CN120144423A - A detection system and method for robot verification browser - Google Patents
A detection system and method for robot verification browser Download PDFInfo
- Publication number
- CN120144423A CN120144423A CN202510622112.9A CN202510622112A CN120144423A CN 120144423 A CN120144423 A CN 120144423A CN 202510622112 A CN202510622112 A CN 202510622112A CN 120144423 A CN120144423 A CN 120144423A
- Authority
- CN
- China
- Prior art keywords
- verification
- browser
- sample
- simulated
- trajectory
- 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 OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3457—Performance evaluation by simulation
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
Abstract
本说明书实施例公开了一种用于机器人验证浏览器的检测通过系统及其方法，其中用于机器人验证浏览器的检测通过系统，包括拟用户模块，收集用户浏览器信息生成模拟行为样本，模拟行为样本调整浏览器环境,规划模型，模拟行为样本通过规划模型生成模拟行为对抗样本。本发明通过模拟行为样本的特征调整浏览器环境内符合当下时间节点的用户特征和/或根据模拟行为对抗样本内的特征调整浏览器环境，以此混淆浏览器检测端，从而提高模拟行为的真实性，通过用户的鼠标移动曲线速率特征与线性差值算法计算生成验证起始坐标移动至验证完成处坐标的路径，如此实现第一验证轨迹验证的模拟效果贴合真人验证，从而提高验证通过率。
The embodiments of this specification disclose a detection system and method for robot verification of browsers, wherein the detection system for robot verification of browsers includes a simulated user module, which collects user browser information to generate simulated behavior samples, adjusts the browser environment with simulated behavior samples, and plans models, and generates simulated behavior adversarial samples with simulated behavior samples through the planning model. The present invention adjusts the user characteristics in the browser environment that conform to the current time node through the characteristics of the simulated behavior samples and/or adjusts the browser environment according to the characteristics in the simulated behavior adversarial samples, thereby confusing the browser detection end, thereby improving the authenticity of the simulated behavior, and calculates and generates a path from the verification starting coordinates to the verification completion coordinates through the user's mouse movement curve rate characteristics and the linear difference algorithm, so as to achieve the simulation effect of the first verification trajectory verification to fit the real person verification, thereby improving the verification pass rate.
Description
Technical Field
Embodiments of the present disclosure relate to the field of robotic verification browsers, and in particular, to a detection passing system and method for a robotic verification browser.
Background
In modern internet environment, modern detection technology may combine means such as browser fingerprint, behavior analysis, machine learning, challenge verification (such as CAPTCHA), etc., internet verification is a basic stone of a digital world trust system, security and user experience are balanced through multi-level technology combination, as technology evolves, a verification mode is being changed from 'passive defense' to 'active adaptation', and combination of privacy protection and automatic wind control will be more focused in the future, so that a simple camouflage method is not effective any more, but as AI technology develops, a plurality of robot automation workflows need to simulate human access means to realize work automation, and robot detection technology causes trouble of automation.
Therefore, how to effectively counter these detection technologies, so that the automation of the normal user and the robot is not affected, and meanwhile, the success rate of bypassing the detection by the robot is improved, which is a problem to be solved.
Disclosure of Invention
In a first aspect, embodiments of the present disclosure provide a detection passing system for a robotic verification browser, comprising:
The simulation user module is used for collecting user browser information to generate a simulation behavior sample, and the simulation behavior sample is used for adjusting the browser environment;
the system comprises a planning model, a simulation behavior sample, a simulation behavior countermeasure sample and a simulation behavior countermeasure sample confusion browser detection end, wherein the simulation behavior countermeasure sample is generated through the planning model;
The identification module records and analyzes the page characteristics of the browser and interacts with the planning model, and inputs the page characteristics of the browser into the planning model to generate a first verification track so as to verify the detection end of the browser;
The first verification track is stripped of interference factors, the recognition module is trained through the simulation behavior countermeasure sample, the trained recognition module recognizes the browser page characteristics again and interacts with the planning model to generate a second verification track, and the browser detection end is verified again;
and the communication module is used for sending out real person verification by the browser detection end, and when the first verification track or the second verification track fails to pass the verification, the communication operator adopts man-machine cooperation or manual auxiliary verification.
Further, the first verification track and the second verification track comprise calculation of a mouse pointer movement track and linear acceleration, webpage features are identified through an identification module, and the distance between the mouse pointer and a browser detection end verification window is determined;
Calculating a first non-Bezier curve of the mouse pointer according to the simulated behavior sample;
the first non-Bezier curves generate a plurality of second non-Bezier curves according to the simulated behavior countermeasure sample;
Calculating a plurality of second non-Bezier curves to match with the equivalent similar simulation behavior samples by using a neural network to obtain an equivalent similarity ratio;
Setting an equivalent similarity threshold E, eliminating second non-Bezier curves lower than the equivalent similarity threshold E, and randomly selecting any one reserved second non-Bezier curve for controlling the movement of the mouse pointer to verify the detection end of the browser.
Further, the simulated behavior sample comprises features, each feature being a different neuron node, whereby an equivalent similarity is calculated from the different neuron nodes by the neural network;
The planning model calculates a simulation behavior sample matched with the current browser detection end by using a neural network, so as to assist the generation of a first verification track and a second verification track.
Further, browser page features include text, images, and symbols;
Recognizing characters, images and symbols through a recognition module to generate a plurality of element blocks, and judging a verification area, a simulation behavior sample and interference factors;
The interference factors comprise input method replacement, popup window, prompt message, network environment and network firewall;
the system comprises an amplifying element block, a plurality of pixel color points, a plurality of image processing units and a plurality of image processing units, wherein the pixel color points in the amplifying element block are distributed, the pixel color points are used as demarcation mark element outlines, each pixel color point in the outline is extracted to form an image outline, and the reduction multiplying power is sequentially reduced to generate a plurality of images;
historical data, judging images by comparing the plurality of images with the historical data and transmitting the images into a planning model;
The determined image is stored in the history data.
Further, the planning model generates an image countermeasure sample, the image countermeasure sample is transmitted into the recognition module, and the recognition module is trained;
The trained recognition module is used for adjusting the characteristic in the simulation behavior sample to adapt to the second verification track and the second non-Bezier curve.
Further, interacting with a browser detection end through a distributed network, so as to generate a plurality of simulated behavior samples;
And training the planning model is enhanced through a plurality of simulation behavior samples, and the data generation accuracy of the first verification track and the second verification track is improved.
And the planning model is used for matching the second verification track according to the simulation behavior sample and the interference factor priority processing habit.
Further, the browser detection end comprises sliding verification, text verification, graphic verification, voice verification, azimuth verification, behavior analysis verification and double-factor verification;
the features within the second verification track containing the simulated behavior samples adapt to the second non-bezier curve.
Further, collecting a plurality of face features, analyzing the face features through a convolutional neural network or generating an antagonistic network to simulate skin aging and wrinkle growth physiological changes, and simulating and generating a second face feature;
The second face feature is used for detecting the end through the browser.
Further, the verification of the true person fails, a matched operator is searched according to the second face characteristics, the second verification track and the simulation behavior sample, and the operator is determined and basic information is obtained;
the communication module contacts an operator through basic information, and the remote auxiliary robot completes verification of the browser detection end.
In a second aspect, embodiments of the present disclosure provide a detection passing method for a robot verification browser, including the steps of:
Collecting user information to generate a simulation sample, and importing the simulation sample into a planning model to generate a simulation countermeasure sample;
Identifying the page characteristics of the browser, importing the page characteristics into a planning model, calculating a first verification track, adjusting the confused browser detection end through a simulation sample and a simulation countermeasures sample, and executing the first verification track to verify the browser detection end;
When verification fails, eliminating interference factors in the first verification track, training a planning model by using a simulation countermeasure sample, generating a second verification track according to the page characteristics of the browser through the trained planning model, and executing the second verification track to verify the detection end of the browser;
When the second verification track is executed and does not pass the true person verification, the browser detection end communicates with an operator and adopts man-machine cooperation or manual auxiliary verification.
The technical scheme provided by some embodiments of the present specification has the following beneficial effects:
in various embodiments of the present disclosure, a detection passing system and a method for a robot verification browser are provided, which adjust user characteristics in a browser environment according to a current time node according to characteristics of a simulated behavior sample and/or adjust the browser environment according to characteristics in an antagonistic sample of a simulated behavior, so as to confuse a browser detection end, thereby improving the authenticity of the simulated behavior;
The current mouse pointer is determined to be located at the verification starting coordinate through the identification module, a path from the verification starting coordinate to the coordinate at the verification completion position is calculated and generated through the mouse movement curve speed characteristic and the linear difference algorithm of the user, and therefore the simulation effect of the first verification track verification is attached to the verification of a real person, and the verification passing rate is improved.
Other features and advantages of various embodiments of the present disclosure will be further disclosed in the following detailed description, the accompanying drawings.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present description, the drawings that are required in the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present description, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of a detection passing system architecture of a robot verification browser according to an embodiment of the present disclosure.
Detailed Description
The technical solutions of the embodiments of the present specification are explained and illustrated below with reference to the drawings of the embodiments of the present specification, but the following embodiments are only preferred embodiments of the present specification, and not all the embodiments. Based on the examples in the implementation manner, those skilled in the art may obtain other examples without making any creative effort, which fall within the protection scope of the present specification.
The terms first, second, third and the like in the description and in the claims and in the above drawings are used for distinguishing between different objects and not necessarily for describing a particular sequential or chronological order. Furthermore, the terms "comprise" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those listed steps or elements but may include other steps or elements not listed or inherent to such process, method, article, or apparatus.
In the following description, directional or positional relationships such as the terms "inner", "outer", "upper", "lower", "left", "right", etc., are presented merely to facilitate describing the embodiments and simplify the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operate in a particular orientation, and therefore should not be construed as limiting the description.
The data related to the application are information and data authorized by the user or fully authorized by all parties, and the collection of the related data complies with related laws and regulations and standards of related countries and regions.
Example 1
Referring to fig. 1, a detection passing system for a robot verification browser includes:
The simulation user module is used for collecting user browser information to generate a simulation behavior sample, and the simulation behavior sample is used for adjusting the browser environment;
the system comprises a planning model, a simulation behavior sample, a simulation behavior countermeasure sample and a simulation behavior countermeasure sample confusion browser detection end, wherein the simulation behavior countermeasure sample is generated through the planning model;
The identification module records and analyzes the page characteristics of the browser and interacts with the planning model, and inputs the page characteristics of the browser into the planning model to generate a first verification track so as to verify the detection end of the browser;
The first verification track is stripped of interference factors, the recognition module is trained through the simulation behavior countermeasure sample, the trained recognition module recognizes the browser page characteristics again and interacts with the planning model to generate a second verification track, and the browser detection end is verified again;
and the communication module is used for sending out real person verification by the browser detection end, and when the first verification track or the second verification track fails to pass the verification, the communication operator adopts man-machine cooperation or manual auxiliary verification.
The verification mode of the browser detection end comprises sliding verification, text verification, graphic verification, voice verification, azimuth verification, behavior analysis verification and double-factor verification;
collecting user browser information to generate a simulation behavior sample, wherein the simulation behavior sample comprises characteristics including a first face characteristic, a voice characteristic, a browser verification start time characteristic, a webpage searching habit characteristic of a webpage browser, a mouse moving curve rate characteristic, a screen brightness characteristic, a screen resolution characteristic, time zone setting, a language preference characteristic, a browser character size characteristic, a device model characteristic, an operating system version characteristic, a character input efficiency characteristic and a browser login IP characteristic;
wherein the characteristic data in the simulated behavior countermeasure sample is similar to the characteristic data;
when verification is needed, user characteristics conforming to the current time node in the browser environment are adjusted according to the characteristics of the simulation behavior sample, and/or the browser environment is adjusted according to the characteristics of the simulation behavior countermeasure sample, so that the detection end of the browser is confused, and the authenticity of the simulation behavior is improved;
Then identifying and recording the page characteristics of the browser through an identification module, firstly determining what type of verification is sent by a browser detection end, determining coordinates, secondly determining the position coordinates of a mouse pointer on the browser page, accordingly determining the linear distance from the mouse to the browser page verification mode, inputting the linear distance and a simulation behavior sample into a planning model, simulating by the planning model through a linear difference algorithm to generate a first verification track, and verifying the browser detection end through the first verification track;
The first verification track and the second verification track comprise a path from the current mouse mark coordinate to the verification starting coordinate and a path from the verification starting coordinate to the verification finishing coordinate;
the linear difference algorithm calculates a path for simulating movement of the attached current mouse mark coordinate to the verification start coordinate according to the speed characteristic of the mouse movement curve of the user, so that the mouse pointer is moved from the current position to the verification start coordinate of the browser detection end;
When the verification is required to be completed by the movement of the mouse, determining that the current mouse pointer is positioned at the initial verification coordinate through the identification module, calculating and generating a path from the initial verification coordinate to the coordinate at the position where the verification is completed through the speed characteristic of the mouse movement curve of the user and a linear difference algorithm, and thus realizing that the simulation effect of the first verification track verification is attached to the verification of a real person, thereby improving the verification passing rate;
It is specially noted that when the verification mode is character input verification, characters are automatically input through matching of character input efficiency characteristics and a planning model, so that character input verification of a true person is simulated;
It is to be specifically noted that, when the verification mode is face verification, the face verification is further assisted by combining the face video with the three-dimensional stereoscopic image through the matching verification of the first face feature and the planning model, wherein the first face feature comprises the three-dimensional stereoscopic image and the face video;
in summary, all features required by the verification modes exist in the simulation behavior sample, and only the features in the simulation behavior sample and the planning model are required to be mobilized to be matched with various verification modes;
It should be specifically noted that, in order to improve the authenticity of the simulation, the simulation behavior countermeasure sample is adopted to confuse the browser detection end, and in a specific embodiment, since the features in the simulation behavior countermeasure sample are similar to but different from the features in the simulation behavior sample, the linear distance, the simulation behavior sample and the simulation behavior countermeasure sample are all input into the planning model, the planning model utilizes the linear difference algorithm to simulate and generate a deformation verification track similar to the first verification track, and the deformation verification track is utilized to verify the browser detection end, so that the browser detection end is confused;
Identifying characters, images and symbols of the page characteristics of the browser through an identification module;
Recognizing characters, images and symbols through a recognition module to generate a plurality of element blocks, and judging a verification area, a simulation behavior sample and interference factors;
When the deformation verification track and the first verification track cannot pass through the browser detection end, removing interference factors in the first verification track, wherein the interference factors comprise input method replacement, popup windows, prompt messages, network environments and network firewalls, for example, popup windows or prompt messages of the current browser page, the first verification track is required to be removed according to a simulation behavior sample and is not processed to continue to perform verification, so that the browser detection end is confused, the coordinates of characters, images and symbols of the browser page characteristics are determined in a specific real-time mode through an identification module, then the coordinates of the current mouse pointer are positioned, the processing path of the mouse pointer is planned through a planning model, and the interference factors are removed or adjusted in sequence, so that the browser detection end is confused further;
because the verification mode of the browser detection end is interfered, the simulation behavior countermeasure sample is led into the identification module, and the adjustment parameters in the identification module are continuously adjusted through a counter-propagation algorithm so as to minimize a loss function, thereby training the identification module and improving the identification accuracy of the identification module;
The verification mode of the browser detection end is identified again by the trained identification module, verification coordinates are determined, a second verification track is generated, the second verification track is more accurate than the first verification track, and then the browser detection end is verified again through the second verification track;
The first face feature comprises a three-dimensional image and a face video, and is used for verifying the true person sent by the detection end, when the first verification track or the second verification track is not verified, the communication operator adopts man-machine cooperation or manual auxiliary verification, so that the situation that the browser detection end is locked and verified due to the fact that the robot verification fails all the time is avoided;
Setting verification recognition failure times, and when the verification recognition fails 3-5 times, communicating with a real operator matched with the current verification feature to verify, so that a browser detection end is further confused, and the subsequent detection passing rate is improved.
It should be noted that the drawing model is called a nonlinear programming model;
the principle of the nonlinear programming model is that an optimal solution is found under the condition of meeting nonlinear or linear constraint by taking an optimized nonlinear objective function as a core;
The mathematical form of the nonlinear programming model contains three elements:
path variable ,Determining a mouse starting point coordinate and a browser verification end point coordinate as coordinate points through an identification module;
an objective function, minimizing a nonlinear cost function, comprising:
Path length:
;
The transverse position and the longitudinal position of the kth time step in the path point are the mouse position coordinates (x, y);
Smoothness penalty-adding curvature term ;
Wherein the method comprises the steps ofIs the corner of the adjacent line segment,Is the step length;
Time efficiency by combining speed variation functions WhereinFor the path segment speed,Is the total path speed;
the shortest straight line path is determined firstly, then the curvature of the straight line path is adjusted based on a simulation behavior sample, and the movement time of the mouse is combined according to the path.
The constraint conditions include:
Interference factor avoidance, interference factor boundary constraint ,As the coordinates of the interference factor,In order to be of a safe radius,The transverse position and the longitudinal position of the kth time step in the path point are the mouse position coordinates (x, y);
Physical limitations:
Mathematical model example: is the maximum acceleration, Is the speed of,Is a curvature;
Start-end-point constraint, start-point Target point;
Overall mathematical model:
;
As curvature, λ is smoothness weight;
The path is expressed as a bezier curve:
;
By optimizing control points Generating a track path;
the whole steps are that control points are initialized, coordinates are adjusted to meet constraint, and then a Bezier curve is utilized to generate a track path.
The challenge samples included:
The GAN is set in the planning model, the GAN is used to generate the challenge network, the generator is used to directly synthesize the challenge samples,
The generator comprises:
input and output, receiving random noise z;
Outputting challenge samples ;
A loss function, wherein the counterloss (such as cross entropy) and disturbance constraint terms are jointly optimized, and the attack effect and the concealment are balanced by jointly optimizing the output countersamples;
The joint optimization formula includes:
;
Wherein, For the weight parameters of the disturbance constraint term,The weight parameter is the cross entropy;
Wherein the cross entropy formula comprises:
;
in order to provide the image vector data, Is model probability data;
The disturbance constraint term formula includes:
;
challenge sample data output for the generator;
x is original input data (picture data);
The strategy of the countermeasure training is combined training of a generator and a discriminator, the discriminator needs to identify a real sample and a countermeasure sample at the same time, the generator optimizes the countermeasure disturbance, and the gradient of the pre-training target model is reversely propagated to the generator to guide the disturbance to generate the countermeasure sample.
Example 2
The embodiment is an improvement made on the basis of embodiment 1, specifically referring to fig. 1, the first verification track and the second verification track include calculation of a movement track of a mouse pointer and linear acceleration, recognition of webpage features by an identification module, and determination of a distance between the mouse pointer and a verification window of a browser detection end;
Calculating a first non-Bezier curve of the mouse pointer according to the simulated behavior sample;
the first non-Bezier curves generate a plurality of second non-Bezier curves according to the simulated behavior countermeasure sample;
Calculating a plurality of second non-Bezier curves to match with the equivalent similar simulation behavior samples by using a neural network to obtain an equivalent similarity ratio;
Setting an equivalent similarity threshold E, eliminating second non-Bezier curves lower than the equivalent similarity threshold E, and randomly selecting any one reserved second non-Bezier curve for controlling the movement of the mouse pointer to verify the detection end of the browser.
Calculating a first non-Bezier curve of the mouse pointer, wherein the calculation formula is as follows:=;
Wherein, 、In order to simulate a mouse moving track coordinate point in a behavior sample characteristic and combine with a mouse moving curve speed characteristic, a first non-Bezier curve is calculated, the mouse mark moving track of a user is completely simulated through the first non-Bezier curve, so that a simulation effect is improved, and due to the fact that the mouse moving curve speed characteristic is referred, a first non-Bezier curve moving mode is firstly and rapidly close to a verification area of a browser detection end, after the first non-Bezier curve moving mode is close to the verification area of the browser detection end, the moving speed of a mouse mark is reduced, the first non-Bezier curve slides into the verification area of the browser detection end at a lower speed, and therefore the operation behavior of the user is accurately simulated, the browser detection end is further confused, and the detection accuracy is improved;
Further, in order to avoid that the browser detection end detects the same verification track for a plurality of times, a first non-Bezier curve and a simulation behavior countermeasure sample are input into a planning model, a plurality of second non-Bezier curves are generated through simulation, the plurality of second non-Bezier curves are calculated and matched with the simulation behavior sample through a neural network, a plurality of identical similarity rates are obtained through the calculation, an identical similarity rate threshold E is set, the second non-Bezier curves lower than the identical similarity rate threshold E are removed, the second non-Bezier curves with high matching degree in the simulation behavior sample are reserved, the second non-Bezier curves selected through the method are different from the simulation behavior sample, but the difference is not large, and the simulation manual verification track is better, so that vigilance of the browser detection end is reduced, and verification passing rate is improved.
The simulated behavior sample comprises features, wherein each feature is a different neuron node, so that an equivalent similarity rate is calculated according to the different neuron nodes through a neural network;
The planning model calculates a simulation behavior sample matched with the current browser detection end by using a neural network, so as to assist the generation of a first verification track and a second verification track;
The browser page features include text, images and symbols;
Recognizing characters, images and symbols through a recognition module to generate a plurality of element blocks, and judging a verification area, a simulation behavior sample and interference factors;
The interference factors comprise input method replacement, popup window, prompt message, network environment and network firewall;
the system comprises an amplifying element block, a plurality of pixel color points, a plurality of image processing units and a plurality of image processing units, wherein the pixel color points in the amplifying element block are distributed, the pixel color points are used as demarcation mark element outlines, each pixel color point in the outline is extracted to form an image outline, and the reduction multiplying power is sequentially reduced to generate a plurality of images;
And the historical data is used for comparing the plurality of images with the historical data to judge the images and transmitting the images into the planning model.
Because the verification picture of the browser is misled, in order to accurately identify characters, images and symbols of the browser page, firstly, all the characters, the images and the symbols of the browser page are framed to form a picture, the picture is composed of a plurality of pixel points, all the pixel points are enlarged, different pixel color points are used as demarcation mark element outlines, the pixel points in each outline are extracted to form an image outline, then the image outline is reduced in sequence, and the identification module is used for determining why the image outline is internally provided with the features, and the features obtained by each reduction judgment adopt the features which appear for a plurality of times as correct features;
it should be noted that, by storing the image in the history data, the auxiliary recognition module can quickly determine why the image is characterized in the outline.
Storing the determined image in the history data;
generating an image countermeasure sample by the planning model, transmitting the image countermeasure sample into the recognition module, and training the recognition module;
The trained recognition module is used for adjusting the characteristic in the simulation behavior sample to adapt to the second verification track and the second non-Bezier curve.
Further, an image countermeasure sample is generated through the planning model, the recognition module is trained through the image countermeasure sample, so that the trained recognition module is used for recognizing why the image profile is the same, when the two judging results are the same, the recognition is accurate, when the two judging results are different, the recognition module is trained through the image countermeasure sample to recognize why the image profile is the same, the process is repeated until the last two recognition results are the same, the detection accuracy of the recognition module is improved, and the fault tolerance of the recognition module is improved.
Interaction is performed between the distributed network and a browser detection end, so that a plurality of simulated behavior samples are generated;
And training the planning model is enhanced through a plurality of simulation behavior samples, and the data generation accuracy of the first verification track and the second verification track is improved.
Further, a distributed network is adopted to interact with a browser detection end, so that a plurality of simulated behavior samples are generated, each simulated behavior sample is different, and each simulated behavior sample is associated with each real operator, so that when the browser detection end needs to be verified, firstly, determining which simulated behavior sample in the range of the current time period can be used for verifying the browser detection end, then connecting the distributed network according to the determined simulated behavior sample, and verifying the browser detection end by the real recovery operation, thereby further improving the probability of passing the robot verification;
the planning model is used for simulating behavior samples according to the first verification track, and the planning model is used for planning the second verification track according to the simulation behavior samples and the interference factors.
When the interference factors are excessive, automatically planning and processing the interference factor lines through a planning model according to the processing habit in the simulation behavior sample, and guiding the interference factor line into a second verification track, so that the second verification track is perfected, and the simulation verification effect of the robot is improved;
the browser detection end comprises sliding verification, text verification, graphic verification, voice verification, azimuth verification, behavior analysis verification and double-factor verification;
The characteristics of the second verification track containing the simulation behavior sample are adapted to a second non-Bezier curve;
Collecting a plurality of face features, analyzing the face features through a convolutional neural network or generating an antagonistic network to simulate skin aging and wrinkle growth physiological changes, and simulating to generate a second face feature;
The second face feature is used for passing through a browser detection end;
The true man verification fails, a matched operator is searched according to the second face characteristics, the second verification track and the simulation behavior sample, and the operator is determined and basic information is obtained;
the communication module contacts an operator through basic information, and the remote auxiliary robot completes verification of the browser detection end.
Further, the facial features are analyzed through a convolutional neural network or a generating countermeasure network to simulate skin aging and wrinkle growth physiological changes, and the hairstyle changes and coat collocation changes of an operator are recorded, so that a second facial feature is generated through simulation, and the simulation effect is further improved through the second facial feature;
the method is characterized in that the hairstyle change and the jacket collocation change are related to weather, the hairstyle change day of an operator is determined, the hairstyle change of the operator is simulated, and the jacket of the operator is adjusted according to the weather of the day, so that the reality is improved;
When the planning module simulates that the simulated three-dimensional head portrait and the three-dimensional head portrait video generated according to the second face features recorded in daily life cannot pass verification, the communication operator adopts man-machine cooperation or manual auxiliary verification.
The invention also provides a detection passing method for the robot verification browser, which comprises the following steps:
Collecting user information to generate a simulation sample, and importing the simulation sample into a planning model to generate a simulation countermeasure sample;
Identifying the page characteristics of the browser, importing the page characteristics into a planning model, calculating a first verification track, adjusting the confused browser detection end through a simulation sample and a simulation countermeasures sample, and executing the first verification track to verify the browser detection end;
When verification fails, eliminating interference factors in the first verification track, training a planning model by using a simulation countermeasure sample, generating a second verification track according to the page characteristics of the browser through the trained planning model, and executing the second verification track to verify the detection end of the browser;
When the second verification track is executed and does not pass the true person verification, the browser detection end communicates with an operator and adopts man-machine cooperation or manual auxiliary verification.
The foregoing is merely a preferred embodiment of the present invention, and it should be noted that it will be apparent to those skilled in the art that modifications and variations can be made without departing from the technical principles of the present invention, and these modifications and variations should also be regarded as the scope of the invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510622112.9A CN120144423B (en) | 2025-05-15 | 2025-05-15 | Detection passing system and method for robot verification browser |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510622112.9A CN120144423B (en) | 2025-05-15 | 2025-05-15 | Detection passing system and method for robot verification browser |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN120144423A true CN120144423A (en) | 2025-06-13 |
| CN120144423B CN120144423B (en) | 2025-08-01 |
Family
ID=95959204
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510622112.9A Active CN120144423B (en) | 2025-05-15 | 2025-05-15 | Detection passing system and method for robot verification browser |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120144423B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108416198A (en) * | 2018-02-06 | 2018-08-17 | 平安科技（深圳）有限公司 | Man-machine identification model establishes device, method and computer readable storage medium |
| CN110768980A (en) * | 2019-10-22 | 2020-02-07 | 武汉极意网络科技有限公司 | Network man-machine verification method, device, equipment and storage medium |
| CN111782068A (en) * | 2019-04-04 | 2020-10-16 | 阿里巴巴集团控股有限公司 | Method, device and system for generating mouse track and data processing method |
| CN112905967A (en) * | 2021-03-05 | 2021-06-04 | 杭州费尔斯通科技有限公司 | Identification method of slider verification code |
| US20240028673A1 (en) * | 2022-07-15 | 2024-01-25 | Nvidia Corporation | Robust trajectory predictions against adversarial attacks in autonomous machines and applications |
-
2025
- 2025-05-15 CN CN202510622112.9A patent/CN120144423B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108416198A (en) * | 2018-02-06 | 2018-08-17 | 平安科技（深圳）有限公司 | Man-machine identification model establishes device, method and computer readable storage medium |
| CN111782068A (en) * | 2019-04-04 | 2020-10-16 | 阿里巴巴集团控股有限公司 | Method, device and system for generating mouse track and data processing method |
| CN110768980A (en) * | 2019-10-22 | 2020-02-07 | 武汉极意网络科技有限公司 | Network man-machine verification method, device, equipment and storage medium |
| CN112905967A (en) * | 2021-03-05 | 2021-06-04 | 杭州费尔斯通科技有限公司 | Identification method of slider verification code |
| US20240028673A1 (en) * | 2022-07-15 | 2024-01-25 | Nvidia Corporation | Robust trajectory predictions against adversarial attacks in autonomous machines and applications |
Also Published As
| Publication number | Publication date |
|---|---|
| CN120144423B (en) | 2025-08-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111340008B (en) | Method and system for generation of counterpatch, training of detection model and defense of counterpatch | |
| CN116561533B (en) | Emotion evolution method and terminal for virtual avatar in educational element universe | |
| CN109508687A (en) | Man-machine interaction control method, device, storage medium and smart machine | |
| CN105426827A (en) | Living body verification method, device and system | |
| JP7412496B2 (en) | Living body (liveness) detection verification method, living body detection verification system, recording medium, and training method for living body detection verification system | |
| CN108447159B (en) | Face image acquisition method and device and entrance and exit management system | |
| CN105518708A (en) | Method and equipment for verifying living human face, and computer program product | |
| CN111178170B (en) | A gesture recognition method and an electronic device | |
| CN108537702A (en) | Foreign language teaching evaluation information generation method and device | |
| CN110524559B (en) | Intelligent man-machine interaction system and method based on personnel behavior data | |
| CN107622252A (en) | information generating method and device | |
| CN119687918A (en) | Navigation obstacle avoidance method and system in low confidence and feature similarity environment | |
| CN110465089B (en) | Map exploration method, map exploration device, map exploration medium and electronic equipment based on image recognition | |
| Galicia et al. | Mexican sign language recognition using movement sensor | |
| KR20200009599A (en) | A method, server and program for apply artist's style to image and provide it | |
| CN119785165B (en) | Multi-mode interaction and image processing method for AI glasses, electronic device and storage medium | |
| KR20230093191A (en) | Method for recognizing joint by error type, server | |
| CN120144423B (en) | Detection passing system and method for robot verification browser | |
| CN107578015A (en) | A kind of first impression identification and feedback system and method based on deep learning | |
| CN117133277A (en) | Virtual character expression control method and system for man-machine interaction | |
| CN113158920B (en) | Training method and device for specific action recognition model and computer equipment | |
| CN113033747B (en) | Figure identification code generation method for man-machine identification | |
| CN108460313A (en) | A kind of gesture identifying device and human-computer interaction system | |
| CN116189304B (en) | Intelligent teaching system based on AI vision technology | |
| CN116881651B (en) | Training method and device for characterization model, and man-machine identification method and device |
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 |