A method and system for identifying fraudulent websites

By building a collision database and entry page feature judgment, bypassing the entrance page of the fraud website, extracting its page image features and comparing it with the official website data, solving the problem of low recognition accuracy of fraud websites in the existing technology, achieving a higher level of recognition effect.

CN114398581BActive Publication Date: 2024-11-05GUANGZHOU CRIMINAL SCIENCE & TECHNOLOGY RESEARCH INSTITUTE +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202210056971.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-18
Publication Date
2024-11-05
Estimated Expiration
2042-01-18

AI Technical Summary

Technical Problem

The existing fraud website identification methods have low accuracy through image matching, especially the fraudsters use the entrance page to evade detection, resulting in incorrect identification results.

Method used

Build a collision database, judge whether the website to be identified through the entrance page characteristics, and use the database data to bypass the entrance page to download the page pictures of the website to be identified, extract text and picture features and comparison databases to determine whether it is a fraud website.

Benefits of technology

It improves the recognition accuracy of fraudulent websites and can effectively identify fraudulent websites that use the entrance page to evade detection, improving the recognition effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114398581B_ABST
    Figure CN114398581B_ABST
Patent Text Reader

Abstract

The present invention proposes a method and system for identifying fraudulent websites, including: collecting web page data of fraudulent websites and official websites, building a collision database and a comparison database; judging whether there is an entry page for the website to be identified, and if there is an entry page, downloading a page image of the website to be identified using data from the collision database; extracting text and image features from the page image of the website to be identified, comparing the text and image features with the comparison database, and judging whether the website to be identified is a fraudulent website; the present invention takes into account the characteristics that fraudulent websites use entry pages to circumvent general fraudulent website detection, makes entry page judgment on the website to be identified, and builds a collision database, bypasses the entry page of the website to be identified and obtains the page image of the website to be identified, and uses the page image of the website to be identified to compare with the comparison data constructed based on the web page data of the official website, thereby improving the recognition effect of fraudulent websites.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of network security technology, and more specifically, to a method and system for identifying fraudulent websites. Background Art

[0002] With the development of information technology and the lightweight of electronic devices, people use terminal devices such as computers and mobile phones to complete asset transactions more and more frequently. At the same time, new types of cybercrime cases represented by telecommunications network fraud have increased sharply, among which fraud impersonating public security, procuratorial and judicial organs is particularly typical. Fraudsters contact victims through telephone and other means. Some fraudsters will ask victims to cooperate with the investigation and guide victims to fill in personal payment information on counterfeit websites and apps to commit fraud. Compared with other types of fraud (such as loan fraud, order fraud, etc.), fraud crimes impersonating public security, procuratorial and judicial organs are more secretive, involve larger amounts of money, and are more confusing. Therefore, the identification of public security, procuratorial and judicial fraud websites has become one of the important issues in preventing fraud crimes. Many existing website recognition methods use web page screenshots and web page element extraction methods, supplemented by machine learning algorithms for recognition.

[0003] There is a method for identifying phishing websites, which proposes to crawl web pages to generate web crawling results; render the web crawling results to form a web page to be identified; perform image matching on the web page to be identified based on the real website web page, and identify the phishing website based on the image matching results. The above method mainly performs image matching on the crawled web page to be identified and the real website web page, and identifies the phishing website based on the image matching results. However, today's fraudsters display less real official website content on the homepage of the fraud website by adding a new entry page, which makes the image matching accuracy low and the final recognition result prone to errors. Summary of the invention

[0004] The present invention provides a method and system for identifying fraudulent websites in order to improve the recognition effect of fraudulent websites.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention provides a method for identifying a fraudulent website, comprising the following steps:

[0007] S1: Collect web page data of known fraudulent websites, and use the web page data of the fraudulent websites to build a collision database;

[0008] S2: Collecting web page data of the official website, and using the web page data of the official website to build a comparison database;

[0009] S3: Based on the page features of the entry page, determine whether the website to be identified has an entry page. If the website to be identified has an entry page, log in to the website to be identified using the data in the collision database, and download the page image of the website to be identified using the data in the collision database; if the website to be identified does not have an entry page, directly enter the website to be identified and download the page image of the website to be identified;

[0010] S4: extracting text features and image features from the page image of the website to be identified, and comparing the text features and / or image features with a comparison database. When the comparison is successful, it is determined that the website to be identified is a fraudulent website.

[0011] Preferably, S1 specifically includes the following steps:

[0012] S1.1: Obtain a batch of known fraudulent websites and use a simulator on the terminal to simulate access to the fraudulent websites;

[0013] S1.2: Record the relative path and name of the homepage of the fraudulent website; if the fraudulent website has an entry page, record the entry page name and entry password of the entry page;

[0014] S1.3: Obtain the HTML source code of the fraud website homepage, and use regular expressions to obtain the relative path and image name of the image element in the HTML source code of the fraud website homepage;

[0015] S1.4: Construct a collision database using the relative path of the homepage, homepage name, entry page name and entry password of the fraudulent website, as well as the relative path and image name of the image elements in the HTML source code.

[0016] Preferably, in S3, the specific steps of determining whether there is an entry page of the website to be identified and downloading the page image of the website to be identified using the data in the collision database are as follows:

[0017] If it is determined that the website to be identified has an entry page, the name of the entry page of the website to be identified is recorded, and the name of the entry page of the fraudulent website is searched in the collision database;

[0018] If there is an entry page name of the fraudulent website that is the same as the entry page name of the website to be identified in the collision database, use the entry password corresponding to the entry page name of the fraudulent website to enter the homepage of the website to be identified, and use the relative path and image name of the image element in the HTML source code of the collision database to use a get request to download the page image of the website to be identified;

[0019] If there is no entry page name of the fraudulent website that is the same as the entry page name of the website to be identified in the collision database or the entry password corresponding to the entry page name of the fraudulent website is invalid, the relative path of the homepage of the fraudulent website in the collision database is used to directly access the homepage of the website to be identified, and the relative path and image name of the image element in the HTML source code in the collision database are used to use a get request to download the page image of the website to be identified.

[0020] Preferably, in S3, the step of determining whether there is an entry page for the website to be identified comprises the following steps:

[0021] Use a simulator on the terminal to simulate access to the website to be identified and obtain the HTML source code of the website to be identified;

[0022] Match the number of input boxes contained in the website to be identified according to the HTML source code of the official website , Number of paragraphs and number of hidden text , then the formula for determining the existence of the entry page is:

[0023]

[0024] in is the set threshold.

[0025] Preferably, S2 specifically includes the following steps:

[0026] S2.1: Obtain a batch of known official websites and use a simulator to simulate access to the official websites on the terminal;

[0027] S2.2: Obtain the HTML source code of the official website, obtain the page paragraphs of the official website according to the HTML source code of the official website, render and screenshot the page paragraphs, and obtain screenshots of each page paragraph of the official website;

[0028] S2.3: using regular expressions to obtain image elements in the HTML source code of the official website, downloading images of the official website according to the image elements, and performing feature extraction on the images of the official website to obtain image features of the official website;

[0029] S2.4: Match and extract Chinese characters in the HTML source code of the official website, perform word segmentation on the extracted Chinese characters to obtain corresponding words, and calculate the weight of each word;

[0030] S2.5: Using screenshots of each page paragraph of the official website, image features, words and word weights as data, a comparison database is constructed.

[0031] Preferably, in S4, Chinese characters are extracted from the page image of the website to be identified, and the extracted Chinese characters are segmented. Then, the words obtained after segmentation are queried in the comparison database for the corresponding word weights. According to the word weights, it is determined whether the website to be identified is a fraudulent website. The judgment formula is as follows:

[0032]

[0033] in, Indicates the top 5 words with the highest weights. Indicates i The word weight of the website to be identified, is a configurable threshold.

[0034] Preferably, in S4, the image features of the page images of the website to be identified are extracted, and the image features are searched based on cosine similarity with the image features of the official website in the comparison database. If the corresponding image feature elements with a cosine similarity higher than a preset cosine similarity threshold are retrieved, the website to be identified is judged to be a fraudulent website.

[0035] In a second aspect, the present invention provides a fraudulent website identification system, comprising:

[0036] A data collection module is used to collect web page data of known fraudulent websites and web page data of official websites;

[0037] Build a collision database to store web page data of fraudulent websites;

[0038] Comparison database, used to store web page data of official websites;

[0039] The entry page detection module is used to determine whether there is an entry page on the website to be identified;

[0040] The page image download module is used to download the page image of the website to be identified; if the website to be identified has an entry page, the website to be identified is logged in using the data in the collision database, and the page image of the website to be identified is downloaded using the data in the collision database; if the website to be identified does not have an entry page, the website to be identified is directly entered and the page image of the website to be identified is downloaded;

[0041] A feature extraction module is used to extract text features and image features from the page images of the website to be identified;

[0042] The identification module is used to compare the text features and image features in the page image of the website to be identified with the comparison database respectively to determine whether the website to be identified is a fraudulent website.

[0043] Preferably, the web page data of the fraudulent website includes the relative path of the home page of the fraudulent website, the name of the home page, the name of the entry page and the entry password, as well as the relative path and image name of the image element in the HTML source code.

[0044] Preferably, the web page data of the official website includes screenshots of each page paragraph of the official website, picture features, words and word weights.

[0045] Compared with the prior art, the beneficial effects of the technical solution of the present invention are: the present invention takes into account the characteristics that fraudulent websites use entry pages to circumvent general fraudulent website detection, and based on the page features and implementation scheme of the entry page, the entry page of the website to be identified is judged, and a collision database is constructed; if the website to be identified has an entry page, the data in the collision database is used to bypass the entry page of the website to be identified, log in to the website to be identified, use the data in the collision database to obtain the page image of the website to be identified, and use the image features and text features of the page image of the website to be identified to compare with the comparison data constructed based on the web page data of the official website, thereby improving the recognition effect of fraudulent websites. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Flowchart of the method for identifying fraudulent websites.

[0047] Figure 2 This is a flow chart for identifying fraudulent websites of public security, procuratorial and judicial organs in Example 2.

[0048] Figure 3 This is the HTML source code of a typical fraud website.

[0049] Figure 4 This is the entrance page of a typical fraud website.

[0050] Figure 5 This is a comparison chart of the portal page detection accuracy of the present invention and the recognition method based on double-end comparison and Chinese code screening.

[0051] Figure 6(a) and Figure 6(b) are comparison diagrams of the official website of the Supreme People's Procuratorate and the fraudulent website of the Supreme People's Procuratorate.

[0052] Figure 7 This is a comparison chart of the accuracy of fraud website identification based on the present invention and text recognition and image recognition.

[0053] Figure 8 This is the architecture diagram of the system for identifying fraudulent websites. DETAILED DESCRIPTION

[0054] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;

[0055] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0056] Example 1

[0057] See also Figure 1 This embodiment provides a method for identifying a fraudulent website, comprising the following steps:

[0058] S1: Collect web page data of known fraudulent websites, and use the web page data of the fraudulent websites to build a collision database;

[0059] S2: Collecting web page data of the official website, and using the web page data of the official website to build a comparison database;

[0060] S3: Based on the page features of the entry page, determine whether the website to be identified has an entry page. If the website to be identified has an entry page, log in to the website to be identified using the data in the collision database, and download the page image of the website to be identified using the data in the collision database; if the website to be identified does not have an entry page, directly enter the website to be identified and download the page image of the website to be identified;

[0061] S4: extracting text features and image features from the page image of the website to be identified, and comparing the text features and / or image features with a comparison database. When the comparison is successful, it is determined that the website to be identified is a fraudulent website.

[0062] The present invention takes into account the characteristic that fraudulent websites utilize entry pages to circumvent general fraudulent website detection, and based on the page features and implementation scheme of the entry page, performs entry page judgment on the website to be identified, and constructs a collision database; if the website to be identified has an entry page, the data in the collision database is used to bypass the entry page of the website to be identified, and the website to be identified is logged in, and the page image of the website to be identified is obtained using the data in the collision database, and the image features and text features of the page image of the website to be identified are compared with the comparison data constructed based on the web page data of the official website, thereby improving the recognition effect of fraudulent websites.

[0063] Example 2

[0064] This embodiment takes the public security, procuratorial and judicial fraud website as an example and proposes a method for identifying a fraud website. Figure 2 As shown, Figure 2 A flowchart of a method for identifying fraudulent websites of public security, procuratorial and judicial organs, including the following steps:

[0065] S1: Collect web page data of known fraudulent websites of public security, procuratorial and judicial organs, and use the web page data of the fraudulent websites to build a collision database. Specifically, the following steps are included:

[0066] S1.1: Obtain a batch of known fraudulent websites for public security, procuratorial and judicial organs, use the Android Studio platform to install a mobile phone simulator on the computer terminal, and use the mobile phone simulator to simulate mobile phone access to the known fraudulent websites for public security, procuratorial and judicial organs;

[0067] S1.2: Record the relative path and name of the homepage of the public security, procuratorial and judicial fraud website; if the public security, procuratorial and judicial fraud website has an entry page, record the entry page name and entry password of the entry page;

[0068] S1.3: Use regular expressions to obtain the relative path and image name of the image element in the HTML source code of the homepage of the public security, procuratorial and judicial fraud website; the structure of the regular expression is:

[0069]

[0070] S1.4: Build a collision database using the relative path and name of the homepage of the public security, procuratorial and judicial fraud website, as well as the relative path and image name of the image elements in the HTML source code.

[0071] S2: Collect web page data from official websites of public security bureaus, procuratorates and courts at the national, provincial and municipal levels, and use the web page data from the official websites to build a comparison database. This includes the following steps:

[0072] S2.1: Obtain a batch of public security, procuratorial and judicial official websites, and use a mobile phone simulator on a computer terminal to simulate mobile phone access to the official websites;

[0073] S2.2: Get the HTML source code of the official website of the public security, procuratorial and judicial organs, and use the HTML tag " "Obtain page paragraphs of the official website of the public security, procuratorial and judicial organs, render and screenshot the page paragraphs, and obtain screenshots of each page paragraph of the official website of the public security, procuratorial and judicial organs;

[0074] S2.3: Use regular expressions to obtain image elements in the HTML source code of the official website of the public security, procuratorial and judicial organs, download images of the official website, and use the SURF algorithm to extract features of the images of the official website to obtain image features of the official website; the structure of the regular expression for obtaining image elements in the HTML source code of the official website of the public security, procuratorial and judicial organs is:

[0075]

[0076] S2.4: Use UTF-8 encoding to match and extract Chinese characters in the HTML source code of the official website of the public security, procuratorial and judicial organs; the matching UTF-8 encoded Chinese characters range is:

[0077]

[0078] Use the THULAC toolkit to segment the extracted Chinese characters, obtain the corresponding words, and calculate the weight of each word; let the number of public security, procuratorial and judicial official websites obtained by the search be , and mark the word after the word The number of times it appears on the homepages of all public security, procuratorial and judicial official websites is , then the word The weight is recorded as:

[0079]

[0080] S2.5: Using screenshots of each page paragraph of the official website of the public security, procuratorial and judicial organs, image features, words and word weights as data, a comparison database is constructed.

[0081] S3: Based on the page features of the entry page, determine whether the website to be identified has an entry page. If the website to be identified has an entry page, use the data in the collision database to log in to the website to be identified, and use the data in the collision database to download the page image of the website to be identified; if the website to be identified does not have an entry page, directly enter the website to be identified and download the page image of the website to be identified. Most of the fraudulent websites of the public security, procuratorial and judicial organs are repeatedly deployed in multiple locations through mirroring and other methods. Fraudulent websites of the same category are likely to have common related entry names, relative paths of pictures, and picture names. Therefore, the page images of the website to be identified can be directly downloaded using the relative paths and picture names recorded in the collision database.

[0082] In this embodiment, the process of determining whether the website to be identified has an entry page includes the following steps:

[0083] Use a mobile phone simulator on a computer terminal to simulate access to the website to be identified and obtain the HTML source code of the website to be identified; Figure 3 As shown, Figure 3 This is the HTML source code of a typical fraud website.

[0084] According to the HTML source code of the official website, using the HTML tag " <form method="post">"Match the number of input boxes contained in the website to be identified ,use" " tag matches the number of paragraphs , use "display: none" to match hidden text data , then the formula for determining the existence of the entry page is:

[0085]

[0086] in is the set threshold, that is, when the matching website to be identified contains 1 input box and the number of paragraphs is less than or equal to or the number of hidden texts is greater than or equal to When , it is determined that the website to be identified has an entry page. Figure 4 and Figure 5 As shown, Figure 4 This is the entrance page of a typical fraud website. Figure 5 This is a comparison chart of the detection accuracy of the portal page of the present invention and the recognition method based on double-end comparison and Chinese code screening. Figure 5 It can be seen that, compared with the recognition method based on double-end comparison and Chinese code screening, the present invention has a higher detection rate in the detection of entry pages.

[0087] If it is determined that the website to be identified has an entry page, the name of the entry page is recorded, and the relevant entry name is searched in the collision database;

[0088] If there is a related entry name of the same category as the entry page name in the collision database, use the entry password corresponding to the related entry name to enter the homepage of the website to be identified, and use the relative path and image name of the image element in the HTML source code of the collision database to use a get request to download the page image of the website to be identified;

[0089] If there is no relevant entrance name identical to the entrance page name in the collision database or the entrance password corresponding to the relevant entrance name is invalid, the homepage of the website to be identified is directly accessed using the relative path of the homepage of the fraudulent website in the collision database, and the relative path and image name of the image element in the HTML source code in the collision database are used to download the page image of the website to be identified using a get request.

[0090] In this embodiment, the entry pages of 10 public security, procuratorial and judicial fraud websites are collided, and the collision results are shown in Table 1.

[0091] Table 1. Collision results of fraudulent websites

[0092]

[0093] As shown in Table 1, Table 1 is the collision result of fraudulent websites. It can be seen from the table that 80% of the entry pages can be logged in through the entry password, 30% of the fraudulent websites can download some pictures through relative paths, and 70% of the fraudulent websites can download a small number of pictures through relative paths.

[0094] S4: extracting text features and image features from the page image of the website to be identified, comparing the text features and image features with the comparison database, and determining whether the website to be identified is a fraudulent website. The following steps are included:

[0095] Extract Chinese characters from the page image of the website to be identified, segment the extracted Chinese characters, and then query the corresponding word weights of the words obtained after segmentation in the comparison database. According to the word weights, determine whether the website to be identified is a fraudulent website. The judgment formula is as follows:

[0096]

[0097] in, Indicates the top 5 words with the highest weights. is a configurable threshold. The sum of the weights of the current 5 words with the highest weights is greater than When the website to be identified is judged to be a fraud website.

[0098] The image features of the page images of the website to be identified are extracted, and the image features are searched based on cosine similarity with the image features of the official website in the comparison database. If the corresponding image feature elements with a cosine similarity higher than 0.9 are retrieved, the website to be identified is judged as a fraudulent website.

[0099] Public security, procuratorial and judicial fraud websites are different from general websites. In order to resist detection, most of them avoid direct identification of web page elements and content by means of browsing device restrictions, adding new entry web pages and inserting screenshots of official websites.

[0100] This embodiment takes into account the fact that fraudulent websites of public security, procuratorial and judicial organs use entry pages to circumvent the detection of general fraudulent websites of public security, procuratorial and judicial organs. Based on the page features and implementation scheme of the entry page, the entry page of the website to be identified is judged, and a collision database is constructed. The collision database is used to bypass the entry page of the website to be identified and obtain the page image of the website to be identified. The page image of the website to be identified is used to compare with the comparison data constructed based on the web page data of the official website, thereby improving the recognition effect of fraudulent websites of public security, procuratorial and judicial organs. As shown in Figure 6(a) and Figure 6(b), Figure 6(a) is the official website of the Supreme People's Procuratorate, and Figure 6(b) is the fraudulent website of the Supreme People's Procuratorate. This embodiment takes into account that fraudulent websites of public security, procuratorial and judicial organs often use paragraph rendering screenshots of the official website of public security, procuratorial and judicial organs to achieve a realistic counterfeiting effect. It proposes to count the Chinese characters of the official website, extract features from the paragraph screenshots of the official website, and construct a comparison database. Figure 7 As shown, Figure 7 This is a comparison chart of the accuracy of fraud website identification based on text recognition and image recognition. Figure 7 It can be seen that the present invention simultaneously considers image recognition and text recognition to identify fraudulent websites, achieving a higher recognition accuracy rate. This solution does not require the collection of a large number of positive and negative samples for training machine learning models, and can ensure the recognition accuracy of various new types of counterfeit public security, procuratorial and judicial websites.

[0101] Example 3

[0102] See also Figure 8 ,This embodiment proposes a fraud website identification system, including: a data acquisition module, a collision database, a comparison database, an entry page detection module, a page image downloading module, a feature extraction module and an identification module.

[0103] During the specific implementation process, the data collection module collects the relative path of the homepage, homepage name, entry page name and entry password of known fraudulent websites, as well as the relative path and image name of image elements in the HTML source code, and stores them in the collision database; collects screenshots, image features, words and word weights of each page paragraph of the official website, and uses the screenshots, image features, words and word weights of each page paragraph to store them in the comparison database.

[0104] The entry page detection module determines whether there is an entry page on the website to be identified based on the page features of the entry page, and outputs the determination result to the page image download module.

[0105] The page image downloading module downloads the page image of the website to be identified according to the judgment result of the entry page detection module; if the website to be identified has an entry page, the website to be identified is logged in using the data in the collision database, and the page image of the website to be identified is downloaded using the data in the collision database; if the website to be identified does not have an entry page, the website to be identified is directly entered and the page image of the website to be identified is downloaded.

[0106] The feature extraction module extracts Chinese characters from the page image of the website to be identified, and segment the extracted Chinese characters to obtain words. The feature extraction module extracts image features from the page image of the website to be identified, and outputs the image features and words to the recognition module.

[0107] The recognition module identifies the website to be identified based on word and image features, specifically:

[0108] The words obtained after word segmentation are queried for corresponding word weights in the comparison database, and the website to be identified is determined to be a fraudulent website based on the word weights.

[0109] The image features are compared with the image features of the official website in the database and searched based on cosine similarity. If a corresponding image feature element with a cosine similarity higher than a preset cosine similarity threshold is retrieved, the website to be identified is judged to be a fraudulent website.

[0110] The terms used in the drawings to describe positional relationships are only used for illustrative purposes and should not be construed as limiting this patent;

[0111] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention. < / form>

Claims

1. A method for identifying a fraudulent website, characterized in that: The following steps are involved: S1: Collecting web page data of known fraudulent websites and using the web page data of the fraudulent websites to build a collision database, including: S1.1: Obtain a batch of known fraudulent websites and use a simulator on the terminal to simulate access to the fraudulent websites; S1.2: Record the relative path and name of the homepage of the fraudulent website; if the fraudulent website has an entry page, record the entry page name and entry password of the entry page; S1.3: Obtain the HTML source code of the fraud website homepage, and use regular expressions to obtain the relative path and image name of the image element in the HTML source code of the fraud website homepage; S1.4: Build a collision database using the relative path, name, entry page name and entry password of the fraudulent website’s homepage, as well as the relative path and image name of the image element in the HTML source code; S2: Collecting web page data of the official website, and using the web page data of the official website to build a comparison database; S3: Based on the page features of the entry page, determine whether the website to be identified has an entry page. If the website to be identified has an entry page, log in to the website to be identified using the data in the collision database, and download the page image of the website to be identified using the data in the collision database; if the website to be identified does not have an entry page, directly enter the website to be identified and download the page image of the website to be identified; The step of determining whether the website to be identified has an entry page includes: Use a simulator on the terminal to simulate access to the website to be identified and obtain the HTML source code of the website to be identified; Match the number of input boxes contained in the website to be identified according to the HTML source code of the official website , Number of paragraphs and number of hidden text , then the formula for determining the existence of the entry page is: in is the set threshold; S4: extracting text features and image features from the page image of the website to be identified, and comparing the text features and / or image features with a comparison database. When the comparison is successful, it is determined that the website to be identified is a fraudulent website.

2. The method for identifying fraudulent websites according to claim 1, characterized in that: In S3, the specific steps of determining whether there is an entry page for the website to be identified and downloading the page image of the website to be identified using the data in the collision database are as follows: If it is determined that the website to be identified has an entry page, the name of the entry page of the website to be identified is recorded, and the name of the entry page of the fraudulent website is searched in the collision database; If there is an entry page name of the fraudulent website that is the same as the entry page name of the website to be identified in the collision database, use the entry password corresponding to the entry page name of the fraudulent website to enter the homepage of the website to be identified, and use the relative path and image name of the image element in the HTML source code of the collision database to use a get request to download the page image of the website to be identified; If there is no entry page name of the fraudulent website that is the same as the entry page name of the website to be identified in the collision database or the entry password corresponding to the entry page name of the fraudulent website is invalid, the relative path of the homepage of the fraudulent website in the collision database is used to directly access the homepage of the website to be identified, and the relative path and image name of the image element in the HTML source code in the collision database are used to use a get request to download the page image of the website to be identified.

3. The method for identifying fraudulent websites according to claim 1, characterized in that: S2 specifically includes the following steps: S2.1: Obtain a batch of known official websites and use a simulator to simulate access to the official websites on the terminal; S2.2: Obtain the HTML source code of the official website, obtain the page paragraphs of the official website according to the HTML source code of the official website, render and screenshot the page paragraphs, and obtain screenshots of each page paragraph of the official website; S2.3: using regular expressions to obtain image elements in the HTML source code of the official website, downloading images of the official website according to the image elements, and performing feature extraction on the images of the official website to obtain image features of the official website; S2.4: Match and extract Chinese characters in the HTML source code of the official website, perform word segmentation on the extracted Chinese characters to obtain corresponding words, and calculate the weight of each word; S2.5: Using screenshots of each page paragraph of the official website, image features, words and word weights as data, a comparison database is constructed.

4. The method for identifying fraudulent websites according to claim 3, characterized in that: In S4, Chinese characters are extracted from the page image of the website to be identified, and the extracted Chinese characters are segmented. Then, the words obtained after segmentation are queried in the comparison database for the corresponding word weights. According to the word weights, it is determined whether the website to be identified is a fraudulent website. The judgment formula is as follows: in, Indicates the top 5 words with the highest weights. Indicates i The word weight of the website to be identified, is a configurable threshold.

5. The method for identifying fraudulent websites according to claim 3, characterized in that: In S4, the image features of the page images of the website to be identified are extracted, and the image features are searched based on cosine similarity with the image features of the official website in the comparison database. If the corresponding image feature elements with a cosine similarity higher than a preset cosine similarity threshold are retrieved, the website to be identified is judged to be a fraudulent website.

6. A fraudulent website identification system, applied to the fraudulent website identification method according to any one of claims 1 to 5, characterized in that: include: A data collection module is used to collect web page data of known fraudulent websites and web page data of official websites; Build a collision database to store web page data of fraudulent websites; Comparison database, used to store web page data of official websites; The entry page detection module is used to determine whether there is an entry page on the website to be identified; The page image download module is used to download the page image of the website to be identified; if the website to be identified has an entry page, the website to be identified is logged in using the data in the collision database, and the page image of the website to be identified is downloaded using the data in the collision database; if the website to be identified does not have an entry page, the website to be identified is directly entered and the page image of the website to be identified is downloaded; A feature extraction module is used to extract text features and image features from the page images of the website to be identified; The identification module is used to compare the text features and image features in the page image of the website to be identified with the comparison database respectively to determine whether the website to be identified is a fraudulent website.

7. The fraud website identification system according to claim 6, characterized in that: The web page data of the fraudulent website includes the relative path of the home page of the fraudulent website, the name of the home page, the name of the entry page and the entry password, as well as the relative path and image name of the image element in the HTML source code.

8. The fraud website identification system according to claim 7, characterized in that: The web page data of the official website includes screenshots of each page paragraph of the official website, picture features, words and word weights.

Citation Information

Patent Citations

  • Method for detecting phishing web pages with spatial mixed index mechanism

    CN102662959A

  • Database collision early warning method based on big data analysis

    CN109450866A

  • Fraud website identification method and device, equipment and storage medium

    CN113779481A

  • Phishing website detection

    US20210099484A1