Cheating detection method and system for answering behaviors of online examination

By constructing a baseline model of behavior and biometric verification, the problem of insufficient real-time and intelligence of cheating detection in computer-based examinations is solved, and the accurate identification of cheating behavior is achieved to ensure the fairness and impartiality of the examination.

CN119026062AInactive Publication Date: 2024-11-26ATA ONLINE (BEIJING) EDUCATION TECH LTD

Patent Information

Application Number
CN202411499147.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-25
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The cheat detection technology for existing computer-based examinations is insufficient in real time, and the intelligence level is limited, making it difficult to accurately identify cheating behaviors, resulting in false alarms and missed reports, affecting the fairness and fairness of the examination.

Method used

By continuously monitoring the candidates' operating behavior, extract the characteristics of keyboard typing mode, mouse action mode, screen browsing mode and application interaction mode, build a behavior baseline model, and use unsupervised learning technology and Bayesian algorithm to evaluate whether the behavior deviates from normal in real time. range and verify with biometric information.

Benefits of technology

It realizes efficient and accurate detection of cheating behaviors in computer-based exams, reduces false alarms and missed reports, and ensures the fairness and impartiality of the exam.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026062A_ABST
    Figure CN119026062A_ABST
Patent Text Reader

Abstract

The invention provides a cheating detection method and system for an answering behavior of a computer examination, and the method comprises the steps: continuously monitoring the operation behavior of an examinee in an examination process; extracting operation behavior characteristics by using the operation behaviors; constructing a behavior baseline model based on the operation behavior characteristics; the behavior baseline model is utilized to evaluate whether the behavior of the current examinee deviates from the normal range in real time, and the cheating behavior of the current examinee is determined when the deviation degree exceeds the preset threshold value, so that the cheating behavior in the online examination process is efficiently and accurately detected, and a fair and just examination environment can be created.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of computer-based examinations, and in particular to a method and system for detecting cheating in answering questions in computer-based examinations. Background Art

[0002] With the development of information technology, computer-based examinations have been widely used in various examination scenarios due to their efficiency and convenience.

[0003] Computer-based exams are widely used in various standardized tests, such as professional qualification certification, academic education exams, language proficiency tests, etc. These exams often involve a large number of objective questions and some subjective questions, requiring candidates to complete the answers within a specified time.

[0004] With the increasing demand for remote examinations and online proctoring, examination organizers urgently need a set of effective technical means to monitor the behavior of examinees to prevent cheating. The technical means must be able to monitor the operation behavior of examinees in real time and accurately identify potential cheating behaviors to ensure the authenticity and fairness of the examination results.

[0005] At present, many examination platforms have adopted methods such as camera monitoring, face recognition, and dual-camera proctoring to prevent cheating. However, these methods mainly focus on physical monitoring, and it is difficult to conduct detailed analysis and evaluation of the specific operating behaviors of candidates on the computer.

[0006] In addition, some systems also try to conduct post-analysis by recording the candidates' operation logs, but this method cannot detect cheating behavior in real time, and lacks sufficient intelligent means in data analysis, making it difficult to distinguish between normal operations and potential cheating behavior.

[0007] The existing solutions have the following defects: Lack of real-time performance: Most existing proctoring solutions rely on post-event analysis and cannot detect and prevent cheating in real time during the exam. This results in the inability to take immediate action even if cheating is discovered.

[0008] Limited level of intelligence: Although some systems can record the operating behaviors of candidates, they lack effective algorithm models to distinguish normal operations from cheating behaviors, and cannot build a behavioral baseline model based on behavioral characteristics, making it difficult to accurately determine whether candidates are suspected of cheating.

[0009] False positive and false negative issues: Due to the lack of sophisticated behavioral feature extraction and analysis mechanisms, existing solutions are prone to false positives (mistaking normal operations for cheating) or false negatives (failure to identify real cheating behavior), which have a negative impact on the fairness of the exam and the psychology of the examinees. Summary of the invention

[0010] The embodiments of the present application provide a cheating detection method and system for computer-based examination answering behavior, which are used to solve the problems existing in the above-mentioned background technology.

[0011] In a first aspect, an embodiment of the present application provides a cheating detection method for answering questions in a computer-based examination, comprising: during the examination, continuously monitoring the examinee's current operating behavior, wherein the operating behavior includes: keyboard tapping mode, mouse movement mode, screen browsing mode and application interaction mode; using the operating behavior to extract operating behavior features, the operating behavior features include: time interval distribution of keyboard tapping, mouse movement speed, click frequency, browsing hot spots and application switching frequency; based on the operating behavior features, constructing a behavior baseline model, wherein the behavior baseline model is a model based on historical data statistics of normal operating behavior for distinguishing normal operation from potential cheating behavior; using the behavior baseline model, real-time evaluation of whether the examinee's behavior deviates from the normal range, and when the degree of deviation exceeds a predetermined threshold, determining that the examinee has cheated.

[0012] Furthermore, after determining that the candidate has cheated, the method also includes: obtaining the candidate's biometric information, the biometric information including: facial image and voice sample; comparing and verifying the candidate's biometric information with the biometric information provided when registering for the exam; if the verification fails, marking the candidate as high risk and having the invigilator conduct a manual review.

[0013] Furthermore, based on the operational behavior characteristics, a behavioral baseline model is constructed, including: using unsupervised learning technology to learn the pattern of normal operational behavior from historical operational behaviors; extracting normal operational behavior characteristics from the pattern of normal operational behavior, and converting the normal operational behavior characteristics into a distribution in a feature space to obtain the behavioral baseline model.

[0014] Furthermore, unsupervised learning technology is used to learn patterns of normal operating behaviors from historical operating behaviors, including: preprocessing the historical operating behaviors to obtain preprocessed operating behaviors, wherein the preprocessing includes: data cleaning and normalization processing; using a feature engineering enhancement model to extract features and fuse features of the preprocessed operating behaviors to obtain feature data; and using the unsupervised learning technology to learn patterns of normal operating behaviors from the feature data.

[0015] Furthermore, the behavior baseline model is used to evaluate in real time whether the examinee's behavior deviates from the normal range, including: defining a behavior deviation measurement index; calculating the examinee's behavior deviation measurement index value based on the expression of the behavior deviation measurement index and the operation behavior characteristics; determining the probability that the operation behavior characteristics are normal operations based on the examinee's behavior deviation measurement index value and the Bayesian algorithm; and evaluating in real time whether the examinee's behavior deviates from the normal range based on the probability that the operation behavior characteristics are normal operations and a preset threshold; wherein the expression of the behavior deviation measurement index is: ; For the The importance weight of the operational behavior feature, is an index to adjust the sensitivity of the bias measure, For the Operational behavior characteristics, is the first The average value of the operational behavior characteristics, is the first The standard deviation of the operational behavior characteristics, is the total number of operational behavior characteristics.

[0016] Furthermore, based on the behavioral deviation measurement index value of the examinee and the Bayesian algorithm, the probability that the operation behavior feature is a normal operation is determined, including: calculating the prior probability of the normal operation behavior based on the historical operation behavior; calculating the marginal probability corresponding to the operation behavior feature; and calculating the probability that the operation behavior feature is a normal operation based on the prior probability, the marginal probability and the behavioral deviation measurement index value of the examinee.

[0017] Furthermore, the method further includes: configuring a dynamic weight for the operational behavior feature; improving the expression of the behavior deviation measurement index based on the dynamic weight and a preset multidimensional distance measurement algorithm to obtain an improved expression of the behavior deviation measurement index, wherein the preset multidimensional distance measurement algorithm includes a weighted Euclidean distance algorithm or a Mahalanobis distance algorithm; wherein the expression of the dynamic weight is: ; For the current moment, For the The occurrence time of the operational behavior characteristics, is the attenuation coefficient, For the The moment when the operation behavior feature was last recorded. is a natural constant, is the total number of operational behavior characteristics.

[0018] In the second aspect, an embodiment of the present application provides a cheating detection system for answering questions in a computer-based examination, comprising: a monitoring unit, for continuously monitoring the current operation behavior of the examinee during the examination, wherein the operation behavior includes: keyboard tapping mode, mouse movement mode, screen browsing mode and application interaction mode; an extraction unit, for extracting operation behavior features using the operation behavior, wherein the operation behavior features include: time interval distribution of keyboard tapping, mouse movement speed, click frequency, browsing hot spots and application switching frequency; a construction unit, for constructing a behavior baseline model based on the operation behavior features, wherein the behavior baseline model is a model based on historical data statistics of normal operation behavior for distinguishing normal operation from potential cheating behavior; a judgment unit, for using the behavior baseline model to evaluate in real time whether the examinee's behavior deviates from the normal range, and when the degree of deviation exceeds a predetermined threshold, determine that the examinee has cheated.

[0019] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the method described in the first aspect above.

[0020] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements the method described in the first aspect above.

[0021] In an embodiment of the present invention, the operating behavior of the examinee is continuously monitored during the examination; the operating behavior is used to extract operating behavior features, and the operating behavior features include: time interval distribution of keyboard tapping, mouse movement speed, click frequency, browsing hot spots, and application switching frequency; based on the operating behavior features, a behavioral baseline model is constructed; using the behavioral baseline model, it is evaluated in real time whether the current examinee's behavior deviates from the normal range, and when the degree of deviation exceeds a predetermined threshold, it is determined that the current examinee has cheated, thereby achieving efficient and accurate detection of cheating behavior during the computer-based examination, which helps to create a fair and just examination environment.

[0022] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 A flowchart of a method for detecting cheating in computer-based examination answering behavior provided in an embodiment of the present application; Figure 2 A schematic diagram of a cheating detection system for computer-based examination answering behavior provided in an embodiment of the present application; Figure 3 A schematic diagram of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Embodiment 1: According to an embodiment of the present invention, an embodiment of a cheating detection method for answering questions in a computer-based examination is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0027] Figure 1 FIG. 1 is a flow chart of a method for detecting cheating in computer-based examination answering behavior according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps: Step S102, during the test, continuously monitoring the examinee's current operation behavior, wherein the operation behavior includes: keyboard tapping mode, mouse action mode, screen browsing mode and application interaction mode; Step S104, extracting operation behavior features using the operation behavior, the operation behavior features including: time interval distribution of keyboard tapping, mouse movement speed, click frequency, browsing hot spots, and application switching frequency; Step S106, constructing a behavior baseline model based on the operation behavior characteristics, wherein the behavior baseline model is a model for distinguishing normal operations from potential cheating behaviors based on historical data statistics of normal operation behaviors; Step S108, using the behavior baseline model, evaluating in real time whether the examinee's behavior deviates from the normal range, and determining that the examinee has cheated when the degree of deviation exceeds a predetermined threshold.

[0028] In an embodiment of the present invention, the monitoring system needs to collect various operation behavior data of the examinee during the examination, including but not limited to keyboard tapping mode, mouse action mode, screen browsing mode and application interaction mode. Keyboard tapping mode refers to the way the examinee taps on the keyboard, including tapping frequency, interval time, etc.; mouse action mode refers to the action characteristics of the examinee when using the mouse, such as movement speed, click frequency, etc.; screen browsing mode refers to the way the examinee browses the screen, such as hot spots where the stay time is longer; application interaction mode refers to the interaction between the examinee and applications other than the examination software, such as the behavior of frequently switching to the browser or other applications to view materials.

[0029] Specifically, the monitoring system can be started after the examinee starts the exam, and the special software installed on the exam terminal can record every operation of the examinee. For example, the system records the timestamp of each keyboard stroke, the mouse movement path, the time of the mouse click, the gaze time of each area of ​​the screen, and whether the examinee has opened other applications. Suppose that within 10 minutes after the start of the exam, a candidate made 50 keyboard strokes, with an average interval of 0.5 seconds between each stroke, moved the mouse 100 times, clicked 20 times, with an average interval of 5 seconds between each click, and the examinee stayed in the lower right corner of the screen for a long time. In addition, the system also records that the examinee did not open any other applications during this period.

[0030] Extract key features from the large amount of collected operation behavior data to build a behavior model. These features may include the time interval distribution of keyboard tapping, mouse movement speed, click frequency, browsing hot spots, and application switching frequency. The purpose of feature extraction is to convert behavior into quantifiable indicators to facilitate subsequent analysis and modeling.

[0031] Specifically, from the data collected above, the distribution of time intervals for keyboard tapping can be extracted. For example, through statistical analysis, it is known that the intervals between candidates' keyboard tapping are mainly distributed between 0.4 seconds and 0.6 seconds, with an average interval of 0.5 seconds per tapping. The mouse movement speed can be obtained by dividing the total distance the mouse moves by the time, assuming that the candidate moves the mouse an average of 50 pixels per second. The click frequency can be obtained by counting the number of clicks per unit time, such as 20 clicks per minute. The browsing hotspot area can be determined by analyzing the screen area where the candidate's mouse stays the longest, for example, the candidate stays the longest in the lower right corner of the screen. The application switching frequency refers to the number of times the candidate switches to other applications, assuming that the candidate does not switch any applications during the exam.

[0032] Based on the extracted operational behavior features, a behavioral baseline model is constructed using historical data statistics. The behavioral baseline model is a model based on historical data statistics of normal operational behaviors, which can distinguish normal operations from potential cheating behaviors. Usually, the model is trained using machine learning or statistical methods to identify which behavioral features are consistent with the performance of candidates under normal test conditions.

[0033] Specifically, a behavioral baseline model is constructed based on the above features. For example, a classifier is trained to distinguish between normal operations and potential cheating behaviors. Assume that in the training data set, the keyboard tapping interval of the normal operating examinee is mainly distributed between 0.3 seconds and 0.7 seconds, with an average tapping interval of 0.5 seconds; the average mouse movement speed is 50 pixels per second; the average click frequency is 20 times per minute; the browsing hotspot area is concentrated in the center of the screen; and the application switching frequency is almost zero. The characteristics of the examinee with cheating behavior may be different, such as very short keyboard tapping intervals (such as less than 0.2 seconds) or very long (such as greater than 1 second), abnormally high click frequency (such as more than 50 times per minute), and frequent switching of applications (such as switching once a minute). By comparing the current examinee's operating characteristics with the characteristics in the baseline model, it is possible to evaluate whether the examinee is likely to have cheating behavior.

[0034] The constructed behavioral baseline model is used to conduct real-time evaluation of the examinee's operating behavior during the examination. If the detected operating behavior characteristics deviate from the normal range to a degree exceeding the predetermined threshold, the examinee is deemed to have cheated. The predetermined threshold here refers to a standard set in advance to distinguish the boundary between normal behavior and abnormal behavior.

[0035] Specifically, it evaluates in real time whether the examinee's behavior deviates from the normal range. For example, suppose that during the exam, the interval between a candidate's keyboard tapping suddenly becomes extremely short, with the average interval between each tapping reduced to less than 0.2 seconds, and the click frequency increases significantly, reaching more than 40 times per minute. At the same time, it is found that the candidate frequently switches applications, switching once every two minutes. Compared with the features in the previously constructed behavioral baseline model, these features clearly exceed the predetermined threshold. Therefore, the system can immediately issue a warning, prompting the invigilator to pay attention to the candidate, and may suspend his or her exam until it is confirmed that there is no cheating behavior.

[0036] In an embodiment of the present invention, when the system detects that the examinee's operation behavior deviates from the normal range and reaches a predetermined threshold, the system will automatically start a further verification procedure. The procedure first obtains the examinee's biometric information, which usually includes a facial image and a voice sample. The facial image can be used to confirm the examinee's identity, while the voice sample can assist in verifying whether the examinee's voice characteristics are consistent with those at the time of registration. At the same time, the microphone starts recording the examinee's voice sample. For example, the system requires the examinee to read a specified text in order to obtain a clear voice sample.

[0037] The acquired biometric information will be sent to the backend server for processing. The system will compare the candidate's current facial image and voice sample with the biometric information provided by the candidate when registering for the exam to verify whether the current candidate is the same person who registered for the exam. This process usually involves image processing technology and voice recognition technology.

[0038] The system transmits the captured facial image and recorded voice sample to the server for analysis. The server uses the Face Recognition Algorithm (FRA) to compare the current facial image with the facial image database saved during registration to check whether there are highly similar facial features. At the same time, the Voice Recognition Algorithm (VRA) is used to compare the current voice sample with the voice sample database saved during registration to confirm whether the sound features are consistent. For example, if the matching score (MS) calculated by the facial recognition algorithm is higher than a certain threshold (such as 0.8), and the similarity score (SS) obtained by the voice recognition algorithm is also higher than another threshold (such as 0.7), the biometric information verification is considered successful.

[0039] If the biometric information comparison result shows that the verification fails, that is, the current candidate's biometrics do not match the biometrics provided when registering for the exam, the system will automatically mark the candidate as a high-risk candidate. Subsequently, the system will notify the invigilator to conduct further manual review to confirm the authenticity of the candidate's identity.

[0040] If the match score calculated by the facial recognition algorithm is lower than 0.8, or the similarity score calculated by the voice recognition algorithm is lower than 0.7, the verification is considered to have failed. At this point, the system will mark the candidate as high risk in the candidate list and send an alert to the invigilator. After receiving the alert, the invigilator will conduct further identity verification on the candidate, which may include asking the candidate for personal information, collecting biometric information again for re-verification, etc. For example, the invigilator can ask the candidate to show his or her ID, and take a facial image and record a voice sample again to confirm the candidate's identity.

[0041] Through the above steps, the system can effectively initiate the biometric verification process in a timely manner when suspicious behavior is detected, and ensure the fairness and authenticity of the exam through manual review.

[0042] In an embodiment of the present invention, based on the operation behavior characteristics, a behavior baseline model is constructed, including: Use unsupervised learning techniques to learn patterns of normal operating behavior from historical operating behaviors; Normal operation behavior features are extracted from the pattern of the normal operation behavior, and the normal operation behavior features are converted into distribution in a feature space to obtain the behavior baseline model.

[0043] In an embodiment of the present invention, historical operation behavior data is collected, including operation records of multiple normal examinees during the examination. These records contain information such as the keyboard tapping time interval, mouse movement speed, click frequency, browsing hot spots, and application switching frequency of the examinees during the examination. For example, historical data of 100 examinees are collected, and each examinee generates about 100,000 operation records during the examination.

[0044] Extract operational behavior features from the collected historical data. For example, count the distribution of each examinee's keyboard tapping time interval, mouse movement speed, click frequency, browsing hotspot area, and application switching frequency. Assume that the keyboard tapping time interval of a examinee is mainly concentrated between 0.3 seconds and 0.7 seconds, with an average tapping interval of 0.5 seconds; the mouse movement speed is about 50 pixels per second; the click frequency is 20 times per minute; the browsing hotspot area is mainly concentrated in the center of the screen; and the application switching frequency is once per hour.

[0045] Use unsupervised learning techniques, such as the K-means Clustering (KMC) algorithm in Cluster Analysis, to analyze the extracted operation behavior features and find out the pattern of normal operation behavior. The K-means algorithm is a commonly used clustering algorithm that divides the data into K clusters through an iterative process to minimize the distance between the data in the cluster. Suppose the KMC algorithm is used to divide the historical data into 5 clusters, each cluster representing a common operation mode. For example, the first cluster may represent that the keyboard tapping time interval is distributed between 0.3 seconds and 0.7 seconds, and the second cluster represents the mouse movement speed distributed at about 50 pixels per second.

[0046] The normal operation behavior features extracted above are converted into distributions in the feature space. For example, the keyboard tapping time interval distribution is converted into a probability density function PDF in the feature space, and the frequency of time interval occurrence is represented by a histogram; the mouse movement speed distribution is converted into a cumulative distribution function CDF in the feature space, which displays the cumulative frequency of different speed levels.

[0047] According to the converted feature space distribution, a behavioral baseline model is constructed. For example, the distribution of all normal operation behavior features is summarized into a comprehensive model, which reflects the operation behavior pattern of normal candidates. When new candidate operation behavior features enter the system, they can be compared with the distribution in the baseline model to determine whether there is any deviation from the normal range.

[0048] When evaluating the examinee's behavior in real time, it is assumed that the examinee's keyboard tapping interval is 0.4 seconds, the mouse movement speed is 55 pixels / second, the click frequency is 25 times / minute, the browsing hotspot area is concentrated in the upper left corner of the screen, and the application switching frequency is 2 times per hour. By comparing with the distribution in the baseline model, it can be found that the browsing hotspot area and application switching frequency are significantly deviated from the normal range, indicating possible cheating behavior.

[0049] In an embodiment of the present invention, unsupervised learning technology is used to learn the pattern of normal operation behavior from historical operation behaviors, including: Preprocessing the historical operation behavior to obtain preprocessed operation behavior, wherein the preprocessing includes: data cleaning and normalization processing; Using a feature engineering enhancement model to perform feature extraction and feature fusion on the preprocessing operation behavior to obtain feature data; The unsupervised learning technique is used to learn patterns of normal operating behaviors from the feature data.

[0050] In an embodiment of the present invention, the historical operation behavior data includes operation records of multiple examinees, for example, information such as each examinee's keyboard tapping time interval, mouse movement speed, click frequency, browsing hot spots, and application switching frequency.

[0051] Data cleaning steps: delete or fill in missing values, remove outliers (such as data points with a keyboard tapping time interval of less than 0.1 seconds or greater than 2 seconds) to ensure the rationality of the data.

[0052] Normalization step: Use the maximum and minimum normalization method to scale all features to the [0, 1] interval.

[0053] A variety of features are extracted from the preprocessed data, such as the time interval distribution of keyboard tapping, the mouse movement speed distribution, the click frequency distribution, the browsing hotspot area distribution, and the application switching frequency distribution.

[0054] The extracted multiple features can be fused, for example, the keyboard tapping time interval distribution and the mouse movement speed distribution can be combined into a new composite feature to better describe the examinee's operating behavior pattern.

[0055] Use unsupervised learning techniques (such as K-Means Clustering [KMC]) to learn patterns of normal operating behavior from feature data.

[0056] The K-means clustering algorithm divides the feature data into multiple clusters, each of which represents an operation mode. For example, the feature data is divided into five clusters, each of which represents a group of similar operation behavior modes.

[0057] For example, the K-means algorithm is used to divide the feature data into five clusters, each cluster representing a common operation mode. For example, the first cluster may represent that the time interval between keyboard tapping is distributed between 0.3 seconds and 0.7 seconds, and the second cluster represents that the mouse movement speed is distributed around 50 pixels per second.

[0058] In the embodiment of the present invention, the behavior baseline model is used to evaluate in real time whether the examinee's behavior deviates from the normal range, including: Define behavioral deviation metrics; Calculating the behavior deviation measurement index value of the examinee based on the expression of the behavior deviation measurement index and the operation behavior characteristics; Based on the behavioral deviation measurement index value of the examinee and the Bayesian algorithm, determining the probability that the operation behavior feature is a normal operation; Based on the probability that the operation behavior characteristics are normal operations and a preset threshold, real-time evaluation is performed to determine whether the examinee's behavior deviates from a normal range; The expression of the behavior deviation measurement index is: ; For the The importance weight of the operational behavior feature, is an index to adjust the sensitivity of the bias measure, For the Operational behavior characteristics, is the first The average value of the operational behavior characteristics, is the first The standard deviation of the operational behavior characteristics, is the total number of operational behavior characteristics.

[0059] In the embodiment of the present invention, the behavior deviation measurement index is a quantitative index used to measure the difference between the examinee's current operation behavior and the historical normal operation behavior. The index evaluates the normality of the examinee's operation behavior by calculating the difference between the examinee's various operation behavior features and the corresponding features in the baseline model.

[0060] When evaluating candidate behavior in real time, for example, the candidate's operational behavior characteristics are as follows: Keyboard tapping interval: 0.4 seconds; Mouse movement speed: 55 pixels / second; Click frequency: 25 times / minute; Browse hot spots: focus on the upper left corner of the screen; Application switching frequency: 2 times per hour.

[0061] The statistics in the behavioral baseline model are as follows: The average time interval between keyboard tapping is 0.5 seconds, and the standard deviation is 0.1 seconds. The average mouse movement speed is 50 pixels / second, and the standard deviation is 10 pixels / second. Click frequency average: 20 times / minute, standard deviation: 5 times / minute; Browsing hot spots are mainly concentrated in the center of the screen; The average application switching frequency is 1 time per hour, with a standard deviation of 0.5 times per hour.

[0062] The behavior deviation measurement index values ​​of each operation behavior feature are calculated by the formula as follows: D1=0.3 (interval between keyboard tapping); D2=0.04 (mouse movement speed); D3=0.25 (click frequency); D4=0.1 (browse hotspot area); D5=1.0 (application switching frequency). The probability that each operation behavior feature is a normal operation is calculated based on the Bayesian algorithm: P1=0.95 (interval between keyboard tapping); P2=0.96 (mouse movement speed); P3=0.9 (click frequency); P4=0.95 (browsing hot spots); P5 = 0.75 (application switching frequency).

[0063] Based on the preset threshold (assuming it is 0.9), assess whether the candidate's behavior deviates from the normal range: Keyboard tapping interval: normal; Mouse movement speed: normal; Click frequency: normal; Browsing hot spots: normal; Application switching frequency: deviates from the normal range.

[0064] According to the above assessment, the examinee's behavior is within the normal range in terms of keyboard tapping interval, mouse movement speed, click frequency, and browsing hot spots, but there is a possibility that the application switching frequency deviates from the normal range. Therefore, the system can take further measures, such as marking it as suspected cheating behavior and reminding the invigilator to conduct manual review.

[0065] In the embodiment of the present invention, based on the behavioral deviation measurement index value of the examinee and the Bayesian algorithm, determining the probability that the operation behavior feature is a normal operation includes: Based on the historical operating behavior, calculating the prior probability of the normal operating behavior; Calculating the marginal probability corresponding to the operation behavior feature; Based on the prior probability, the marginal probability and the behavioral deviation measurement index value of the examinee, the probability that the operation behavior feature is a normal operation is calculated.

[0066] In the embodiment of the present invention, it can be assumed that the behavior deviation metric D follows a certain distribution (such as a normal distribution), and it can be estimated based on D , that is, the probability that the operation behavior feature is a normal operation.

[0067] For example, assuming that D follows a normal distribution , the cumulative distribution function (CDF) can be used to estimate ,in, , is the cumulative distribution function of the standard normal distribution. Then, we use the Bayesian formula to calculate the probability that the current behavior belongs to normal operation. , ,in, The prior probability of normal operating behavior, is the marginal probability.

[0068] in, ; , is the probability of the current behavior feature occurring under abnormal behavior conditions.

[0069] In summary .

[0070] In an embodiment of the present invention, the method further includes: configuring dynamic weights for the operational behavior characteristics; Based on the dynamic weight and the preset multidimensional distance measurement algorithm, the expression of the behavior deviation measurement index is improved to obtain an improved expression of the behavior deviation measurement index, wherein the preset multidimensional distance measurement algorithm includes a weighted Euclidean distance algorithm or a Mahalanobis distance algorithm; The expression of the dynamic weight is: ; For the current moment, For the The occurrence time of the operational behavior characteristics, is the attenuation coefficient, For the The moment when the operation behavior feature was last recorded. is a natural constant, is the total number of operational behavior characteristics.

[0071] In the embodiment of the present invention, by configuring dynamic weights for the operational behavior features, the system can more accurately reflect the recent behavioral features of the examinee, because the recent behavioral features are more likely to represent the examinee's current operational mode. This method can effectively improve the accuracy of detecting cheating behavior and avoid false positives caused by the improper influence of past behavioral features on current behavior.

[0072] The use of dynamic weights allows the system to make a personalized assessment of each candidate's behavior pattern. This means that each candidate's behavior is considered independently rather than simply compared with the group average. This personalized approach can better adapt to the differences between individuals and reduce misjudgments caused by individual differences.

[0073] Embodiment 2: An embodiment of the present invention also provides a cheating detection system for answering questions in a computer-based examination. The cheating detection system for answering questions in a computer-based examination is used to execute the cheating detection method for answering questions in a computer-based examination provided in the above content of the embodiment of the present invention. The following is a specific introduction to the cheating detection system for answering questions in a computer-based examination provided in an embodiment of the present invention.

[0074] like Figure 2 As shown, Figure 2 The schematic diagram of the above-mentioned cheating detection system for computer-based examination answering behavior is as follows: The monitoring unit 10 is used to continuously monitor the current operation behavior of the examinee during the examination, wherein the operation behavior includes: keyboard tapping mode, mouse action mode, screen browsing mode and application interaction mode; An extraction unit 20 is used to extract operation behavior features using the operation behavior, wherein the operation behavior features include: time interval distribution of keyboard tapping, mouse movement speed, click frequency, browsing hot spots, and application switching frequency; A construction unit 30 is used to construct a behavior baseline model based on the operation behavior characteristics, wherein the behavior baseline model is a model based on historical data statistics of normal operation behaviors and used to distinguish normal operations from potential cheating behaviors; The determination unit 40 is used to use the behavior baseline model to evaluate in real time whether the examinee's behavior deviates from the normal range, and determine that the examinee has cheated when the degree of deviation exceeds a predetermined threshold.

[0075] Embodiment three: The embodiment of the present invention further provides a computing device for executing a program of the method described in the first embodiment, such as Figure 3 As shown, the computing device may include a storage component 41 and a processing component 42; The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 42 .

[0076] The processing component 42 may include one or more processors to execute computer instructions to complete all or part of the steps in the method of embodiment 1. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0077] The storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0078] The display component 43 may be an electroluminescent (EL) element, a liquid crystal display or a micro display having a similar structure, or a retinal direct display or a similar laser scanning display.

[0079] Of course, a computing device may also include other components, such as input / output interfaces, communication components, etc.

[0080] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0081] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0082] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0083] Embodiment 4: The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The method of the illustrated embodiment.

[0084] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0085] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0086] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A cheating detection method for answering questions in a computer-based test. It is characterized in that include: During the examination, the examinee's current operation behavior is continuously monitored, where the operation behavior includes: keyboard tapping mode, mouse movement mode, screen browsing mode, and application interaction mode; Extracting operation behavior features using the operation behavior, the operation behavior features including: time interval distribution of keyboard tapping, mouse movement speed, click frequency, browsing hot spots, and application switching frequency; Based on the operation behavior characteristics, a behavior baseline model is constructed, wherein the behavior baseline model is a model based on historical data statistics of normal operation behaviors and used to distinguish normal operations from potential cheating behaviors; Using the behavior baseline model, assess in real time whether the examinee's behavior deviates from the normal range, and if the degree of deviation exceeds a predetermined threshold, determine that the examinee has cheated; The behavior baseline model is used to evaluate in real time whether the examinee's behavior deviates from the normal range, including: Define behavioral deviation metrics; Calculating the behavior deviation measurement index value of the examinee based on the expression of the behavior deviation measurement index and the operation behavior characteristics; Based on the behavioral deviation measurement index value of the examinee and the Bayesian algorithm, determining the probability that the operation behavior feature is a normal operation; Based on the probability that the operational behavior characteristics are normal operations and a preset threshold, a real-time assessment is made as to whether the examinee's behavior deviates from the normal range.

2. The method according to claim 1, It is characterized in that After determining that the examinee has cheated, the method further includes: Obtaining biometric information of the examinee, the biometric information including: facial image and voice sample; Comparing and verifying the candidate's biometric information with the biometric information provided when registering for the exam; If the verification fails, the candidate will be marked as high risk and will be manually reviewed by the proctor.

3. The method according to claim 1, It is characterized in that Based on the operational behavior characteristics, a behavior baseline model is constructed, including: Use unsupervised learning techniques to learn patterns of normal operating behavior from historical operating behaviors; Normal operation behavior features are extracted from the pattern of the normal operation behavior, and the normal operation behavior features are converted into distribution in a feature space to obtain the behavior baseline model.

4. The method according to claim 3, It is characterized in that Use unsupervised learning techniques to learn patterns of normal operating behavior from historical operating behaviors, including: Preprocessing the historical operation behavior to obtain preprocessed operation behavior, wherein the preprocessing includes: data cleaning and normalization processing; Using a feature engineering enhancement model to perform feature extraction and feature fusion on the preprocessing operation behavior to obtain feature data; The unsupervised learning technique is used to learn patterns of normal operating behaviors from the feature data.

5. The method according to claim 1, It is characterized in that in, The expression of the behavior deviation metric is: ; For the The importance weight of the operational behavior feature, is an index to adjust the sensitivity of the bias measure, For the Operational behavior characteristics, is the first The average value of the operational behavior characteristics, is the first The standard deviation of the operational behavior characteristics, is the total number of operational behavior characteristics.

6. The method according to claim 5, It is characterized in that Based on the behavioral deviation measurement index value of the examinee and the Bayesian algorithm, the probability that the operation behavior feature is a normal operation is determined, including: Based on the historical operating behavior, calculating the prior probability of the normal operating behavior; Calculating the marginal probability corresponding to the operation behavior feature; Based on the prior probability, the marginal probability and the behavioral deviation measurement index value of the examinee, the probability that the operation behavior feature is a normal operation is calculated.

7. The method according to claim 5, It is characterized in that The method further comprises: Configuring dynamic weights for the operational behavior characteristics; Based on the dynamic weight and the preset multidimensional distance measurement algorithm, the expression of the behavior deviation measurement index is improved to obtain an improved expression of the behavior deviation measurement index, wherein the preset multidimensional distance measurement algorithm includes a weighted Euclidean distance algorithm or a Mahalanobis distance algorithm; The expression of the dynamic weight is: ; For the current moment, For the The occurrence time of the operational behavior characteristics, is the attenuation coefficient, For the The moment when the operation behavior feature was last recorded. is a natural constant, is the total number of operation behavior characteristics.

8. A cheating detection system for answering questions in a computer-based exam. It is characterized in that include: The monitoring unit is used to continuously monitor the examinee's current operation behavior during the examination, wherein the operation behavior includes: keyboard tapping mode, mouse action mode, screen browsing mode and application interaction mode; An extraction unit, configured to extract operation behavior features using the operation behavior, wherein the operation behavior features include: time interval distribution of keyboard tapping, mouse movement speed, click frequency, browsing hot spots, and application switching frequency; A construction unit, configured to construct a behavior baseline model based on the operation behavior characteristics, wherein the behavior baseline model is a model for distinguishing normal operation from potential cheating behavior based on historical data statistics of normal operation behavior; The determination unit is used to use the behavior baseline model to evaluate in real time whether the examinee's behavior deviates from the normal range, and determine that the examinee has cheated when the degree of deviation exceeds a predetermined threshold.

9. A computing device, It is characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the method according to any one of claims 1 to 7.

10. A computer storage medium, It is characterized in that A computer program is stored, and when the computer program is executed by a computer, the method according to any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • Detection method and corresponding apparatus for cheating behavior of mobile terminal

    CN106096996A

  • Online examination system

    CN112686311A

  • Online examination cheating identification method, electronic equipment and storage medium

    CN117058612A

  • Online examination intelligent management system and method

    CN117711432A

  • Invigilation management system and method based on information security

    CN118277983A

Cited By

  • Personal health abnormal behavior monitoring system fused with unsupervised learning

    CN120277542A