the MD5 value of the history acquisition without the webpage, the above method can not be examined
Measure the webpage.So the precision that current scheme detection website is tampered is very low.
Summary of the invention
The embodiment of the present application provides a kind of method and device for detecting website, to improve the essence whether website is tampered
Degree.The technical solution is as follows:
On the one hand, method that the detection website based on deep learning distorts that this application provides a kind of, which comprises
Crawl the content of pages in website to be detected address information and the content of pages；
The type of the content of pages is determined according to the suffix of the address information；
When the type is picture, the picture is inputted into sensitive image detection model, the sensitive image detects mould
Type obtains the testing result of the sensitive image detection model output, In for detecting whether the picture includes sensitive image
The testing result is the picture when including sensitive image, determines that the website to be detected is tampered；
When the type is text information, the text information is input to text detection model, the text detection
Model obtains the detection knot of the text detection model output for whether detecting in the text information including sensitive information
Fruit determines that the website to be detected is tampered when the testing result is that the text information includes sensitive information.
Optionally, described that the content of pages is input to sensitive information detection model, obtain the sensitive information detection
The testing result of model output, comprising:
When the content of pages is picture, the text information in the picture is extracted；
The text information is input to text detection model, the text detection model is for detecting the text information
In whether include sensitive information, obtain the testing result of text detection model output.
It is optionally, described that the text information is input to text detection model, comprising:
There are when sensitive word in the word that the text information includes, obtained and the sensitivity from the text information
X adjacent word of word, x are the integer greater than 1；
The term vector of each word of x+1 word is obtained, the term vector of word is the semantic expressiveness of the word, described
X+1 word includes the sensitive word and the x word；
The term vector of each word is input to text by sequence of each word in the text information
Detection model.
Optionally, the x word include in the text information be located at the sensitive word before and with the sensitive word
Adjacent x/2 word and after the sensitive word and the x/2 word adjacent with the sensitive word.
It is optionally, described that the picture is inputted into sensitive image detection model, comprising:
The picture gray processing is obtained into grayscale image, the size conversion that the grayscale image is converted will turn into pre-set dimension
The grayscale image after changing is input to the sensitive image detection model.
On the other hand, this application provides a kind of device that the detection website based on deep learning distorts, described device packet
It includes:
Crawl module, for crawl the content of pages in website to be detected address information and the content of pages；
Determining module, for determining the type of the content of pages according to the suffix of the address information；
First obtains module, for when the type is picture, the picture to be inputted sensitive image detection model, institute
Sensitive image detection model is stated for detecting whether the picture includes sensitive image, it is defeated to obtain the sensitive image detection model
Testing result out determines that the website to be detected is tampered when it includes sensitive image that the testing result, which is the picture,；
Second obtains module, for when the type is text information, the text information to be input to text detection
Model, the text detection model obtain the text detection for whether detecting in the text information including sensitive information
The testing result of model output determines described to be detected when the testing result is that the text information includes sensitive information
Website is tampered.
Optionally, described device includes:
Extraction module, for extracting the text information in the picture when the content of pages is picture；By the text
This information input arrives text detection model, and whether the text detection model is for detecting in the text information including sensitive letter
Breath obtains the testing result of the text detection model output.
Optionally, described second module is obtained, is used for:
There are when sensitive word in the word that the text information includes, obtained and the sensitivity from the text information
X adjacent word of word, x are the integer greater than 1；
The term vector of each word of x+1 word is obtained, the term vector of word is the semantic expressiveness of the word, described
X+1 word includes the sensitive word and the x word；
The term vector of each word is input to text by sequence of each word in the text information
Detection model.
Optionally, the x word include in the text information be located at the sensitive word before and with the sensitive word
Adjacent x/2 word and after the sensitive word and the x/2 word adjacent with the sensitive word.
Optionally, described first module is obtained, is used for:
The picture gray processing is obtained into grayscale image, the size conversion that the grayscale image is converted will turn into pre-set dimension
The grayscale image after changing is input to the sensitive image detection model.
On the other hand, this application provides a kind of electronic equipment, including at least one processor and at least one processor,
For storing at least one instruction, at least one described instruction is added at least one processor by least one described processor
It carries and runs, to realize above-mentioned method.
On the other hand, this application provides a kind of computer readable storage mediums, described for storing at least one instruction
At least one instruction is loaded and is run by processor, to realize above-mentioned method.
Technical solution provided by the embodiments of the present application can include the following benefits:
By crawling the content of pages in website to be detected；Content of pages is input to sensitive information detection model, it is sensitive
Infomation detection model is used to obtain the infomation detection mould based on whether including sensitive information in content of pages detection content of pages
The testing result of type output；When the testing result is that content of pages includes sensitive information, determination is described to be tampered to website.By
It is used to be detected based on content of pages in sensitive information detection model, it in this way can be based on the meaning detection that content of pages is expressed
Sensitive information improves the precision of detection sensitive information, and then improves the precision whether detection website is tampered.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The application can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the application
Example, and together with specification it is used to explain the principle of the application.
Fig. 1 is a kind of existing page schematic diagram；
Fig. 2 is existing another page schematic diagram；
Fig. 3 is the method flow diagram that the detection website provided by the embodiments of the present application based on deep learning is distorted；
Fig. 4 is the flow chart provided by the embodiments of the present application for crawling content of pages；
Fig. 5 is the flow chart of detection picture provided by the embodiments of the present application；
Fig. 6 is the flow chart of the first deep learning of training network provided by the embodiments of the present application；
Fig. 7 is the flow chart of detection text information provided by the embodiments of the present application；
Fig. 8 is the flow chart of the second deep learning of training network provided by the embodiments of the present application；
Fig. 9 is the apparatus structure schematic diagram that the detection website provided by the embodiments of the present application based on deep learning is distorted；
Figure 10 is a kind of terminal structure schematic diagram provided by the embodiments of the present application.
Through the above attached drawings, it has been shown that the specific embodiment of the application will be hereinafter described in more detail.These attached drawings
It is not intended to limit the range of the application design in any manner with verbal description, but is by referring to specific embodiments
Those skilled in the art illustrate the concept of the application.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the application.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the application.
The content in website may be distorted by undesirable at present, make include in website with pornographic, gambling, it is sudden and violent probably, politics
Relevant sensitive information.
In order to detect the sensitive information in the website being tampered, sensitive information detection model is trained in this application,
Sensitive information detection model can detecte whether the content of pages in the website includes sensitive information.Content of pages master in website
It to include two kinds, one is pictures, and one is text informations.Picture can be the picture including picture material, including text envelope
The picture of breath and while the picture including text information and picture material.
Sensitive information includes two classes, and a kind of sensitive information is text information, for example, for publicize pornographic, gambling, it is sudden and violent probably or
The text information of extreme politics.Another kind of sensitive information is picture material, such as pornographic image, gambling image etc..It trains
Sensitive information detection model includes sensitive image detection model and text detection model.Sensitive image detection model is for detecting figure
Sensitive image in piece, whether text detection model is for detecting in text information including sensitive information.
Sensitive image detection model is the intelligent measurement model by being trained to the first deep learning network.Text
This detection model is the model of mind model by being trained to the second deep learning network.Training obtains sensitive image
It after detection model and text detection model, can detect whether website is distorted by undesirable by following any embodiment, lead
Causing website includes sensitive information.
Referring to Fig. 3, the embodiment of the present application provides a kind of method that the detection website based on deep learning is distorted, comprising:
Step 101: the address information and the content of pages of the content of pages of website to be detected are crawled, according to the address information
Suffix determine and determine the type of the content of pages surely.
Website to be detected includes website homepage, includes the address letter for belonging to each page of website to be detected in the website homepage
It ceases, may include the content of pages for belonging to website to be detected in each page.
For any one page, the content of pages of the page may include the ground of text information, picture or other pages
At least one of location information etc..When the content of pages is picture, the address information including the picture in the page.
It should be understood that the address information for the page for including for website homepage, the address information of this page of page is corresponding
The page may be the page for belonging to website to be detected, it is also possible to belong to the page of other websites.It optionally, may also in the page
Including other pages address information, other corresponding pages of the address information of other pages may be to belong to website to be detected
The page, it is also possible to belong to the page of other websites.
The address information of the website homepage of website to be detected is the domain-name information of website to be detected.Belong in website to be detected
The address information of any one page include the storage address and suffix of the domain-name information, the page in website to be detected.It is right
In the content of pages for belonging to the website to be detected, if the content of pages is picture, the address information of the picture includes to be detected
The storage address and suffix of the domain-name information of website, the picture in website to be detected, the suffix can be the picture of the picture
Type.Picture type can be jpg, jpeg, png, bmp or gif etc..
In this step, referring to fig. 4, the domain-name information for obtaining website to be detected is believed according to the domain name of the website to be detected
Breath crawls the content of pages that the website homepage of website to be detected includes.It is crawled in the website homepage to an address information, is sentenced
The domain-name information that the address information includes that breaks is identical as the domain information of website to be detected.
If it is different, then continuing to crawl the other content in website homepage.If identical, the suffix of the address information is judged
Whether be picture type, if it is picture type, crawl the corresponding picture of the address information, by the address information and the picture it
Between corresponding relationship be stored in picture list.If not picture type, the page in the corresponding page of the address information is crawled
Face content.
The content of pages of the page may include text information, the address information of picture or address information of other pages etc.
At least one of.When crawling text information in the page, by the corresponding pass between the address information and text information
System is saved in text list.Address information is being crawled, when the suffix of the address information is picture type, is indicating the address
Information is the address information of picture, crawls the picture based on the address information, is stored in the address information is corresponding with the picture
Picture list.When the suffix of the address information is not picture type, show that the address information is the address information of the page, judgement
Whether the domain-name information that the address information includes is the domain-name information of website to be detected, if so, climbing this based on the address information
The page.
For any one address information, after having crawled all the elements in the corresponding page of the address information, returning to should
The page where address information continues the content for crawling the page.Until crawling all the elements in website to be detected.
For any one page, if including address information in the page, which is located at preset address tag
Later.Preset address tag is href and src etc..
Step 102: when the type of the content of pages is picture, extracting the text information in the picture, the text is believed
The address information of breath and picture correspondence is stored in text list.
Picture can be the picture including picture material, the picture including text information and simultaneously include text information and figure
As the picture of content.
For the picture including picture material, then extract from the picture less than text information.For including text information
Picture or simultaneously include text information and picture material picture, then extract text information from the picture.
In this step, picture address information corresponding with the picture is read from picture list, can use OCR
(Optical Character Recognition, optical character identification) technology extracts text information from the picture, will extract
Text information corresponding with the address information be stored in text list.
In this step, which can be obtained into grayscale image, it can be using OCR technique from the grayscale image
Extract text information.
Step 103: the picture crawled being input to sensitive image detection model, sensitive image detection model is for detecting this
Whether picture includes sensitive image, obtains the testing result of sensitive image detection model output.
In this step, referring to Fig. 5, first picture saved in picture list can be read, first picture is defeated
Enter into sensitive image detection model, obtains the sensitive image detection model and export the corresponding testing result of first picture.It reads
Second picture for taking the picture list to save, second picture is input in sensitive image detection model, the sensitivity is obtained
Image detection model exports the corresponding testing result of second picture.It repeats the above process, until obtaining sensitive image detection
Model exports the corresponding testing result of the last one picture in the picture list.
Optionally, before the picture crawled is input to sensitive image detection model, the picture that can also will crawl
Gray processing is carried out, the grayscale image of the picture is obtained.It is pre-set dimension by the size conversion of the grayscale image, then by the ash after conversion
Degree figure is input to sensitive image detection model.Sensitive image detection model detects the grayscale image of input, detects whether
Including sensitive image.
Sensitive image detection model is trained to obtain to the first deep learning network in advance.It can be arranged in advance more
A image pattern and the corresponding markup information of each image pattern.It is sensitivity in the image pattern for any one image pattern
When image, the markup information of the image pattern indicates that the image pattern is sensitive image, which can be and pornographic, gambling
It is rich, sudden and violent probably or the relevant image such as politics.When the image pattern is not sensitive image, the markup information of the image pattern is indicated
The image pattern is not sensitive image.
When the image pattern is sensitive image, the markup information of the image pattern can be indicated with numerical value 1, in the image
When sample is not sensitive image, the markup information of the image pattern can be indicated with numerical value 0.Alternatively, being quick in the image pattern
When feeling image, the markup information of the image pattern can be indicated with numerical value 0, it, can be with when the image pattern is not sensitive image
The markup information of the image pattern is indicated with numerical value 1.
In this step, using the markup information of multiple image pattern and each image pattern the first deep learning of training
Network.Referring to Fig. 6, the training process is as follows:
1031: in the first deep learning network of training, by multiple image pattern and the corresponding mark of each image pattern
Information input is infused to the first deep learning network.
1032: the first deep learning networks detect whether each image pattern is sensitive image, obtain each image pattern
Testing result.
1033: the first deep learning networks are compared the testing result of each image pattern and markup information to obtain difference
Different information adjusts the network parameter of itself according to the different information.
By the different information that the testing result of each image pattern and markup information are compared can be one to
Amount, the element of the vector is the corresponding comparison result of each image pattern.The corresponding comparison result of image pattern can use numerical value 0
Or 1 indicate.It can indicate that the testing result of the image pattern is identical as markup information with numerical value 0, indicate the image with numerical value 1
The testing result of sample is different from markup information.Alternatively, can indicate that the testing result of the image pattern and mark are believed with numerical value 1
Manner of breathing is same, indicates that the testing result of the image pattern is different from markup information with numerical value 0.
The different information is input to preset cost function by 1034: the first deep learning networks, calculates cost value, In
When the cost value is not the minimum cost value of the preset cost function, 1032 are returned.
When the cost value is not the minimum cost value of the preset cost function, then the first deep learning network repeats
The above process detects whether each image pattern is sensitive image.
In the minimum cost value that the cost value is the preset cost function, then stop above-mentioned the first deep learning of training net
The operation of network, using the first deep learning network as sensitive image detection model.
Optionally, the first deep learning network can be that ResNet can be set ResNet's before training ResNet
Convolution number of layers is 32, filter blocks are 16 and the size of each filter is 3.The ResNet of setting is instructed in this way
Sensitive image detection model is got out, the precision of sensitive image detection model detection is higher.
Step 104: when it includes sensitive image that the testing result, which is the picture, obtaining the address letter of the picture crawled
Breath, determines that website to be detected is tampered, using the picture as content of evidence, by the address information of the picture and the content of evidence pair
It should be stored in the corresponding relationship of address information and content of evidence.
Each picture in 103 and 104 pairs of picture lists detects through the above steps, whereby it can be detected that including
Each picture of sensitive image.Since sensitive image detection model can be detected based on picture material in picture, so as to
Accurately detect whether picture includes sensitive image, can be improved in this way detection website in whether include sensitive image essence
Degree.
Step 105: when the type of the content of pages is text information, the text information crawled being input to text detection
Model, text detection model obtain the output of text detection model for detecting whether text information includes sensitive information
Testing result.
Referring to Fig. 7, in this step, first text information saved in text list can be read, by first text
It is corresponding to obtain text infomation detection model first text information of output into sensitive image detection model for this information input
Testing result.Second text information that text list saves is read, second text information is input to text detection mould
In type, obtains text detection model and export the corresponding testing result of second text information.It repeats the above process, until obtaining
Text detection model exports the corresponding testing result of the last one text information in text list.
In this step, before the text information crawled being input to text detection model, text information is divided
Word obtains each word that text information includes, and text information is obtained from the corresponding relationship of word and term vector and includes
The term vector of each word.The term vector of word is the semantic expressiveness of the word, then suitable in text information by each word
The term vector of each word is input to text detection model by sequence.Text detection model detects the text based on the term vector of each word
Whether information includes sensitive information.
The corresponding relationship of word and term vector can be downloaded from network and be obtained.
Sensitive information generally includes sensitive word, sensitive word be to pornographic, gambling, it is sudden and violent probably and politics etc. at least one is relevant
Word.Sensitive information is usually the text information or several text informations etc. for including the sensitive word.
The term vector of each word in text information is input to text detection model, in this way the detection of the increase text
The detection limit of model.
In order to solve this problem, when participle obtains each word that text information includes, judge that text information includes
Each word in the presence or absence of the sensitive word in sensitive dictionary.If there are sensitive word in each word that text information includes,
The x word adjacent with the sensitive word is obtained, x is the integer greater than 1.It is quick from this is obtained in the corresponding relationship of word and term vector
Feel the term vector of word and the term vector of each word in the x word.Further according to the sequence in text information to text
Detection model inputs the term vector of the sensitive word and the term vector of each word in the x word.Text detection model is based on
Whether the term vector detection text information of each word in the term vector of the sensitive word and the x word includes sensitive letter
Breath.It can reduce the word number of text detection module detection in this way, improve detection efficiency.
If there are sensitive words in the word that text information includes, available before the sensitive word and quick with this
Feel the adjacent x/2 word of word and positioned at the sensitive word later and the x/2 word adjacent with the sensitive word.
If the word number before being located at the sensitive word in text information is less than x/2, obtains and be located at the sensitivity
All words before word.If the word number after being located at the sensitive word in text information is less than x/2, position is obtained
All words after the sensitive word.
When text information includes sensitive word, show to may include sensitive information in text information, it is also possible to not include quick
Feel information.For example, drugs are sensitive words, and for text information " severe beat drugs ", although text information includes sensitive word,
But text information is not sensitive information, so the term vector of the adjacent word of the sensitive word is input to text detection model.
Whether the text information of context detection input of the text detection model based on sensitive word is sensitive information, to improve detection
Precision.
Text detection model is trained to obtain to the second deep learning network in advance.Multiple texts can be set in advance
This sample markup information corresponding with each samples of text.For any one samples of text, being in text sample includes sensitivity
When the sensitive information of word, the markup information of text sample indicates that samples of text is sensitive information, the sensitive information can be with
Pornographic, gambling, it is sudden and violent probably or the relevant text information such as politics.Text sample be include the non-sensitive information of sensitive word when,
The markup information of text sample indicates that samples of text is not sensitive information.
When text sample is sensitive information, the markup information of text sample can be indicated with numerical value 1, in the text
When sample is is not sensitive information, the markup information of text sample can be indicated with numerical value 0.Alternatively, being in text sample
When sensitive information, the markup information of text sample can be indicated with numerical value 0, it, can when text sample is not sensitive information
To indicate the markup information of text sample with numerical value 1.
In this step, using the markup information of multiple samples of text and each samples of text the second deep learning of training
Network.Referring to Fig. 8, the training process is as follows:
1051: in the second deep learning network of training, by multiple samples of text and the corresponding mark of each samples of text
Information input is infused to the second deep learning network.
1052: the second deep learning networks detect whether each samples of text is sensitive information, obtain each samples of text
Testing result.
1053: the second deep learning networks are compared the testing result of each samples of text and markup information to obtain difference
Different information.Second deep learning network adjusts the network parameter of itself according to the different information.
By the different information that the testing result of each samples of text and markup information are compared can be one to
Amount, the element of the vector is the corresponding comparison result of each samples of text.The corresponding comparison result of samples of text can with numerical value 0 or
1.It can indicate that the testing result of text sample is identical as markup information with numerical value 0, the inspection of text sample is indicated with numerical value 1
It is different from markup information to survey result.Alternatively, can indicate that the testing result of text sample is identical as markup information with numerical value 1,
Indicate that the testing result of text sample is different from markup information with numerical value 0.
The different information is input to preset cost function by 1054: the second deep learning networks, calculates cost value.In
When the cost value is not the minimum cost value of the preset cost function, returns and execute 1052.
When the cost value is not the minimum cost value of the preset cost function, the second deep learning network repeats above-mentioned
Process detects whether each samples of text is sensitive information.In the minimum cost value that the cost value is the preset cost function,
The operation for then stopping above-mentioned the second deep learning of training network, using the second deep learning network as text detection model.
Optionally, the second deep learning network can be LSTM, and before training LSTM, LSTM, which can be set, to be allowed to export
Word number be x+1, and setting allow input term vector dimension.Such as setting allows the dimension of the term vector of input
It is 100,200 or 300 etc..That is, the dimension of the term vector of each word is the dimension of setting.
Step 106: when it includes sensitive information that the testing result, which is text information, determine that website to be detected is tampered,
The address information of the page where obtaining text information, using text information as content of evidence, by the address information of the page
It is stored in the corresponding relationship of address information and content of evidence with content of evidence correspondence.
In this step, the sensitive word and the x word adjacent with the sensitive word in text detection model can be will enter into
Language is as content of evidence.
Each text information in 105 and 106 pairs of text lists detects through the above steps.Due to text detection mould
Type can be detected based on the semanteme of text information, so as to accurately detect whether text information includes sensitive letter
Breath, can be improved in this way detection website in whether include sensitive information precision.
In the present embodiment, the picture including sensitive image is being detected, or when the text information including sensitive information, really
Fixed website to be detected is tampered.Address information pass corresponding with content of evidence can be sent to the corresponding terminal of administrator of website
System.In this way administrator can content in the corresponding relationship by checking the address information and content of evidence, and treat measuring station
Point is handled.
In the embodiment of the present application, the address information of the content of pages and the content of pages in website to be detected is crawled；In
The suffix of the address information is picture type, indicates that the content of pages is picture, otherwise, indicates that the content of pages is text envelope
Breath.Picture is input to sensitive image detection model, sensitive image detection model is used to hold detection figure based on the image in picture
Whether include sensitive image in piece, obtains the testing result of sensitive image detection model output；It is comprising quick in the testing result
When feeling image, determination is tampered to website.Since sensitive image detection model is used to be examined based on the picture material in picture
It surveys, improves the precision of detection sensitive image in this way, and then improve the precision whether detection website is tampered.And by text envelope
Breath is input to text detection model, text detection model in the Semantic detection text information based on text information whether include
Sensitive information obtains the testing result of text detection model output；When the testing result is comprising sensitive information, determine wait stand
Point is tampered.Since text detection model is used to be detected based on the content in text information, the sensitive letter of detection is improved in this way
The precision of breath, and then improve the precision whether detection website is tampered.
Following is the application Installation practice, can be used for executing the application embodiment of the method.It is real for the application device
Undisclosed details in example is applied, the application embodiment of the method is please referred to.
Referring to Fig. 9, this application provides a kind of device 200 that the detection website based on deep learning distorts, described device
200 include:
Crawl module 201, for crawl the content of pages in website to be detected address information and the content of pages；
Determining module 202, for determining the type of the content of pages according to the suffix of the address information；
First obtains module 203, for when the type is picture, picture input sensitive image to be detected mould
Type, the sensitive image detection model obtain the sensitive image detection for detecting whether the picture includes sensitive image
The testing result of model output determines the website to be detected when it includes sensitive image that the testing result, which is the picture,
It is tampered；
Second obtains module 204, for when the type is text information, the text information to be input to text inspection
Model is surveyed, the text detection model obtains the text inspection for whether detecting in the text information including sensitive information
The testing result for surveying model output determines described to be checked when the testing result is that the text information includes sensitive information
Survey station point is tampered.
Optionally, described device 200 further include:
Extraction module, for extracting the text information in the picture when the type is picture；By the text envelope
Breath is input to text detection model, and whether it includes sensitive information that the text detection model is used to detect in the text information,
Obtain the testing result of the text detection model output.
Optionally, described second module 204 is obtained, is used for:
There are when sensitive word in the word that the text information includes, obtained and the sensitivity from the text information
X adjacent word of word, x are the integer greater than 1；
The term vector of each word of x+1 word is obtained, the term vector of word is the semantic expressiveness of the word, described
X+1 word includes the sensitive word and the x word；
The term vector of each word is input to text by sequence of each word in the text information
Detection model.
Optionally, the x word include in the text information be located at the sensitive word before and with the sensitive word
Adjacent x/2 word and after the sensitive word and the x/2 word adjacent with the sensitive word.
Optionally, described first module 203 is obtained, is used for:
The picture gray processing is obtained into grayscale image, the size conversion that the grayscale image is converted will turn into pre-set dimension
The grayscale image after changing is input to the sensitive image detection model.
In the embodiment of the present application, by crawling the content of pages in website to be detected；Content of pages is input to sensitivity
Infomation detection model, whether it includes sensitive information that sensitive information detection model is used to detect based on content of pages in content of pages,
Obtain the testing result of the infomation detection model output；When the testing result is that content of pages includes sensitive information, determine
It is described to be tampered to website.Since sensitive information detection model is used to be detected based on content of pages, it can be based on page in this way
The meaning of face content expression detects sensitive information, improves the precision of detection sensitive information, and then improves whether detection website is usurped
The precision changed.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
Figure 10 shows the structural block diagram of the terminal 300 of an illustrative embodiment of the invention offer.The terminal 300 is used for
The method for executing above-mentioned detection website, can be smart phone, tablet computer, laptop or desktop computer.Terminal 300 is also
Other titles such as user equipment, portable terminal, laptop terminal, terminal console may be referred to as.
In general, terminal 300 includes: processor 301 and memory 302.
Processor 301 may include one or more processing cores, such as 4 core processors, 8 core processors etc..Place
Reason device 301 can use DSP (Digital Signal Processing, Digital Signal Processing), FPGA (Field-
Programmable Gate Array, field programmable gate array), PLA (Programmable Logic Array, may be programmed
Logic array) at least one of example, in hardware realize.Processor 301 also may include primary processor and coprocessor, master
Processor is the processor for being handled data in the awake state, also referred to as CPU (Central Processing
Unit, central processing unit)；Coprocessor is the low power processor for being handled data in the standby state.In
In some embodiments, processor 301 can be integrated with GPU (Graphics Processing Unit, image processor),
GPU is used to be responsible for the rendering and drafting of content to be shown needed for display screen.In some embodiments, processor 301 can also be wrapped
AI (Artificial Intelligence, artificial intelligence) processor is included, the AI processor is for handling related machine learning
Calculating operation.
Memory 302 may include one or more computer readable storage mediums, which can
To be non-transient.Memory 302 may also include high-speed random access memory and nonvolatile memory, such as one
Or multiple disk storage equipments, flash memory device.In some embodiments, the non-transient computer in memory 302 can
Storage medium is read for storing at least one instruction, at least one instruction performed by processor 301 for realizing this Shen
Please in embodiment of the method provide detection website method.
In some embodiments, terminal 300 is also optional includes: peripheral device interface 303 and at least one peripheral equipment.
It can be connected by bus or signal wire between processor 301, memory 302 and peripheral device interface 303.Each peripheral equipment
It can be connected by bus, signal wire or circuit board with peripheral device interface 303.Specifically, peripheral equipment includes: radio circuit
304, at least one of touch display screen 305, camera 306, voicefrequency circuit 307, positioning component 308 and power supply 309.
Peripheral device interface 303 can be used for I/O (Input/Output, input/output) is relevant outside at least one
Peripheral equipment is connected to processor 301 and memory 302.In some embodiments, processor 301, memory 302 and peripheral equipment
Interface 303 is integrated on same chip or circuit board；In some other embodiments, processor 301, memory 302 and outer
Any one or two in peripheral equipment interface 303 can realize on individual chip or circuit board, the present embodiment to this not
It is limited.
Radio circuit 304 is for receiving and emitting RF (Radio Frequency, radio frequency) signal, also referred to as electromagnetic signal.It penetrates
Frequency circuit 304 is communicated by electromagnetic signal with communication network and other communication equipments.Radio circuit 304 turns electric signal
It is changed to electromagnetic signal to be sent, alternatively, the electromagnetic signal received is converted to electric signal.Optionally, radio circuit 304 wraps
It includes: antenna system, RF transceiver, one or more amplifiers, tuner, oscillator, digital signal processor, codec chip
Group, user identity module card etc..Radio circuit 304 can be carried out by least one wireless communication protocol with other terminals
Communication.The wireless communication protocol includes but is not limited to: WWW, Metropolitan Area Network (MAN), Intranet, each third generation mobile communication network (2G, 3G,
4G and 5G), WLAN and/or WiFi (Wireless Fidelity, Wireless Fidelity) network.In some embodiments, it penetrates
Frequency circuit 304 can also include NFC (Near Field Communication, wireless near field communication) related circuit, this
Application is not limited this.
Display screen 305 is for showing UI (User Interface, user interface).The UI may include figure, text, figure
Mark, video and its their any combination.When display screen 305 is touch display screen, display screen 305 also there is acquisition to show
The ability of the touch signal on the surface or surface of screen 305.The touch signal can be used as control signal and be input to processor
301 are handled.At this point, display screen 305 can be also used for providing virtual push button and/or dummy keyboard, also referred to as soft button and/or
Soft keyboard.In some embodiments, display screen 305 can be one, and the front panel of terminal 300 is arranged；In other embodiments
In, display screen 305 can be at least two, be separately positioned on the different surfaces of terminal 300 or in foldover design；In still other reality
It applies in example, display screen 305 can be flexible display screen, be arranged on the curved surface of terminal 300 or on fold plane.Even, it shows
Display screen 305 can also be arranged to non-rectangle irregular figure, namely abnormity screen.Display screen 305 can use LCD (Liquid
Crystal Display, liquid crystal display), OLED (Organic Light-Emitting Diode, Organic Light Emitting Diode)
Etc. materials preparation.
CCD camera assembly 306 is for acquiring image or video.Optionally, CCD camera assembly 306 include front camera and
Rear camera.In general, the front panel of terminal is arranged in front camera, the back side of terminal is arranged in rear camera.One
In a little embodiments, rear camera at least two is main camera, depth of field camera, wide-angle camera, focal length camera shooting respectively
Any one in head, to realize that main camera and the fusion of depth of field camera realize background blurring function, main camera and wide-angle
Camera fusion realizes that pan-shot and VR (Virtual Reality, virtual reality) shooting function or other fusions are clapped
Camera shooting function.In some embodiments, CCD camera assembly 306 can also include flash lamp.Flash lamp can be monochromatic warm flash lamp,
It is also possible to double-colored temperature flash lamp.Double-colored temperature flash lamp refers to the combination of warm light flash lamp and cold light flash lamp, can be used for not
With the light compensation under colour temperature.
Voicefrequency circuit 307 may include microphone and loudspeaker.Microphone is used to acquire the sound wave of user and environment, and will
Sound wave, which is converted to electric signal and is input to processor 301, to be handled, or is input to radio circuit 304 to realize voice communication.
For stereo acquisition or the purpose of noise reduction, microphone can be separately positioned on the different parts of terminal 300 to be multiple.Mike
Wind can also be array microphone or omnidirectional's acquisition type microphone.Loudspeaker is then used to that processor 301 or radio circuit will to be come from
304 electric signal is converted to sound wave.Loudspeaker can be traditional wafer speaker, be also possible to piezoelectric ceramic loudspeaker.When
When loudspeaker is piezoelectric ceramic loudspeaker, the audible sound wave of the mankind can be not only converted electrical signals to, it can also be by telecommunications
Number the sound wave that the mankind do not hear is converted to carry out the purposes such as ranging.In some embodiments, voicefrequency circuit 307 can also include
Earphone jack.
Positioning component 308 is used for the current geographic position of positioning terminal 300, to realize navigation or LBS (Location
Based Service, location based service).Positioning component 308 can be the GPS (Global based on the U.S.
Positioning System, global positioning system), China dipper system or Russia Galileo system positioning group
Part.
Power supply 309 is used to be powered for the various components in terminal 300.Power supply 309 can be alternating current, direct current,
Disposable battery or rechargeable battery.When power supply 309 includes rechargeable battery, which can be wired charging electricity
Pond or wireless charging battery.Wired charging battery is the battery to be charged by Wireline, and wireless charging battery is by wireless
The battery of coil charges.The rechargeable battery can be also used for supporting fast charge technology.
In some embodiments, terminal 300 further includes having one or more sensors 310.The one or more sensors
310 include but is not limited to: acceleration transducer 311, gyro sensor 312, pressure sensor 313, fingerprint sensor 314,
Optical sensor 315 and proximity sensor 316.
The acceleration that acceleration transducer 311 can detecte in three reference axis of the coordinate system established with terminal 300 is big
It is small.For example, acceleration transducer 311 can be used for detecting component of the acceleration of gravity in three reference axis.Processor 301 can
With the acceleration of gravity signal acquired according to acceleration transducer 311, touch display screen 305 is controlled with transverse views or longitudinal view
Figure carries out the display of user interface.Acceleration transducer 311 can be also used for the acquisition of game or the exercise data of user.
Gyro sensor 312 can detecte body direction and the rotational angle of terminal 300, and gyro sensor 312 can
To cooperate with acquisition user to act the 3D of terminal 300 with acceleration transducer 311.Processor 301 is according to gyro sensor 312
Following function may be implemented in the data of acquisition: when action induction (for example changing UI according to the tilt operation of user), shooting
Image stabilization, game control and inertial navigation.
The lower layer of side frame and/or touch display screen 305 in terminal 300 can be set in pressure sensor 313.Work as pressure
When the side frame of terminal 300 is arranged in sensor 313, user can detecte to the gripping signal of terminal 300, by processor 301
Right-hand man's identification or prompt operation are carried out according to the gripping signal that pressure sensor 313 acquires.When the setting of pressure sensor 313 exists
When the lower layer of touch display screen 305, the pressure operation of touch display screen 305 is realized to UI circle according to user by processor 301
Operability control on face is controlled.Operability control includes button control, scroll bar control, icon control, menu
At least one of control.
Fingerprint sensor 314 is used to acquire the fingerprint of user, collected according to fingerprint sensor 314 by processor 301
The identity of fingerprint recognition user, alternatively, by fingerprint sensor 314 according to the identity of collected fingerprint recognition user.It is identifying
When the identity of user is trusted identity out, the user is authorized to execute relevant sensitive operation, the sensitive operation packet by processor 301
Include solution lock screen, check encryption information, downloading software, payment and change setting etc..Terminal can be set in fingerprint sensor 314
300 front, the back side or side.When being provided with physical button or manufacturer Logo in terminal 300, fingerprint sensor 314 can be with
It is integrated with physical button or manufacturer Logo.
Optical sensor 315 is for acquiring ambient light intensity.In one embodiment, processor 301 can be according to optics
The ambient light intensity that sensor 315 acquires controls the display brightness of touch display screen 305.Specifically, when ambient light intensity is higher
When, the display brightness of touch display screen 305 is turned up；When ambient light intensity is lower, the display for turning down touch display screen 305 is bright
Degree.In another embodiment, the ambient light intensity that processor 301 can also be acquired according to optical sensor 315, dynamic adjust
The acquisition parameters of CCD camera assembly 306.
Proximity sensor 316, also referred to as range sensor are generally arranged at the front panel of terminal 300.Proximity sensor 316
For acquiring the distance between the front of user Yu terminal 300.In one embodiment, when proximity sensor 316 detects use
When family and the distance between the front of terminal 300 gradually become smaller, touch display screen 305 is controlled from bright screen state by processor 301
It is switched to breath screen state；When proximity sensor 316 detects user and the distance between the front of terminal 300 becomes larger,
Touch display screen 305 is controlled by processor 301 and is switched to bright screen state from breath screen state.
It will be understood by those skilled in the art that the restriction of the not structure paired terminal 300 of structure shown in Figure 10, can wrap
It includes than illustrating more or fewer components, perhaps combine certain components or is arranged using different components.
Those skilled in the art will readily occur to its of the application after considering specification and practicing application disclosed herein
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the application, these modifications, purposes or
Person's adaptive change follows the general principle of the application and including the undocumented common knowledge in the art of the application
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the application are by following
Claim is pointed out.
It should be understood that the application is not limited to the precise structure that has been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.Scope of the present application is only limited by the accompanying claims.
Claims (10)
1. a kind of method that the detection website based on deep learning is distorted, which is characterized in that the described method includes:
Crawl the content of pages in website to be detected address information and the content of pages；
The type of the content of pages is determined according to the suffix of the address information；
When the type is picture, the picture is inputted into sensitive image detection model, the sensitive image detection model is used
Whether include sensitive image in detecting the picture, the testing result of the sensitive image detection model output is obtained, described
Testing result is the picture when including sensitive image, determines that the website to be detected is tampered；
When the type is text information, the text information is input to text detection model, the text detection model
For whether detecting in the text information including sensitive information, the testing result of the text detection model output, In are obtained
When the testing result is that the text information includes sensitive information, determine that the website to be detected is tampered.
2. the method as described in claim 1, which is characterized in that the method also includes:
When the type is picture, the text information in the picture is extracted；The text information is input to text detection
Model, the text detection model obtain the text detection for whether detecting in the text information including sensitive information
The testing result of model output.
3. method according to claim 1 or 2, which is characterized in that described that the text information is input to text detection mould
Type, comprising:
There are when sensitive word in the word that the text information includes, obtained and the sensitive word phase from the text information
X adjacent word, x are the integer greater than 1；
The term vector of each word of x+1 word is obtained, the term vector of word is the semantic expressiveness of the word, the x+1
A word includes the sensitive word and the x word；
The term vector of each word is input to text detection by sequence of each word in the text information
Model.
4. method as claimed in claim 3, which is characterized in that the x word includes in the text information positioned at described
Before sensitive word and the x/2 word adjacent with the sensitive word and to be located at the sensitive word later and adjacent with the sensitive word
X/2 word.
5. the method as described in claim 1, which is characterized in that described that the picture is inputted sensitive image detection model, packet
It includes:
The picture gray processing is obtained into grayscale image, the size conversion that the grayscale image is converted is pre-set dimension, after conversion
The grayscale image be input to the sensitive image detection model.
6. a kind of device that the detection website based on deep learning is distorted, which is characterized in that described device includes:
Crawl module, for crawl the content of pages in website to be detected address information and the content of pages；
Determining module, for determining the type of the content of pages according to the suffix of the address information；
First obtains module, described quick for when the type is picture, the picture to be inputted sensitive image detection model
Sense image detection model obtains the sensitive image detection model output for detecting whether the picture includes sensitive image
Testing result determines that the website to be detected is tampered when it includes sensitive image that the testing result, which is the picture,；
Second obtains module, for when the type is text information, the text information to be input to text detection model,
It is defeated to obtain the text detection model for whether detecting in the text information including sensitive information for the text detection model
Testing result out determines the website quilt to be detected when the testing result is that the text information includes sensitive information
It distorts.
7. device as claimed in claim 6, which is characterized in that described device further include:
Extraction module, for extracting the text information in the picture when the type is picture；The text information is defeated
Enter to text detection model, whether the text detection model is for detecting in the text information including sensitive information, acquisition
The testing result of the text detection model output.
8. device as claimed in claims 6 or 7, which is characterized in that described second obtains module, is used for:
There are when sensitive word in the word that the text information includes, obtained and the sensitive word phase from the text information
X adjacent word, x are the integer greater than 1；
The term vector of each word of x+1 word is obtained, the term vector of word is the semantic expressiveness of the word, the x+1
A word includes the sensitive word and the x word；
The term vector of each word is input to text detection by sequence of each word in the text information
Model.
9. a kind of electronic equipment, which is characterized in that including at least one processor and at least one processor, it is described at least one
For storing at least one instruction, at least one described instruction is loaded and is run by least one described processor memory, with
Realize such as method as claimed in any one of claims 1 to 6.
10. a kind of computer readable storage medium, which is characterized in that for storing at least one instruction, it is described at least one refer to
Order is loaded and is run by processor, to realize such as method as claimed in any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910741015.6A CN110442771B (en) | 2019-08-12 | 2019-08-12 | Deep learning-based method and device for detecting station tampering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910741015.6A CN110442771B (en) | 2019-08-12 | 2019-08-12 | Deep learning-based method and device for detecting station tampering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110442771A true CN110442771A (en) | 2019-11-12 |
| CN110442771B CN110442771B (en) | 2023-09-29 |
Family
ID=68434708
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910741015.6A Active CN110442771B (en) | 2019-08-12 | 2019-08-12 | Deep learning-based method and device for detecting station tampering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110442771B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111198983A (en) * | 2019-12-30 | 2020-05-26 | 中国银行股份有限公司 | Sensitive information detection method, device and storage medium |
| CN111209613A (en) * | 2020-01-10 | 2020-05-29 | 杭州涂鸦信息技术有限公司 | Rapid design method and system of intelligent product |
| CN113177409A (en) * | 2021-05-06 | 2021-07-27 | 上海慧洲信息技术有限公司 | Intelligent sensitive word recognition system |
| CN113221032A (en) * | 2021-04-08 | 2021-08-06 | 北京智奇数美科技有限公司 | Link risk detection method, device and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106411578A (en) * | 2016-09-12 | 2017-02-15 | 国网山东省电力公司电力科学研究院 | Website monitoring system and method applicable to power industry |
| CN106407260A (en) * | 2016-08-24 | 2017-02-15 | 乐视控股（北京）有限公司 | Processing method and apparatus for obtaining file type |
| CN107547555A (en) * | 2017-09-11 | 2018-01-05 | 北京匠数科技有限公司 | A kind of web portal security monitoring method and device |
-
2019
- 2019-08-12 CN CN201910741015.6A patent/CN110442771B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106407260A (en) * | 2016-08-24 | 2017-02-15 | 乐视控股（北京）有限公司 | Processing method and apparatus for obtaining file type |
| CN106411578A (en) * | 2016-09-12 | 2017-02-15 | 国网山东省电力公司电力科学研究院 | Website monitoring system and method applicable to power industry |
| CN107547555A (en) * | 2017-09-11 | 2018-01-05 | 北京匠数科技有限公司 | A kind of web portal security monitoring method and device |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111198983A (en) * | 2019-12-30 | 2020-05-26 | 中国银行股份有限公司 | Sensitive information detection method, device and storage medium |
| CN111209613A (en) * | 2020-01-10 | 2020-05-29 | 杭州涂鸦信息技术有限公司 | Rapid design method and system of intelligent product |
| CN111209613B (en) * | 2020-01-10 | 2023-05-12 | 杭州涂鸦信息技术有限公司 | Rapid design method and system for intelligent product |
| CN113221032A (en) * | 2021-04-08 | 2021-08-06 | 北京智奇数美科技有限公司 | Link risk detection method, device and storage medium |
| CN113177409A (en) * | 2021-05-06 | 2021-07-27 | 上海慧洲信息技术有限公司 | Intelligent sensitive word recognition system |
| CN113177409B (en) * | 2021-05-06 | 2024-05-31 | 上海慧洲信息技术有限公司 | Intelligent sensitive word recognition system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110442771B (en) | 2023-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110442771A (en) | A kind of method and device that the detection website based on deep learning is distorted | |
| CN109829456A (en) | Image-recognizing method, device and terminal | |
| CN110020140A (en) | Recommendation display methods, apparatus and system | |
| CN110413501B (en) | Browser compatibility testing method, device and system and readable storage medium | |
| CN109976930A (en) | Detection method, system and the storage medium of abnormal data | |
| EP3629198A1 (en) | Method and device for searching for audio | |
| CN109886208A (en) | Method, apparatus, computer equipment and the storage medium of object detection | |
| CN108304506A (en) | Search method, device and equipment | |
| CN110490179A (en) | Licence plate recognition method, device and storage medium | |
| CN110046018A (en) | Show method, apparatus, electronic equipment and the medium of message | |
| CN109992685A (en) | A kind of method and device of retrieving image | |
| CN109948087A (en) | Acquisition methods, device and the terminal of web page resources | |
| CN110414232A (en) | Rogue program method for early warning, device, computer equipment and storage medium | |
| CN108829488A (en) | Generate Web can interaction page method, apparatus and storage medium | |
| CN108320756A (en) | It is a kind of detection audio whether be absolute music audio method and apparatus | |
| CN110096525A (en) | Calibrate method, apparatus, equipment and the storage medium of interest point information | |
| CN110457571A (en) | Obtain method, apparatus, equipment and the storage medium of interest point information | |
| CN110334736A (en) | Image-recognizing method, device, electronic equipment and medium | |
| CN110288689A (en) | The method and apparatus that electronic map is rendered | |
| CN110059686A (en) | Character identifying method, device, equipment and readable storage medium storing program for executing | |
| CN110175653A (en) | Method, apparatus, equipment and the storage medium of image recognition | |
| CN109933594A (en) | Obtain method, apparatus, electronic equipment and the medium of data | |
| CN110109770A (en) | Adjustment method, device, electronic equipment and medium | |
| CN110348318A (en) | Image-recognizing method, device, electronic equipment and medium | |
| CN109871497A (en) | The method and apparatus of sharing information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |