An applet privacy protection policy display violation detection method

Abstract

The application discloses a method for detecting display violation of a mini-program privacy protection policy, which comprises the following steps: analyzing the user interaction components and the UI tree structure of the mini-program by an Airtest dynamic engine, understanding the component types and designing a classification analysis method; searching and verifying the existing form of the privacy protection policy by combining the dynamic engine tool, judging whether the privacy protection policy exists according to the characteristics of the mini-program page, judging whether the user will be prompted to read the privacy protection policy when starting the mini-program for the first time, judging whether there is a problem of default user agreement to the privacy protection policy in the case where the privacy protection policy exists, and generating an analysis result of the display violation of the mini-program privacy protection policy. The application can detect the display violation of the mini-program privacy protection policy, determine whether the mini-program fails to provide the privacy protection policy to the user and whether there is a problem of default user agreement to the privacy protection policy, and improve the safety guarantee of the user when using the mini-program.

Description

Technical Field

[0001] This invention relates to the field of mini-program privacy policy analysis in mobile applications, and in particular to a method for detecting demonstrative violations of mini-program privacy policies. Background Technology

[0002] With the development of internet technology, mini-programs, applications integrated into third-party platforms such as WeChat, Alipay, Douyin, and Baidu, have gradually become an indispensable service in people's lives and continue to evolve. According to relevant data, by the end of 2022, there were over 7.8 million internet mini-programs, with a daily active user count exceeding 800 million and an average usage frequency of over 12 times per user. Mini-programs have become integrated into all aspects of netizens' lives. However, as the internet mini-program market continues to expand, the types of mini-programs are becoming increasingly complex, user data is becoming increasingly insecure, and there are more and more instances of a lack of privacy protection policies and default consent to privacy policies.

[0003] To protect user data security, national regulations stipulate that mobile applications such as mini-programs must have specific privacy protection policies that describe the permissions involved in user privacy during application use. Applications can only collect and use user data with the user's consent. However, current security analysis technologies primarily focus on the application's own data usage, with very few analyses combining privacy protection policies with the software's presentational characteristics.

[0004] The privacy policy of mini-programs has the following main problems in terms of analyzing display violations:

[0005] 1) The mini-program does not provide users with a readable privacy policy or user agreement related to user information and data;

[0006] 2) When the mini-program is run for the first time, it does not prompt users to read the privacy protection policy and other rules for the collection and use of privacy data through obvious means such as pop-ups;

[0007] 3) On the interface prompting users to read the privacy policy, obtain user consent in a non-explicit manner, such as by having the user default to agreeing to the privacy policy; Summary of the Invention

[0008] The purpose of this invention is to propose a method for detecting the display violations of privacy protection policies in mini-programs, in order to solve the above-mentioned technical problems. This invention analyzes the display violations of privacy protection policies in mini-programs in three steps: (1) Analyzing the user interaction components and UI tree structure of the mini-program through the Airtest dynamic engine, understanding the component types and designing a classification analysis method; (2) Searching and verifying the existence of privacy protection policies using dynamic engine tools, judging whether privacy protection policies exist based on the characteristics of the mini-program page, and judging whether the user is prompted to read the privacy protection policy when the mini-program is first launched; (3) Judging whether there is a problem of defaulting user consent to the privacy protection policy based on the information between mini-program components and combined with a deep learning image recognition model, and generating analysis results of display violations of privacy protection policies in mini-programs. Through the above methods, the display violations of privacy protection policies in mini-programs can be detected, it can be determined whether the mini-program has not provided privacy protection policies to users and whether there is a problem of defaulting user consent to the privacy protection policy, thereby improving the security of users when using mini-programs.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A method for detecting violations of privacy protection policies in mini-programs includes the following steps:

[0011] 1) Privacy policy reading prompt detection based on component information feature point traversal:

[0012] Step S101: For the mini-program P to be tested, obtain its component information M and UI tree information K using the Airtest dynamic engine tool;

[0013] Step S102: According to step S101, add the mini-program P to be detected to the detection thread list, start the monitoring thread for page authorization and prompt pop-up, identify key components of the authorization and prompt pop-up of the current page to determine the pop-up status, and close the corresponding pop-up according to the default rejection and cancellation policy;

[0014] Step S103: Based on the judgment information of the component information M and UI tree information K in the mini program P after the page authorization and prompt pop-up monitoring thread in step S102 in step S101, perform feature point identification and monitoring of the privacy protection policy on the homepage of mini program P; and determine whether mini program P prompts users to read the privacy protection policy and other collection and usage rule results E on the homepage based on the identification and test results.

[0015] Step S104: Based on the inter-component information M and UI tree information K in step S101, perform a depth-first search based on the nodes obtained by the dynamic tool Airtest. Under the guidance of the privacy policy keyword feature dictionary, determine whether the current inter-component information M and UI tree information K contain a privacy policy entry or relay page. Based on the step-by-step test click results and the explicit representation features of the privacy policy, determine the reachable path S of the privacy policy of the mini-program P.

[0016] 2) Analysis of default consent for privacy policies based on image recognition checkboxes:

[0017] Step S105: Based on the reachability path S of the privacy policy determined in step S104, analyze the information M between page components and the UI tree information K of the mini-program P. Click to jump to the privacy policy page, and filter out the core component coordinate region R(T) based on the existence of coordinate X in the privacy policy. l T r B l B r The screenshot is then used to determine, via a pre-trained privacy policy agreement checkbox status classifier F, whether the current page indicates default agreement to the privacy policy (result C).

[0018] Step S106: Based on the status judgment result C of whether the user agrees to the privacy policy by default in step S105, and combined with the result E of whether the user is prompted to read the privacy policy and other collection and usage rules on the homepage in step S104, the final result set of the display violation issues E and C of the mini program P can be output.

[0019] Furthermore, step S102 specifically includes:

[0020] Step S201: Determine from the dynamic page traversal of the current mini-program P and the Activity name of the current mini-program that, in the absence of page navigation, read the authorization window component information identifier and authorization window keyword ontology information of the current page.

[0021] Step S202: If an authorization pop-up exists on the current page, search for the keyword K containing one of the six types of authorization pop-ups: "Request", "Authorize", "Get Your", "Information", "Allow", and "Deny". a ;

[0022] Step S203: Using keywords, click the corresponding authorization window feature value K. a The authorization window closing function is triggered at the location of component A;

[0023] Step S204: If a pop-up notification window exists on the current page, search for keywords K containing three types of notification windows: "Notification," "OK," and "Cancel." t ;

[0024] Step S205: Using the keyword, click the corresponding feature value K in the prompt window. t The prompt window closes at the location of component T;

[0025] Step S206: Finally output the function result of dynamically closing the authorization and prompting pop-up window on the current page of the mini-program P.

[0026] Furthermore, the method for detecting the homepage privacy policy button and pop-up based on feature point recognition in step S103 and the method for searching the privacy policy relay page and features based on node depth traversal in step S104 are specifically as follows:

[0027] Step S301: After obtaining the information M between components on the current page of the mini-program P and the UI tree information K, generate a list H of all nodes and relationships within the current page of the mini-program P.

[0028] Step S302: Perform a node depth traversal search method based on the dynamic tool Airtest for all nodes and relationship list H in the current page of the mini program P. If there is no login or privacy policy click button on the current page, randomly select node n from the layout file.

[0029] Step S303: During the traversal of node n, dynamically compare the page changes before and after the click based on the node characteristics of different pages, and search for the relay page Page(n) of the privacy protection policy. x (This continues until all nodes have been traversed or the user is redirected to the privacy policy display page.)

[0030] Step S304: If the current page contains a login button or a privacy policy click button, then click the corresponding privacy policy relay page Page(n) x The entry point or corresponding button;

[0031] Step S305: Simultaneously, combining the page characteristics of the privacy policy, attempt to log in and click on the privacy policy to determine whether the privacy policy exists, and generate the reachable path S of the privacy policy based on the click order.

[0032] Furthermore, in step S105, the privacy policy agreement checkbox status recognition method based on image recognition will pre-construct a privacy policy checkbox status judgment image recognition classifier F according to the following method:

[0033] To construct the image recognition classifier F for judging the status of privacy policy checkboxes, it is necessary to first determine the types and representations of privacy policy checkboxes, including eight attributes: checkbox shape, checked state graphic, checked state color, checked state display intensity, relative position of the checkbox and the privacy policy, and checkbox and privacy policy component information. Then, based on the checked state of the checkboxes in the image, corresponding image data and corresponding label datasets are prepared. After training the corresponding model using an image recognition deep learning algorithm, the final input needs to be a screenshot containing the text features of the privacy policy and the overall area of ​​the privacy policy checkboxes. The output is the judgment result of the privacy policy checkbox status, namely, no privacy policy checkbox, privacy policy checkbox present but not checked, and privacy policy checkbox present and checked.

[0034] Furthermore, the method for recognizing the privacy policy consent checkbox status based on image recognition in step S105 is specifically as follows:

[0035] Step S401: Based on the node information and component information M and the coordinate X information of the privacy policy within the mini-program P page, generate the privacy policy and checkbox area R(T). l T r B l B r ), where T l T r B l B r These are the coordinates of the four vertices of the rectangle representing the privacy policy and the checkbox area;

[0036] Step S402: Based on page features, privacy policy, and checkbox area R(T) l T r B l B r The corresponding coordinates of ) are used to determine the coordinates of the vertices (T) of the rectangle corresponding to screenshot G. l T t B l B r );

[0037] Step S403: Input the screenshot G generated in step S402 into the privacy policy checkbox status judgment image recognition classifier F to obtain the privacy policy checkbox status judgment result;

[0038] The privacy policy checkbox status will be determined in one of the following three ways: no privacy policy checkbox, a privacy policy checkbox is present but not checked, or a privacy policy checkbox is present and checked.

[0039] Step S404: Based on the result of step S403, determine whether the current page has a privacy protection policy checkbox. If the result is true, proceed to S405; otherwise, proceed to S406.

[0040] Step S405: If it is determined that the current page has a privacy policy checkbox, based on the keyword features of the privacy policy text on the current page, combined with the output result of the image recognition classifier F of the privacy policy checkbox status judgment, the checkbox status bar is judged and the result is generated;

[0041] Step S406: Based on the judgment result of the privacy policy checkbox status image recognition classifier F in step S404, and after logical unification with the judgment result in step S405, the default consent result C of the privacy policy of the mini-program P is generated.

[0042] A further improvement of the present invention is that: the Airtest dynamic engine-based tool in step S101 can generate inter-component information M and UI tree information K of the mini-program P to be tested; and the page authorization and pop-up monitoring thread in step S102 can first determine whether there is a mini-program jump problem on the current page based on the current Activity, and then specifically determine and close page pop-ups and prompt windows in mini-program P, so as to facilitate the sequential execution of the test process.

[0043] A further improvement of the present invention is that: in steps S103 and S104, the method of detecting the privacy policy button and pop-up on the homepage is combined with the feature point recognition of the privacy policy, and the method of searching the relay page and features of the privacy policy by using the node depth traversal based on the dynamic tool Airtest is used to establish the collection and use rule results E such as whether the mini-program P under test prompts the user to read the privacy policy on the homepage.

[0044] A further improvement of the present invention is that: in steps S105 and S106, a privacy protection policy checkbox status judgment image recognition classifier F is pre-constructed, and based on the current page component information M of the mini-program and the privacy protection policy existence coordinate X information, a privacy protection policy and checkbox area R(T) is generated. l T t B l B r By combining the classification results of the image recognition classifier F based on the privacy policy checkbox status judgment with the text features of the page privacy policy, the default consent result C of the privacy policy was established.

[0045] Compared with the prior art, the present invention has the following advantages:

[0046] 1) The method of the present invention can automatically analyze the privacy protection policies of different types of mini programs and different display formats, which improves efficiency and accuracy compared with the manual review of privacy protection policy display violations;

[0047] 2) This invention combines user interaction components and page feature display information for analysis, incorporates deep learning model algorithms and dynamic testing and verification methods, and simulates human click operations for dynamic analysis, making the analysis results more intuitive and comprehensive;

[0048] 3) Based on the relevant provisions of national cybersecurity laws and regulations, this invention defines and proposes methods to solve three types of problems: mini-programs do not provide privacy protection policies, do not prompt users to read privacy protection policies upon first launch, and default users to agree to privacy protection policies;

[0049] 4) This invention provides a solution to the problem of display violations of privacy protection policies in mini-programs, namely, combining mini-program page component analysis, deep learning image recognition algorithms and dynamic testing and verification methods, which is an application of emerging artificial intelligence technology in the field of software security. Attached Figure Description

[0050] Figure 1 This is an overall flowchart of the method for judging and analyzing privacy protection policy violations based on mini-programs according to the present invention;

[0051] Figure 2 This is a flowchart of the monitoring thread method for authorization and prompt pop-up windows on the current page of the mini-program of this invention;

[0052] Figure 3 This is a flowchart of the method of the present invention for detecting the privacy policy button and pop-up on the homepage based on feature point recognition and searching for the relay page and features of the privacy policy based on node depth traversal obtained by the dynamic tool Airtest.

[0053] Figure 4 This is a flowchart of the method of the present invention for recognizing the status of the privacy policy consent checkbox based on image recognition and determining whether the privacy policy is agreed to by default by combining the text features of the privacy policy on the page. Detailed Implementation

[0054] The following describes in detail, with reference to the accompanying drawings, the specific implementation method of the present invention for detecting violations of the privacy protection policy of mini-programs.

[0055] Figure 1 This is an overall flowchart of the method for judging and analyzing privacy protection policy violations based on mini-programs according to the present invention;

[0056] This invention discloses a method for detecting violations of privacy protection policies displayed in mini-programs, comprising the following steps:

[0057] Step S101: For the mini-program P to be tested, obtain its component information M and UI tree information K using the Airtest dynamic engine tool;

[0058] Step S102: According to step S101, add the mini-program P to be detected to the detection thread list, start the monitoring thread for page authorization and prompt pop-up, identify key components of the authorization and prompt pop-up of the current page to determine the pop-up status, and close the corresponding pop-up according to the default rejection and cancellation policy.

[0059] Figure 2 This is a flowchart of the thread monitoring method for authorization and prompt pop-up windows on the current page of the mini-program of this invention.

[0060] Specifically, the method for enabling page authorization and pop-up monitoring thread in step S102 can be divided into the following steps:

[0061] Step S201: Determine from the dynamic page traversal of the current mini-program P and the Activity name of the current mini-program that, in the absence of page navigation, read the authorization window component information identifier and authorization window keyword ontology information of the current page.

[0062] Step S202: If an authorization pop-up exists on the current page, search for the keyword K containing one of the six types of authorization pop-ups: "Request", "Authorize", "Get Your", "Information", "Allow", and "Deny". a ;

[0063] Step S203: Using keywords, click the corresponding authorization window feature value K. a The authorization window closing function is triggered at the location of component A;

[0064] Step S204: If a pop-up notification window exists on the current page, search for keywords K containing three types of notification windows: "Notification," "OK," and "Cancel." t ;

[0065] Step S205: Using the keyword, click the corresponding feature value K in the prompt window. t The prompt window closes at the location of component T;

[0066] Step S206: Finally output the function result of dynamically closing the authorization and prompting pop-up window on the current page of the mini-program P.

[0067] Step S103: Based on the judgment information of the component information M and UI tree information K in the mini program P after the page authorization and prompt pop-up monitoring thread in step S102 in step S101, perform feature point identification and monitoring of the privacy protection policy on the homepage of mini program P; and determine whether mini program P prompts users to read the privacy protection policy and other collection and usage rule results E on the homepage based on the identification and test results.

[0068] Step S104: Based on the component information M and UI tree information K in step S103, perform a depth-first search based on the nodes obtained by the dynamic tool Airtest. Under the guidance of the privacy policy keyword feature dictionary, determine whether the current component information M and UI tree information K contain a privacy policy entry or relay page. Based on the step-by-step test click results and the explicit representation features of the privacy policy, determine the reachable path S of the privacy policy of the mini-program P.

[0069] Figure 3 This is a flowchart of the method for detecting the privacy policy button and pop-up on the homepage based on feature point recognition, and searching for the privacy policy relay page and features based on the node depth traversal obtained by the dynamic tool Airtest.

[0070] Specifically, the method for detecting the homepage privacy policy button and pop-up based on feature point recognition in step S103 and the method for searching the privacy policy relay page and features based on node depth traversal in step S104 can be divided into the following steps:

[0071] Step S301: After obtaining the information M between components on the current page of the mini-program P and the UI tree information K, generate a list H of all nodes and relationships within the current page of the mini-program P.

[0072] Step S302: Perform a node depth traversal search method based on the dynamic tool Airtest for all nodes and relationship list H in the current page of the mini program P. If there is no login or privacy policy click button on the current page, randomly select node n from the layout file.

[0073] Step S303: During the traversal of node n, dynamically compare the page changes before and after the click based on the node characteristics of different pages, and search for the relay page Page(n) of the privacy protection policy. x (This continues until all nodes have been traversed or the user is redirected to the privacy policy display page.)

[0074] Step S304: If the current page contains a login button or a privacy policy click button, then click the corresponding privacy policy relay page Page(n) x The entry point or corresponding button;

[0075] Step S305: Simultaneously, combining the page characteristics of the privacy policy, attempt to log in and click on the privacy policy to determine whether the privacy policy exists, and generate the reachable path S of the privacy policy based on the click order.

[0076] Step S105: Based on the reachability path S of the privacy policy determined in step S104, analyze the information M between page components and the UI tree information K of the mini-program P. Click to jump to the privacy policy page, and filter out the core component coordinate region R(T) based on the existence of coordinate X in the privacy policy. l T r B l B r The screenshot is then used to determine, via a pre-trained privacy policy agreement checkbox status classifier F, whether the current page indicates default agreement to the privacy policy (result C).

[0077] Step S106: Based on the status judgment result C of whether the user agrees to the privacy policy by default in step S105, and combined with the result E of whether the user is prompted to read the privacy policy and other collection and usage rules on the homepage in step S104, the final result set of the display violation issues E and C of the mini program P can be output.

[0078] Figure 4 This is a flowchart of the method of the present invention for recognizing the status of the privacy policy consent checkbox based on image recognition and determining whether the privacy policy is agreed to by default by combining the text features of the privacy policy on the page.

[0079] Specifically, the privacy policy agreement checkbox status recognition method based on image recognition in step S105 will pre-construct a privacy policy checkbox status judgment image recognition classifier F according to the following method:

[0080] To construct the image recognition classifier F for judging the status of privacy policy checkboxes, it is necessary to first determine the types and representations of privacy policy checkboxes, including eight attributes: checkbox shape, checked state graphic, checked state color, checked state display intensity, relative position of checkbox and privacy policy, and checkbox and privacy policy component information. Then, based on the checked state of the checkboxes in the image, corresponding image data and corresponding label datasets are prepared. After training the corresponding model using an image recognition deep learning algorithm, the final input needs to be a screenshot containing the text features of the privacy policy and the overall area of ​​the privacy policy checkboxes, and the output should be the judgment result of the privacy policy checkbox status.

[0081] Step S401: Based on the node information and component information M and the coordinate X information of the privacy policy within the mini-program P page, generate the privacy policy and checkbox area R(T).l T r B l B r ), where T l T r B l B r These are the coordinates of the four vertices of the rectangle representing the privacy policy and the checkbox area;

[0082] Step S402: Based on page features, privacy policy, and checkbox area R(T) l T r B l B r The corresponding coordinates of ) are used to determine the coordinates of the vertices (T) of the rectangle corresponding to screenshot G. l T r B l B r );

[0083] Step S403: Input the screenshot G generated in step S402 into the privacy policy checkbox status judgment image recognition classifier F to obtain the privacy policy checkbox status judgment result;

[0084] The privacy policy checkbox status will be determined in one of the following three ways: no privacy policy checkbox, a privacy policy checkbox is present but not checked, or a privacy policy checkbox is present and checked.

[0085] Step S404: Based on the result of step S403, determine whether the current page has a privacy protection policy checkbox. If the result is true, proceed to S405; otherwise, proceed to S406.

[0086] Step S405: If it is determined that the current page has a privacy policy checkbox, based on the keyword features of the privacy policy text on the current page, combined with the output result of the image recognition classifier F of the privacy policy checkbox status judgment, the checkbox status bar is judged and the result is generated;

[0087] Step S406: Based on the judgment result of the privacy policy checkbox status image recognition classifier F in step S404, and after logical unification with the judgment result in step S405, the default consent result C of the privacy policy of the mini-program P is generated.

[0088] Based on the status judgment result C of whether the user agrees to the privacy policy by default in step S105, and combined with the result E of whether the user is prompted to read the privacy policy and other collection and usage rules in step S104, the final output set of display violation issues E and C of the mini-program P can be obtained.

Claims

1. A method for detecting violations of privacy protection policies displayed in mini-programs, characterized in that, Includes the following steps: 1) Privacy policy reading prompt detection based on component information feature point traversal: Step S101: For the mini-program P to be tested, obtain its component information M and UI tree information K using the Airtest dynamic engine tool; Step S102: According to step S101, add the mini-program P to be detected to the detection thread list, start the monitoring thread for page authorization and prompt pop-up, identify key components of the authorization and prompt pop-up of the current page to determine the pop-up status, and close the corresponding pop-up according to the default rejection and cancellation policy; Step S103: Based on the judgment information of the component information M and UI tree information K in the mini program P after the page authorization and prompt pop-up monitoring thread in step S102 in step S101, perform feature point recognition and monitoring of the privacy protection policy on the homepage of mini program P, and determine whether mini program P prompts the user to read the privacy protection policy result E on the homepage based on the recognition and test results; Step S104: Based on the inter-component information M and UI tree information K in step S101, perform a depth-first search based on the nodes obtained by the dynamic tool Airtest. Under the guidance of the privacy policy keyword feature dictionary, determine whether the current inter-component information M and UI tree information K contain a privacy policy entry or relay page. Based on the step-by-step test click results and the explicit representation features of the privacy policy, determine the reachable path S of the privacy policy of the mini-program P. 2) Analysis of default consent to privacy policies based on image recognition checkboxes: Step S105: Based on the reachable path S of the privacy policy determined in step S104, analyze the information M between page components and the UI tree information K of the mini-program P. Click to jump to the privacy policy page, and filter out the core component coordinate area R based on the existence of coordinate X in the privacy policy. , , , And take a screenshot, among which , , , The coordinates of the four vertices of the privacy policy and checkbox area rectangle are given; the state C of whether the current page agrees to the privacy policy by default is determined by the pre-trained privacy policy agreement checkbox state recognition classifier F; Step S106: Based on the status judgment result C of whether the user agrees to the privacy policy by default in step S105, and combined with the result E of whether the user is prompted to read the privacy policy and other collection and usage rules on the homepage in step S104, the final result set of the display violation issues E and C of the mini program P can be output.

2. The method according to claim 1, characterized in that, Step S101 uses the Airtest dynamic engine, a cross-platform, image recognition-based UI automation testing framework that also includes the Poco UI control recognition-based automation testing framework. It can obtain the inter-component information M and UI tree information K of the mini-program P to be tested.

3. The method according to claim 1, characterized in that, Step S102 specifically involves: Step S201: Determine from the dynamic page traversal of the current mini-program P and the Activity name of the current mini-program that, in the absence of page navigation, read the authorization window component information identifier and authorization window keyword ontology information of the current page. Step S202: If an authorization pop-up exists on the current page, search for authorization pop-up keywords containing six types, including "Request", "Authorize", "Get Your", "Information", "Allow", and "Deny". ; Step S203: Using keywords, click the corresponding authorization window feature value. The authorization window closing function is triggered at the location of component A; Step S204: If a pop-up notification window exists on the current page, search for keywords containing the three types of notification windows: "notification," "OK," and "cancel." ; Step S205: Using keywords, click on the corresponding feature value in the prompt window. The prompt window closes at the location of component T; Step S206: Finally output the function result of dynamically closing the authorization and prompting pop-up window on the current page of the mini-program P.

4. The method according to claim 3, characterized in that, The specific method of step S102 is as follows: (1) Dynamically identify whether the current page of the mini program to be tested contains an authorization API call pop-up of the main App to obtain the user's personal information and close the authorization pop-up; (2) Dynamically identify whether the current page of the mini program to be tested contains a function pop-up designed by the mini program body to complete the specific function or notification of the mini program and close the function pop-up.

5. The method according to claim 1, characterized in that, The method for detecting the homepage privacy policy button and pop-up based on feature point recognition in step S103 and the method for searching the privacy policy relay page and features based on node depth traversal in step S104 are as follows: Step S301: After obtaining the information M between components on the current page of the mini-program P and the UI tree information K, generate a list H of all nodes and relationships within the current page of the mini-program P. Step S302: Perform a node depth traversal search method based on the dynamic tool Airtest for all nodes and relationship list H in the current page of the mini program P. If there is no login or privacy policy click button on the current page, randomly select node n from the layout file. Step S303: During the traversal of node n, dynamically compare the page changes before and after the click based on the node characteristics of different pages, and search for the relay page Page(n) of the privacy policy until all nodes have been traversed or the privacy policy display page has been reached. Step S304: If the current page has a login button or a privacy policy click button, click the entry point of the corresponding privacy policy relay page Page(n) or the corresponding button; Step S305: Simultaneously, combining the page characteristics of the privacy policy, attempt to log in and click on the privacy policy to determine whether the privacy policy exists, and generate the reachable path S of the privacy policy based on the click order.

6. The method according to claim 5, wherein step S302 performs a node depth-first traversal search method based on all nodes and the relationship list H within the current page of the mini-program P, obtained using the dynamic tool Airtest, characterized in that... Traverse and search the tree, traversing the nodes of the tree along the depth of the relation tree, and searching the branches of the tree as deeply as possible. When all the edges containing node v have been explored or the node does not meet the conditions during the search, the search will backtrack to the starting node of the edge that found node v. The whole process is repeated until all nodes have been visited.

7. The method according to claim 1, characterized in that, The privacy policy agreement checkbox status recognition method in step S105 requires calling the snapshot function in the Airtest dynamic engine framework to extract the page features of the privacy policy based on OCR technology, take a screenshot based on the resolution of the test device, and then preprocess the captured image. By combining the image text coordinate area, the text part in the image is removed, and the checkbox status image of the privacy policy is retained.

8. The method according to claim 7, characterized in that, The method for recognizing the privacy policy consent checkbox status based on image recognition in step S105 is as follows: Step S401: Based on the node information, component information M, and privacy policy coordinates X within the mini-program page P, generate the privacy policy and checkbox area R. , , , ),in , , , These are the coordinates of the four vertices of the rectangle representing the privacy policy and the checkbox area; Step S402: Based on page features, privacy policy, and checkbox area R ( , , , The corresponding coordinates of ) are used to determine the vertex coordinates of the rectangle corresponding to screenshot G. , , , ); Step S403: Input the screenshot G generated in step S402 into the privacy policy checkbox status judgment image recognition classifier F to obtain the privacy policy checkbox status judgment result; The privacy policy checkbox status will be determined in one of the following three ways: no privacy policy checkbox, a privacy policy checkbox is present but not checked, or a privacy policy checkbox is present and checked. Step S404: Based on the result of step S403, determine whether the current page has a privacy protection policy checkbox. If the result is true, proceed to S405; otherwise, proceed to S406. Step S405: If it is determined that the current page has a privacy policy checkbox, based on the keyword features of the privacy policy text on the current page, combined with the output result of the image recognition classifier F of the privacy policy checkbox status judgment, the checkbox status bar is judged and the result is generated; Step S406: Based on the judgment result of the privacy policy checkbox status image recognition classifier F in step S404, and after logical unification with the judgment result in step S405, the default consent result C of the privacy policy of the mini-program P is generated.

9. The method according to claim 8, characterized in that, The image recognition classifier F for determining the privacy protection policy checkbox status in step S403 needs to be built in advance. The specific construction method is as follows: To construct the image recognition classifier F for judging the status of privacy policy checkboxes, it is necessary to first determine the types and representations of privacy policy checkboxes, including six attributes: checkbox shape, checked state graphic, checked state color, checked state display intensity, relative position of the checkbox and the privacy policy, and checkbox and privacy policy component information. Then, based on the checked state of the checkboxes in the image, corresponding image data and corresponding label datasets are prepared. After training the corresponding model using an image recognition deep learning algorithm, the final input needs to be a screenshot containing the text features of the privacy policy and the overall area of ​​the privacy policy checkboxes. The output is the judgment result of the privacy policy checkbox status, namely, no privacy policy checkbox, privacy policy checkbox present but not checked, and privacy policy checkbox present and checked.

10. The method according to claim 8, characterized in that, The logical unification method between the judgment result of the privacy policy checkbox status image recognition classifier F in step S404 and the judgment result in step S405 in step S406 is as follows: Step S404 first takes a screenshot of the privacy policy and checkbox area coordinates, and then uses the privacy policy checkbox status judgment image recognition classifier to identify whether the checkbox is checked. Then, step S405 combines the privacy policy keyword features of the current page, such as "privacy policy" and "privacy agreement", which are clickable, redirectible, and can be identified as privacy policy URLs, to jointly generate the judgment result of whether the checkbox is checked. The purpose is to eliminate the interference of checkboxes such as "remember password", and combine the results of the two steps to generate a more accurate default consent result C of the privacy policy of the mini program P.

Images