Artificial intelligence-based gaze track anti-fraud method, device, equipment and medium

Abstract

The application provides a gaze track anti-fraud method and device based on artificial intelligence, an electronic device and a storage medium. The gaze track anti-fraud method based on artificial intelligence comprises the following steps: acquiring gaze video data to generate a gaze sample track set; segmenting the gaze sample track set to obtain a track initial segment set; screening the track initial segment set to obtain a positive track effective segment set and a negative track effective segment set; constructing a gaze track fraud identifier based on the positive track effective segment set and the negative track effective segment set; performing anti-fraud detection on a target gaze track to be detected based on the gaze track fraud identifier to obtain a detection result; and updating the gaze track fraud identifier based on the detection result to improve the detection accuracy of the gaze track fraud identifier. The gaze track fraud identifier is generated, and the accuracy of identifying the fraud risk of the gaze track is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a gaze track anti-fraud method and device based on artificial intelligence, an electronic device, and a storage medium. BACKGROUND

[0002] Fraud risk identification is a long-standing and difficult risk control direction in the financial field. In traditional fraud risk identification, it is generally started from the correlation analysis of the customer's text, browsing records, device IP, etc. However, these methods have certain time delay and are easy to be broken, resulting in low accuracy of fraud risk identification.

[0003] In various researches and actual applications of anti-fraud, there are very few through visual identification of fraud risks. In related applications, micro-expression is an attempt based on video images, but the effect is always unstable and not accurate enough. Therefore, how to effectively identify the fraud risks existing in the gaze track through visual technology becomes particularly important. SUMMARY

[0004] In view of the above, it is necessary to propose a gaze track anti-fraud method and device based on artificial intelligence, an electronic device, and a storage medium, to solve the technical problem of how to improve the accuracy of fraud risk identification in the field of financial lending.

[0005] The present application provides a gaze track anti-fraud method based on artificial intelligence, which comprises:

[0006] Obtaining gaze video data to generate a gaze sample track set, the gaze sample track set comprising a gaze positive sample track set and a gaze negative sample track set;

[0007] Segmenting all gaze sample tracks in the gaze sample track set to obtain a track initial segment set, the track initial segment set comprising a positive track initial segment set and a negative track initial segment set;

[0008] Screening the track initial segment set to obtain a positive track effective segment set and a negative track effective segment set;

[0009] Constructing a gaze track fraud identifier based on the positive track effective segment set and the negative track effective segment set;

[0010] Performing anti-fraud detection on a target gaze track to be detected based on the gaze track fraud identifier to obtain a detection result;

[0011] In some embodiments, the obtaining of the video data to generate the gaze sample track set comprises:

[0012] Collecting gaze video data;

[0013] marking all eye video data according to a preset manner, and taking all eye video data marked with positive sample as an eye positive sample data set and all eye video data marked with negative sample as an eye negative sample data set;

[0014] generating eye positive sample tracks based on eye video data in the eye positive sample data set, and taking all eye positive sample tracks as an eye positive sample track set;

[0015] generating eye negative sample tracks based on eye video data in the eye negative sample data set, and taking all eye negative sample tracks as an eye negative sample track set;

[0016] taking the eye positive sample track set and the eye negative sample track set as an eye sample track set.

[0017] In some embodiments, the splitting all eye sample tracks in the eye sample track set to obtain a track initial segment set comprises:

[0018] extracting low-frequency information of each eye sample track in the eye sample track set to obtain an eye sample low-frequency track;

[0019] calculating extreme points of the eye sample low-frequency track, and splitting the eye sample low-frequency track according to all extreme points to obtain a plurality of track initial segments of the eye sample low-frequency track;

[0020] respectively taking all track initial segments corresponding to each eye sample track in the eye positive sample track set and the eye negative sample track set as a positive track initial segment set and a negative track initial segment set;

[0021] taking the positive track initial segment set and the negative track initial segment set as a track initial segment set.

[0022] In some embodiments, the screening the track initial segment set to obtain a positive track effective segment set and a negative track effective segment set comprises:

[0023] calculating a segment similarity between a target positive track initial segment and each positive track initial segment in the positive track initial segment set, the target positive track initial segment being any one in the positive track initial segment set;

[0024] obtaining a positive track similar segment set corresponding to the target positive track initial segment based on a preset threshold and the segment similarity;

[0025] iterating each positive track initial segment in the positive track initial segment set to obtain a positive track similar segment set corresponding to each positive track initial segment;

[0026] Calculate the segment similarity between the initial segment of the target negative trajectory and each initial segment of the negative trajectory in the set of initial segments of the negative trajectory, wherein the initial segment of the target negative trajectory is any one of the initial segments of the negative trajectory;

[0027] Based on a preset threshold and the segment similarity, a set of negative trajectory similar segments corresponding to the initial segment of the target negative trajectory is obtained;

[0028] By traversing each initial negative trajectory segment in the set of initial negative trajectory segments, a set of similar negative trajectory segments corresponding to each initial negative trajectory segment is obtained;

[0029] The effective segments of positive trajectories and the effective segments of negative trajectories are obtained by filtering all sets of similar segments of positive trajectories and all sets of similar segments of negative trajectories respectively.

[0030] In some embodiments, the step of filtering all positive trajectory similarity fragment sets and all negative trajectory similarity fragment sets to obtain a positive trajectory valid fragment set and a negative trajectory valid fragment set includes:

[0031] Count the number of segments included in each positive trajectory similarity segment set and each negative trajectory similarity segment set;

[0032] Sort all positive trajectory similarity fragment sets and all negative trajectory similarity fragment sets in descending order of the number of fragments to obtain positive trajectory sorting results and negative trajectory sorting results;

[0033] Based on the positive trajectory sorting results and negative trajectory sorting results, all positive trajectory similar segment sets and all negative trajectory similar segment sets are filtered to obtain the same number of multiple positive trajectory valid segment sets and negative trajectory valid segment sets.

[0034] In some embodiments, constructing a gaze trajectory fraud detector based on the set of valid positive trajectory segments and the set of valid negative trajectory segments includes:

[0035] Extract the longest common sequence of all segments in the set of valid segments for each positive trajectory;

[0036] The longest common sequence of all valid segments of the positive trajectory is weighted and summed to obtain the positive trajectory reference template;

[0037] Extract the longest common sequence of all segments in the set of valid segments for each negative trajectory;

[0038] The longest common sequence of all valid segments of the negative trajectory is weighted and summed to obtain the negative trajectory reference template;

[0039] The positive trajectory reference template and the negative trajectory reference template are used as eye trajectory fraud detectors.

[0040] In some embodiments, obtaining detection results by performing anti-fraud detection on the target's eye trajectory based on the eye trajectory fraud detector includes:

[0041] The similarity between the target's gaze trajectory to be detected and the positive trajectory reference template and the negative trajectory reference template in the gaze trajectory fraud detector is calculated to obtain the positive trajectory similarity and the negative trajectory similarity.

[0042] If the positive trajectory similarity is greater than the negative trajectory similarity, then the target gaze trajectory to be detected has a fraud risk, and the detection result is a high risk of fraud.

[0043] If the positive trajectory similarity is not greater than the negative trajectory similarity, then the target gaze trajectory to be detected does not pose a fraud risk, and the detection result is low risk of fraud.

[0044] This application also provides an artificial intelligence-based eye-tracking anti-fraud device, the device comprising:

[0045] An acquisition unit is used to acquire eye gaze video data to generate an eye gaze sample trajectory set, the eye gaze sample trajectory set including an eye gaze positive sample trajectory set and an eye gaze negative sample trajectory set.

[0046] The segmentation unit is used to segment all gaze sample trajectories in the gaze sample trajectory set to obtain an initial trajectory fragment set, which includes a positive trajectory initial fragment set and a negative trajectory initial fragment set.

[0047] The filtering unit is used to filter the initial trajectory segment set to obtain a set of valid positive trajectory segments and a set of valid negative trajectory segments;

[0048] The construction unit is used to construct a gaze trajectory fraud detector based on the set of valid positive trajectory segments and the set of valid negative trajectory segments;

[0049] The detection unit is used to perform anti-fraud detection on the target eye trajectory to be detected based on the eye trajectory fraud detector to obtain the detection result.

[0050] This application embodiment also provides an electronic device, the electronic device comprising:

[0051] Memory, storing at least one instruction;

[0052] The processor executes the instructions stored in the memory to implement the AI-based eye-tracking anti-fraud method.

[0053] This application also provides a computer-readable storage medium storing at least one instruction, which is executed by a processor in an electronic device to implement the artificial intelligence-based eye-tracking anti-fraud method.

[0054] This application extracts positive and negative gaze sample trajectories from the acquired gaze video data, and combines the positive and negative gaze sample trajectories after segmentation, thereby obtaining a gaze trajectory fraud detector that can detect whether the gaze trajectory contains fraudulent information, effectively improving the accuracy of identifying fraud risks in time-series gaze trajectories. Attached Figure Description

[0055] Figure 1 This is a flowchart of a preferred embodiment of the artificial intelligence-based eye trajectory anti-fraud method involved in this application.

[0056] Figure 2 This is a functional block diagram of a preferred embodiment of the artificial intelligence-based eye-tracking anti-fraud device involved in this application.

[0057] Figure 3 This is a schematic diagram of the structure of an electronic device that is a preferred embodiment of the artificial intelligence-based eye trajectory anti-fraud method involved in this application.

[0058] Figure 4 This is a schematic diagram of a gaze trajectory involved in this application.

[0059] Figure 5 This is a computational diagram of the dynamic time warping algorithm involved in this application. Detailed Implementation

[0060] To better understand the purpose, features, and advantages of this application, a detailed description of the application is provided below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of this application can be combined with each other. Numerous specific details are set forth in the following description to provide a thorough understanding of this application; the described embodiments are only a part of the embodiments of this application, and not all of them.

[0061] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0063] This application provides an artificial intelligence-based eye-tracking anti-fraud method, which can be applied to one or more electronic devices. An electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions. Its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0064] Electronic devices can be any electronic product that allows human-computer interaction with a customer, such as personal computers, tablets, smartphones, personal digital assistants (PDAs), game consoles, interactive network television (IPTV), smart wearable devices, etc.

[0065] Electronic devices may also include network devices and / or client devices. The network devices include, but are not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of hosts or network servers.

[0066] The networks in which electronic devices are located include, but are not limited to, the Internet, wide area networks, metropolitan area networks, local area networks, and virtual private networks (VPNs).

[0067] like Figure 1 The diagram shown is a flowchart of a preferred embodiment of the artificial intelligence-based eye-tracking anti-fraud method of this application. The order of the steps in this flowchart can be changed, and some steps can be omitted, depending on different requirements.

[0068] S10, acquire eye gaze video data to generate an eye gaze sample trajectory set, the eye gaze sample trajectory set including an eye gaze positive sample trajectory set and an eye gaze negative sample trajectory set.

[0069] In an optional embodiment, acquiring gaze video data to generate a gaze sample trajectory set includes:

[0070] S101, collects gaze video data;

[0071] S102, mark all gaze video data as positive and negative samples according to a preset method, and take all gaze video data with positive sample labels as gaze positive sample dataset and all gaze video data with negative sample labels as gaze negative sample dataset.

[0072] S103, Generate positive gaze trajectory based on the gaze video data in the positive gaze sample dataset, and use all positive gaze trajectories as a positive gaze trajectory set;

[0073] S104, Generate gaze negative sample trajectories based on gaze video data in the gaze negative sample dataset, and use all gaze negative sample trajectories as a gaze negative sample trajectory set;

[0074] S105, the positive gaze sample trajectory set and the negative gaze sample trajectory set are used as the gaze sample trajectory set.

[0075] In this optional embodiment, in scenarios such as large-amount transfers and loan applications in the financial field, it is necessary to pay close attention to whether customers are at risk of fraud. Therefore, when customers conduct financial business remotely, such as applying for a loan, it is necessary to record video data in real time for review and filing. In this solution, a large amount of recorded video with customer gaze data is collected as gaze video data.

[0076] In this optional embodiment, all gaze video data can be manually labeled as positive or negative samples based on whether they contain fraudulent behavior, thereby dividing all gaze video data into two categories. Specifically, all gaze video data containing fraudulent behavior is labeled as positive samples, and all gaze video data not containing fraudulent behavior is labeled as negative samples. All gaze video data with positive sample labels is collected as the gaze positive sample dataset, and all gaze video data with negative sample labels is collected as the gaze negative sample dataset.

[0077] In this optional embodiment, for the obtained positive gaze sample dataset and negative gaze sample dataset, a preset number of frames can be extracted from all gaze video data first, such as extracting one frame as the target frame for every five consecutive frames. This can avoid the problem of excessive system resource consumption caused by processing a large number of video frames at the same time.

[0078] In this optional embodiment, after extracting a preset number of consecutive target frames from each gaze video data, the gaze focus in each target frame can be located using a microwave displacement sensor as the target point coordinates. Then, the target point coordinates of all target frames are connected sequentially according to the chronological order to generate a gaze movement trajectory. Specifically, a gaze negative sample trajectory can be generated based on the gaze video data in the gaze negative sample dataset, and all gaze negative sample trajectories are used as a gaze negative sample trajectory set, with each gaze negative sample trajectory corresponding one-to-one with the gaze video data in the gaze negative sample dataset. Simultaneously, a gaze positive sample trajectory is generated based on the gaze video data in the gaze positive sample dataset, and all gaze positive sample trajectories are used as a gaze positive sample trajectory set, with each gaze positive sample trajectory corresponding one-to-one with the gaze video data in the gaze positive sample dataset.

[0079] In this optional embodiment, the obtained positive gaze sample trajectory set and the negative gaze sample trajectory set are used as the gaze sample trajectory set.

[0080] In this way, we can obtain two sets of eye trajectory samples, one positive and one negative, with fraud labels, providing accurate data support for subsequent processes to obtain an eye trajectory fraud detector capable of detecting fraudulent information.

[0081] S11, the initial trajectory fragment set is obtained by segmenting all gaze sample trajectories in the gaze sample trajectory set, the initial trajectory fragment set includes a positive trajectory initial fragment set and a negative trajectory initial fragment set.

[0082] In an optional embodiment, the step of segmenting all gaze sample trajectories in the gaze sample trajectory set to obtain an initial trajectory fragment set, the initial trajectory fragment set including a positive trajectory initial fragment set and a negative trajectory initial fragment set, includes:

[0083] S111, extract the low-frequency information of each gaze sample trajectory in the gaze sample trajectory set to obtain the gaze sample low-frequency trajectory;

[0084] S112, calculate the extreme points of the low-frequency trajectory of the gaze sample, and divide the low-frequency trajectory of the gaze sample according to all the extreme points to obtain multiple initial trajectory segments of the low-frequency trajectory of the gaze sample.

[0085] S113, take all the initial segments of the trajectory corresponding to each gaze sample trajectory in the positive gaze sample trajectory set and the negative gaze sample trajectory set as the positive trajectory initial segment set and the negative trajectory initial segment set, respectively.

[0086] S114, the positive trajectory initial segment set and the negative trajectory initial segment set are used as the trajectory initial segment set.

[0087] In this optional embodiment, since gaze trajectories change rapidly over time and carry complex and variable high-frequency information, wavelet transform can be used to extract low-frequency information from each gaze sample trajectory in the gaze sample trajectory set to effectively extract meaningful gaze trajectory information. Wavelet transform is a commonly used signal frequency analysis tool that can decompose each gaze sample trajectory in the gaze sample trajectory set to obtain both high-frequency and low-frequency information. In this scheme, the low-frequency information of each gaze sample trajectory in the gaze sample trajectory set is used as the low-frequency gaze sample trajectory. The low-frequency gaze sample trajectory corresponds one-to-one with the gaze sample trajectory, and the gaze sample trajectory in the gaze sample trajectory set is segmented using the low-frequency gaze sample trajectory. The reason for choosing the low-frequency information is that the high-frequency information carried by the gaze trajectory changes frequently, making it difficult to extract effective gaze trajectory features, while the low-frequency information changes less and can effectively reflect the gaze trajectory features.

[0088] In this optional embodiment, for each obtained low-frequency trajectory of a light sample, the extreme points of each low-frequency trajectory of the light sample can be calculated, and the low-frequency trajectory of the light sample can be divided into multiple segments based on the extreme points, thereby obtaining multiple initial segments of the low-frequency trajectory of the light sample. For example... Figure 4 As shown, this is an acquired gaze sample trajectory. Black dots A, B, C, and D are the extreme points of the low-frequency trajectory of the gaze sample obtained after wavelet transform processing. Finally, the gaze sample trajectory is divided into 5 initial trajectory segments.

[0089] In this optional embodiment, after segmenting all gaze sample trajectories in the positive gaze sample trajectory set and the negative gaze sample trajectory set, this scheme uses all the initial trajectory segments corresponding to each gaze sample trajectory in the obtained positive gaze sample trajectory set and the negative gaze sample trajectory set as the initial positive trajectory segment set and the initial negative trajectory segment set, and uses the initial positive trajectory segment set and the initial negative trajectory segment set as the initial trajectory segment set.

[0090] In this way, all gaze sample trajectories can be segmented to obtain a large number of practically meaningful initial trajectory segments, providing data support for subsequent processes to select more effective initial trajectory segments.

[0091] S12, the initial trajectory fragment set is filtered to obtain a set of valid positive trajectory fragments and a set of valid negative trajectory fragments.

[0092] In an optional embodiment, the step of filtering the initial trajectory fragment set to obtain a set of valid positive trajectory fragments and a set of valid negative trajectory fragments includes:

[0093] S121, calculate the segment similarity between the initial segment of the target positive trajectory and each initial segment of the positive trajectory in the set of initial segments of the positive trajectory, wherein the initial segment of the target positive trajectory is any one of the initial segments of the positive trajectory in the set of initial segments of the positive trajectory;

[0094] S122, Based on a preset threshold and the segment similarity, obtain a set of similar segments corresponding to the initial segment of the target positive trajectory;

[0095] S123, traverse each positive trajectory initial segment in the set of positive trajectory initial segments to obtain the set of positive trajectory similar segments corresponding to each positive trajectory initial segment;

[0096] S124, calculate the segment similarity between the initial segment of the target negative trajectory and each initial segment of the negative trajectory in the set of initial segments of the negative trajectory, wherein the initial segment of the target negative trajectory is any one of the initial segments of the negative trajectory;

[0097] S125, Based on a preset threshold and the segment similarity, obtain a set of negative trajectory similar segments corresponding to the initial segment of the target negative trajectory;

[0098] S126, traverse each negative trajectory initial segment in the negative trajectory initial segment set to obtain the negative trajectory similar segment set corresponding to each negative trajectory initial segment;

[0099] S127, filter all positive trajectory similarity segment sets and all negative trajectory similarity segment sets respectively to obtain the positive trajectory valid segment set and the negative trajectory valid segment set.

[0100] In this optional embodiment, the segment similarity between the initial segment of the target positive trajectory and each initial segment of the positive trajectory in the set of initial segments of the positive trajectory can be calculated based on the Dynamic Time Warping (DTW) algorithm, wherein the initial segment of the target positive trajectory is any one of the initial segments of the positive trajectory in the set of initial segments of the positive trajectory.

[0101] In this optional embodiment, the DTW algorithm is a method for measuring the similarity between two time series. Since different time series may only have displacement on the time axis, compared to traditional methods for calculating the Euclidean distance between two time series, the DTW algorithm can calculate the similarity between them by extending and shortening the time series. Figure 5As shown, the two solid lines represent the initial segment of the target positive trajectory and any other initial segment of the positive trajectory in the set of initial segments excluding the target positive trajectory, respectively. The dashed line between the two initial segments represents similar points between them. The DTW algorithm measures the similarity between the two initial segments by calculating the sum of the distances between all these similar points, which is taken as the segment similarity. Since the DTW algorithm has various variations, the specific calculation process in this scheme can be consistent with the calculation process of the conventional DTW algorithm.

[0102] In this optional embodiment, after obtaining the segment similarity between the initial segment of the target positive trajectory and other initial segments of positive trajectories, all segment similarities corresponding to the initial segment of the target positive trajectory can be filtered based on a preset threshold. Initial segments of positive trajectories with similarities greater than the preset threshold are considered as similar segments of the target positive trajectory; initial segments of positive trajectories with similarities less than the preset threshold are filtered out. Finally, all similar segments of positive trajectories corresponding to the initial segment of the target positive trajectory are considered as a set of similar segments. The preset threshold for segment similarity can be 0.6, meaning that initial segments of positive trajectories with similarities greater than 0.6 are considered as similar segments of the target positive trajectory; initial segments of positive trajectories with similarities less than 0.6 are filtered out.

[0103] In this optional embodiment, a set of similar positive trajectory segments corresponding to each initial positive trajectory segment can be obtained by traversing each initial positive trajectory segment in the initial positive trajectory segment set. Similarly, a set of similar negative trajectory segments corresponding to each initial negative trajectory segment can be calculated.

[0104] In this optional embodiment, all sets of positive trajectory similar segments and all sets of negative trajectory similar segments can be sorted in descending order of the number of segments to obtain positive trajectory sorting results and negative trajectory sorting results. Then, the sets of positive trajectory similar segments and negative trajectory similar segments with the highest ranking in the positive trajectory sorting results and negative trajectory sorting results are selected respectively. For example, the highest ranking in the positive trajectory sorting results can be 50, that is, based on the number of segments, a total of 50 sets of positive trajectory similar segments and 50 sets of negative trajectory similar segments are finally obtained. In this scheme, each set of positive trajectory similar segments obtained after final screening is regarded as a valid set of positive trajectory segments, and each set of negative trajectory similar segments is regarded as a valid set of negative trajectory segments.

[0105] In this way, the initial trajectory fragment set can be filtered according to the number of fragments to obtain more effective initial trajectory fragments, which is beneficial to improving the detection accuracy of the eye trajectory fraud detector generated based on this in the subsequent process.

[0106] S13, Construct a gaze trajectory fraud detector based on the set of valid positive trajectory segments and the set of valid negative trajectory segments.

[0107] In an optional embodiment, constructing the eye-tracking fraud detector based on the set of valid positive trajectory segments and the set of valid negative trajectory segments includes:

[0108] S131, extract the longest common sequence of all segments in the set of valid segments of each positive trajectory;

[0109] S132, the longest common sequence of all valid segments of the positive trajectory is weighted and summed to obtain the positive trajectory reference template;

[0110] S133, extract the longest common sequence of all segments in the set of valid segments of each negative trajectory;

[0111] S134, the longest common sequence of all valid segments of the negative trajectory is weighted and summed to obtain the negative trajectory reference template;

[0112] S135, the positive trajectory reference template and the negative trajectory reference template are used as eye trajectory fraud detectors.

[0113] In this optional embodiment, since there are multiple similar valid trajectory segments in each set of positive trajectory valid segments and set of negative trajectory valid segments, and each valid trajectory segment is not exactly the same, in order to obtain the common characteristics of all valid trajectory segments in each set of positive trajectory valid segments and set of negative trajectory valid segments, the longest common subsequence (LCS) algorithm is used in this scheme to extract the longest common sequence of all valid trajectory segments in each set of positive trajectory valid segments and set of negative trajectory valid segments.

[0114] In this optional embodiment, the LCSS algorithm is used to find the longest subsequence among all sequences in a sequence set. For example, if str1 = "ABCBDCA" and str2 = "DABDA", their common subsequence is AA, and the longest common subsequence is ABDA. The characters in the common subsequence do not need to be consecutive in the original sequence.

[0115] In this optional embodiment, after extracting the longest common subsequence of all segments in each set of valid positive trajectory segments and each set of valid negative trajectory segments, the corresponding longest common subsequence can be weighted according to the number of valid trajectory segments included in the set of valid positive trajectory segments or the set of valid negative trajectory segments to which each longest common subsequence is located. The longest common subsequence of all sets of valid positive trajectory segments and all sets of valid negative trajectory segments that have obtained weights is then summed in a weighted manner to obtain the positive trajectory reference template and the negative trajectory reference template respectively.

[0116] In this optional embodiment, taking the acquisition of positive trajectory reference templates as an example, the specific process of assigning weights to the longest common subsequences corresponding to each positive trajectory reference template is as follows: count the total number of positive trajectory valid segments included in the set of positive trajectory valid segments where each longest common subsequence is located, and sum all the total numbers to obtain the total number of segments. In this scheme, the ratio of the total number of positive trajectory valid segments corresponding to each longest common subsequence to the total number of segments is used as the weight of the corresponding longest common subsequence.

[0117] In this optional embodiment, the process of obtaining the negative trajectory reference template is consistent with the process of obtaining the positive trajectory reference template, and the finally obtained positive trajectory reference template and the negative trajectory reference template are used together as an eye trajectory fraud detector to perform anti-fraud detection on the eye trajectory to be detected.

[0118] In this way, an eye-track fraud detector can be generated, which includes both positive and negative trajectory reference templates, enabling accurate identification of fraud risks associated with eye-tracking.

[0119] S14, based on the eye trajectory fraud detector, perform anti-fraud detection on the eye trajectory of the target to be detected to obtain the detection result.

[0120] In an optional embodiment, the step of obtaining the detection result by performing anti-fraud detection on the target's eye trajectory based on the eye trajectory fraud detector includes:

[0121] S141, calculate the similarity between the target gaze trajectory to be detected and the positive trajectory reference template and the negative trajectory reference template in the gaze trajectory fraud detector to obtain the positive trajectory similarity and the negative trajectory similarity;

[0122] S142, if the positive trajectory similarity is greater than the negative trajectory similarity, then the target gaze trajectory to be detected has a fraud risk, and the detection result is high risk of fraud.

[0123] S143, if the positive trajectory similarity is not greater than the negative trajectory similarity, then the target gaze trajectory to be detected does not have a fraud risk, and the detection result is low risk of fraud.

[0124] In this optional embodiment, for the target gaze trajectory to be detected, the similarity between the target gaze trajectory to be detected and the positive trajectory reference template and the negative trajectory reference template in the gaze trajectory fraud detector can be calculated using the DTW algorithm as the positive trajectory similarity and the negative trajectory similarity.

[0125] In this optional embodiment, the obtained positive trajectory similarity and negative trajectory similarity are compared. If the positive trajectory similarity is greater than the negative trajectory similarity, it indicates that the target gaze trajectory to be detected is closer to the positive trajectory reference template in the gaze trajectory fraud detector, and has a fraud risk. The detection result is high-risk fraud. If the positive trajectory similarity is not greater than the negative trajectory similarity, the target gaze trajectory to be detected is closer to the negative trajectory reference template in the gaze trajectory fraud detector, and has no fraud risk. The detection result is low-risk fraud.

[0126] In this way, the eye trajectory fraud detector can quickly detect the eye trajectory of the target and obtain accurate detection results, thus effectively distinguishing the fraud risk of the eye trajectory of the target.

[0127] In an optional embodiment, the eye-tracking fraud detector can be updated based on the detection results to improve its detection accuracy.

[0128] In this optional embodiment, the eye trajectory fraud detector can be updated based on the detection result. If the detection result indicates a high risk of fraud, the target eye trajectory corresponding to the high risk of fraud is marked as a positive eye sample and added to the positive eye sample trajectory set. A new positive trajectory reference template is then generated through a series of processing steps as shown in this scheme to update the eye trajectory fraud detector. If the detection result indicates a low risk of fraud, the target eye trajectory corresponding to the high risk of fraud is marked as a negative eye sample and added to the negative eye sample trajectory set. Finally, a new negative trajectory reference template is generated to update the eye trajectory fraud detector.

[0129] In this way, the detection accuracy of the eye trajectory fraud detector can be continuously improved through continuous updates.

[0130] Please see Figure 2 , Figure 2This is a functional block diagram of a preferred embodiment of the AI-based eye-tracking anti-fraud device of this application. The AI-based eye-tracking anti-fraud device 11 includes an acquisition unit 110, a segmentation unit 111, a filtering unit 112, a construction unit 113, and a detection unit 114. The module / unit referred to in this application refers to a series of computer-readable instruction segments that can be executed by the processor 13 and perform a fixed function, and are stored in the memory 12. In this embodiment, the functions of each module / unit will be described in detail in subsequent embodiments.

[0131] In an optional embodiment, the acquisition unit 110 is used to acquire gaze video data to generate a gaze sample trajectory set, the gaze sample trajectory set including a gaze positive sample trajectory set and a gaze negative sample trajectory set.

[0132] In an optional embodiment, acquiring gaze video data to generate a gaze sample trajectory set includes:

[0133] Collect eye-tracking video data;

[0134] According to a preset method, all gaze video data are labeled as positive and negative samples, and all gaze video data with positive sample labels are used as gaze positive sample datasets, and all gaze video data with negative sample labels are used as gaze negative sample datasets.

[0135] Based on the gaze video data in the gaze positive sample dataset, gaze positive sample trajectories are generated, and all gaze positive sample trajectories are used as the gaze positive sample trajectory set.

[0136] Based on the gaze video data in the gaze negative sample dataset, gaze negative sample trajectories are generated, and all gaze negative sample trajectories are used as a gaze negative sample trajectory set.

[0137] The positive gaze sample trajectory set and the negative gaze sample trajectory set are used as the gaze sample trajectory set.

[0138] In this optional embodiment, in scenarios such as large-amount transfers and loan applications in the financial field, it is necessary to pay close attention to whether customers are at risk of fraud. Therefore, when customers conduct financial business remotely, such as applying for a loan, it is necessary to record video data in real time for review and filing. In this solution, a large amount of recorded video with customer gaze data is collected as gaze video data.

[0139] In this optional embodiment, all gaze video data can be manually labeled as positive or negative samples based on whether they contain fraudulent behavior, thereby dividing all gaze video data into two categories. Specifically, all gaze video data containing fraudulent behavior is labeled as positive samples, and all gaze video data not containing fraudulent behavior is labeled as negative samples. All gaze video data with positive sample labels is collected as the gaze positive sample dataset, and all gaze video data with negative sample labels is collected as the gaze negative sample dataset.

[0140] In this optional embodiment, for the obtained positive gaze sample dataset and negative gaze sample dataset, a preset number of frames can be extracted from all gaze video data first, such as extracting one frame as the target frame for every five consecutive frames. This can avoid the problem of excessive system resource consumption caused by processing a large number of video frames at the same time.

[0141] In this optional embodiment, after extracting a preset number of consecutive target frames from each gaze video data, the gaze focus in each target frame can be located using a microwave displacement sensor as the target point coordinates. Then, the target point coordinates of all target frames are connected sequentially according to the chronological order to generate a gaze movement trajectory. Specifically, a gaze negative sample trajectory can be generated based on the gaze video data in the gaze negative sample dataset, and all gaze negative sample trajectories are used as a gaze negative sample trajectory set, with each gaze negative sample trajectory corresponding one-to-one with the gaze video data in the gaze negative sample dataset. Simultaneously, a gaze positive sample trajectory is generated based on the gaze video data in the gaze positive sample dataset, and all gaze positive sample trajectories are used as a gaze positive sample trajectory set, with each gaze positive sample trajectory corresponding one-to-one with the gaze video data in the gaze positive sample dataset.

[0142] In this optional embodiment, the obtained positive gaze sample trajectory set and the negative gaze sample trajectory set are used as the gaze sample trajectory set.

[0143] In an optional embodiment, the segmentation unit 111 is used to segment all gaze sample trajectories in the gaze sample trajectory set to obtain an initial trajectory fragment set, which includes a positive trajectory initial fragment set and a negative trajectory initial fragment set.

[0144] In an optional embodiment, the step of segmenting all gaze sample trajectories in the gaze sample trajectory set to obtain an initial trajectory fragment set, the initial trajectory fragment set including a positive trajectory initial fragment set and a negative trajectory initial fragment set, includes:

[0145] The low-frequency information of each gaze sample trajectory in the gaze sample trajectory set is extracted to obtain the gaze sample low-frequency trajectory;

[0146] Calculate the extreme points of the low-frequency trajectory of the gaze sample, and divide the low-frequency trajectory of the gaze sample according to all the extreme points to obtain multiple initial trajectory segments of the low-frequency trajectory of the gaze sample.

[0147] Each initial segment of the trajectory corresponding to each gaze sample trajectory in the positive gaze sample trajectory set and the negative gaze sample trajectory set is respectively used as the positive trajectory initial segment set and the negative trajectory initial segment set;

[0148] The positive trajectory initial segment set and the negative trajectory initial segment set are used as the trajectory initial segment set.

[0149] In this optional embodiment, since gaze trajectories change rapidly over time and carry complex and variable high-frequency information, wavelet transform can be used to extract low-frequency information from each gaze sample trajectory in the gaze sample trajectory set to effectively extract meaningful gaze trajectory information. Wavelet transform is a commonly used signal frequency analysis tool that can decompose each gaze sample trajectory in the gaze sample trajectory set to obtain both high-frequency and low-frequency information. In this scheme, the low-frequency information of each gaze sample trajectory in the gaze sample trajectory set is used as the low-frequency gaze sample trajectory. The low-frequency gaze sample trajectory corresponds one-to-one with the gaze sample trajectory, and the gaze sample trajectory in the gaze sample trajectory set is segmented using the low-frequency gaze sample trajectory. The reason for choosing the low-frequency information is that the high-frequency information carried by the gaze trajectory changes frequently, making it difficult to extract effective gaze trajectory features, while the low-frequency information changes less and can effectively reflect the gaze trajectory features.

[0150] In this optional embodiment, for each obtained low-frequency trajectory of a light sample, the extreme points of each low-frequency trajectory of the light sample can be calculated, and the low-frequency trajectory of the light sample can be divided into multiple segments based on the extreme points, thereby obtaining multiple initial segments of the low-frequency trajectory of the light sample. For example... Figure 4 As shown, this is an acquired gaze sample trajectory. Black dots A, B, C, and D are the extreme points of the low-frequency trajectory of the gaze sample obtained after wavelet transform processing. Finally, the gaze sample trajectory is divided into 5 initial trajectory segments.

[0151] In this optional embodiment, after segmenting all gaze sample trajectories in the positive gaze sample trajectory set and the negative gaze sample trajectory set, this scheme uses all the initial trajectory segments corresponding to each gaze sample trajectory in the obtained positive gaze sample trajectory set and the negative gaze sample trajectory set as the initial positive trajectory segment set and the initial negative trajectory segment set, and uses the initial positive trajectory segment set and the initial negative trajectory segment set as the initial trajectory segment set.

[0152] In an optional embodiment, the filtering unit 112 is used to filter the initial trajectory fragment set to obtain a set of valid positive trajectory fragments and a set of valid negative trajectory fragments.

[0153] In an optional embodiment, the step of filtering the initial trajectory fragment set to obtain a set of valid positive trajectory fragments and a set of valid negative trajectory fragments includes:

[0154] Calculate the segment similarity between the initial segment of the target positive trajectory and each initial segment of the positive trajectory in the set of initial segments of the positive trajectory, wherein the initial segment of the target positive trajectory is any one of the initial segments of the positive trajectory in the set of initial segments of the positive trajectory;

[0155] Based on a preset threshold and the segment similarity, a set of similar segments corresponding to the initial segment of the target positive trajectory is obtained;

[0156] By traversing each initial segment of the positive trajectory in the set of initial segments of the positive trajectory, a set of similar segments of the positive trajectory corresponding to each initial segment of the positive trajectory is obtained;

[0157] Calculate the segment similarity between the initial segment of the target negative trajectory and each initial segment of the negative trajectory in the set of initial segments of the negative trajectory, wherein the initial segment of the target negative trajectory is any one of the initial segments of the negative trajectory;

[0158] Based on a preset threshold and the segment similarity, a set of negative trajectory similar segments corresponding to the initial segment of the target negative trajectory is obtained;

[0159] By traversing each initial negative trajectory segment in the set of initial negative trajectory segments, a set of similar negative trajectory segments corresponding to each initial negative trajectory segment is obtained;

[0160] The effective segments of positive trajectories and the effective segments of negative trajectories are obtained by filtering all sets of similar segments of positive trajectories and all sets of similar segments of negative trajectories respectively.

[0161] In this optional embodiment, the segment similarity between the initial segment of the target positive trajectory and each initial segment of the positive trajectory in the set of initial segments of the positive trajectory can be calculated based on the Dynamic Time Warping (DTW) algorithm, wherein the initial segment of the target positive trajectory is any one of the initial segments of the positive trajectory in the set of initial segments of the positive trajectory.

[0162] In this optional embodiment, the DTW algorithm is a method for measuring the similarity between two time series. Since different time series may only have displacement on the time axis, compared to traditional methods for calculating the Euclidean distance between two time series, the DTW algorithm can calculate the similarity between them by extending and shortening the time series. Figure 5As shown, the two solid lines represent the initial segment of the target positive trajectory and any other initial segment of the positive trajectory in the set of initial segments excluding the target positive trajectory, respectively. The dashed line between the two initial segments represents similar points between them. The DTW algorithm measures the similarity between the two initial segments by calculating the sum of the distances between all these similar points, which is taken as the segment similarity. Since the DTW algorithm has various variations, the specific calculation process in this scheme can be consistent with the calculation process of the conventional DTW algorithm.

[0163] In this optional embodiment, after obtaining the segment similarity between the initial segment of the target positive trajectory and other initial segments of positive trajectories, all segment similarities corresponding to the initial segment of the target positive trajectory can be filtered based on a preset threshold. Initial segments of positive trajectories with similarities greater than the preset threshold are considered as similar segments of the target positive trajectory; initial segments of positive trajectories with similarities less than the preset threshold are filtered out. Finally, all similar segments of positive trajectories corresponding to the initial segment of the target positive trajectory are considered as a set of similar segments. The preset threshold for segment similarity can be 0.6, meaning that initial segments of positive trajectories with similarities greater than 0.6 are considered as similar segments of the target positive trajectory; initial segments of positive trajectories with similarities less than 0.6 are filtered out.

[0164] In this optional embodiment, a set of similar positive trajectory segments corresponding to each initial positive trajectory segment can be obtained by traversing each initial positive trajectory segment in the initial positive trajectory segment set. Similarly, a set of similar negative trajectory segments corresponding to each initial negative trajectory segment can be calculated.

[0165] In this optional embodiment, all sets of positive trajectory similar segments and all sets of negative trajectory similar segments can be sorted in descending order of the number of segments to obtain positive trajectory sorting results and negative trajectory sorting results. Then, the sets of positive trajectory similar segments and negative trajectory similar segments with the highest ranking in the positive trajectory sorting results and negative trajectory sorting results are selected respectively. For example, the highest ranking in the positive trajectory sorting results can be 50, that is, based on the number of segments, a total of 50 sets of positive trajectory similar segments and 50 sets of negative trajectory similar segments are finally obtained. In this scheme, each set of positive trajectory similar segments obtained after final screening is regarded as a valid set of positive trajectory segments, and each set of negative trajectory similar segments is regarded as a valid set of negative trajectory segments.

[0166] In an optional embodiment, the construction unit 113 is used to construct a gaze trajectory fraud detector based on the set of valid positive trajectory segments and the set of valid negative trajectory segments.

[0167] In an optional embodiment, constructing the eye-tracking fraud detector based on the set of valid positive trajectory segments and the set of valid negative trajectory segments includes:

[0168] Extract the longest common sequence of all segments in the set of valid segments for each positive trajectory;

[0169] The longest common sequence of all valid segments of the positive trajectory is weighted and summed to obtain the positive trajectory reference template;

[0170] Extract the longest common sequence of all segments in the set of valid segments for each negative trajectory;

[0171] The longest common sequence of all valid segments of the negative trajectory is weighted and summed to obtain the negative trajectory reference template;

[0172] The positive trajectory reference template and the negative trajectory reference template are used as eye trajectory fraud detectors.

[0173] In this optional embodiment, since there are multiple similar valid trajectory segments in each set of positive trajectory valid segments and set of negative trajectory valid segments, and each valid trajectory segment is not exactly the same, in order to obtain the common characteristics of all valid trajectory segments in each set of positive trajectory valid segments and set of negative trajectory valid segments, the longest common subsequence (LCS) algorithm is used in this scheme to extract the longest common sequence of all valid trajectory segments in each set of positive trajectory valid segments and set of negative trajectory valid segments.

[0174] In this optional embodiment, the LCSS algorithm is used to find the longest subsequence among all sequences in a sequence set. For example, if str1 = "ABCBDCA" and str2 = "DABDA", their common subsequence is AA, and the longest common subsequence is ABDA. The characters in the common subsequence do not need to be consecutive in the original sequence.

[0175] In this optional embodiment, after extracting the longest common subsequence of all segments in each set of valid positive trajectory segments and each set of valid negative trajectory segments, the corresponding longest common subsequence can be weighted according to the number of valid trajectory segments included in the set of valid positive trajectory segments or the set of valid negative trajectory segments to which each longest common subsequence is located. The longest common subsequence of all sets of valid positive trajectory segments and all sets of valid negative trajectory segments that have obtained weights is then summed in a weighted manner to obtain the positive trajectory reference template and the negative trajectory reference template respectively.

[0176] In this optional embodiment, taking the acquisition of positive trajectory reference templates as an example, the specific process of assigning weights to the longest common subsequences corresponding to each positive trajectory reference template is as follows: count the total number of positive trajectory valid segments included in the set of positive trajectory valid segments where each longest common subsequence is located, and sum all the total numbers to obtain the total number of segments. In this scheme, the ratio of the total number of positive trajectory valid segments corresponding to each longest common subsequence to the total number of segments is used as the weight of the corresponding longest common subsequence.

[0177] In this optional embodiment, the process of obtaining the negative trajectory reference template is consistent with the process of obtaining the positive trajectory reference template, and the finally obtained positive trajectory reference template and the negative trajectory reference template are used together as an eye trajectory fraud detector to perform anti-fraud detection on the eye trajectory to be detected.

[0178] In an optional embodiment, the detection unit 114 is used to perform anti-fraud detection on the target eye trajectory to be detected based on the eye trajectory fraud detector to obtain the detection result.

[0179] In an optional embodiment, the step of obtaining the detection result by performing anti-fraud detection on the target's eye trajectory based on the eye trajectory fraud detector includes:

[0180] The similarity between the target's gaze trajectory to be detected and the positive trajectory reference template and the negative trajectory reference template in the gaze trajectory fraud detector is calculated to obtain the positive trajectory similarity and the negative trajectory similarity.

[0181] If the positive trajectory similarity is greater than the negative trajectory similarity, then the target gaze trajectory to be detected has a fraud risk, and the detection result is a high risk of fraud.

[0182] If the positive trajectory similarity is not greater than the negative trajectory similarity, then the target gaze trajectory to be detected does not pose a fraud risk, and the detection result is low risk of fraud.

[0183] In this optional embodiment, the obtained positive trajectory similarity and negative trajectory similarity are compared. If the positive trajectory similarity is greater than the negative trajectory similarity, it indicates that the target gaze trajectory to be detected is closer to the positive trajectory reference template in the gaze trajectory fraud detector, and has a fraud risk. The detection result is high-risk fraud. If the positive trajectory similarity is not greater than the negative trajectory similarity, the target gaze trajectory to be detected is closer to the negative trajectory reference template in the gaze trajectory fraud detector, and has no fraud risk. The detection result is low-risk fraud.

[0184] In an optional embodiment, the eye-tracking fraud detector can be updated based on the detection results to improve its detection accuracy.

[0185] In this optional embodiment, the eye trajectory fraud detector can be updated based on the detection result. If the detection result indicates a high risk of fraud, the target eye trajectory corresponding to the high risk of fraud is marked as a positive eye sample and added to the positive eye sample trajectory set. A new positive trajectory reference template is then generated through a series of processing steps as shown in this scheme to update the eye trajectory fraud detector. If the detection result indicates a low risk of fraud, the target eye trajectory corresponding to the high risk of fraud is marked as a negative eye sample and added to the negative eye sample trajectory set. Finally, a new negative trajectory reference template is generated to update the eye trajectory fraud detector.

[0186] As can be seen from the above technical solutions, this application can extract positive and negative gaze sample trajectories from the acquired gaze video data, and combine the positive and negative gaze sample trajectories after segmentation, thereby obtaining a gaze trajectory fraud detector that can detect whether there is fraudulent information in the gaze trajectory, effectively improving the accuracy of identifying fraud risks in time-series gaze trajectories.

[0187] Please see Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device 1 includes a memory 12 and a processor 13. The memory 12 is used to store computer-readable instructions, and the processor 13 executes the computer-readable instructions stored in the memory to implement the artificial intelligence-based eye-tracking anti-fraud method described in any of the above embodiments.

[0188] In an alternative embodiment, the electronic device 1 further includes a bus, a computer program stored in the memory 12 and executable on the processor 13, such as an AI-based eye-tracking anti-fraud program.

[0189] Figure 3 Only electronic device 1 with memory 12 and processor 13 is shown. It will be understood by those skilled in the art that... Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0190] Combination Figure 1 The memory 12 in the electronic device 1 stores a plurality of computer-readable instructions to implement an artificial intelligence-based eye-tracking anti-fraud method, and the processor 13 can execute the plurality of instructions to achieve:

[0191] Acquire eye-viewing video data to generate an eye-viewing sample trajectory set, which includes a positive eye-viewing sample trajectory set and a negative eye-viewing sample trajectory set;

[0192] All gaze sample trajectories in the gaze sample trajectory set are segmented to obtain an initial trajectory fragment set, which includes a positive trajectory initial fragment set and a negative trajectory initial fragment set.

[0193] The initial set of trajectory segments is filtered to obtain a set of valid positive trajectory segments and a set of valid negative trajectory segments;

[0194] A gaze trajectory fraud detector is constructed based on the set of valid positive trajectory segments and the set of valid negative trajectory segments.

[0195] Based on the eye trajectory fraud detector, anti-fraud detection is performed on the eye trajectory of the target to be detected to obtain the detection result.

[0196] Specifically, the processor 13's implementation method for the above instructions can be found in [reference needed]. Figure 1 The descriptions of the relevant steps in the corresponding embodiments are not repeated here.

[0197] Those skilled in the art will understand that the schematic diagram is merely an example of electronic device 1 and does not constitute a limitation on electronic device 1. Electronic device 1 can be a bus-type structure or a star-type structure. Electronic device 1 may also include more or fewer other hardware or software than shown in the diagram, or different component arrangements. For example, electronic device 1 may also include input / output devices, network access devices, etc.

[0198] It should be noted that electronic device 1 is only an example. Other existing or future electronic products that are suitable for this application should also be included within the scope of protection of this application and are incorporated herein by reference.

[0199] The memory 12 includes at least one type of readable storage medium, which can be non-volatile or volatile. The readable storage medium includes flash memory, portable hard drives, multimedia cards, card-type memory (e.g., SD or DX memory), magnetic storage, magnetic disks, optical disks, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a portable hard drive of the electronic device 1. In other embodiments, the memory 12 can be an external storage device of the electronic device 1, such as a plug-in portable hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 1. The memory 12 can be used not only to store application software and various types of data installed on the electronic device 1, such as the code of an AI-based eye-tracking anti-fraud program, but also to temporarily store data that has been output or will be output.

[0200] In some embodiments, the processor 13 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits packaged with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 13 is the control unit of the electronic device 1, connecting various components of the electronic device 1 via various interfaces and lines. It executes programs or modules stored in the memory 12 (e.g., executing an AI-based eye-tracking anti-fraud program) and calls data stored in the memory 12 to perform various functions and process data in the electronic device 1.

[0201] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the various embodiments of the AI-based eye-tracking anti-fraud method described above, for example... Figure 1 The steps are shown.

[0202] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 12 and executed by the processor 13 to complete this application. The one or more modules / units may be a series of computer-readable instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into an acquisition unit 110, a segmentation unit 111, a filtering unit 112, a construction unit 113, and a detection unit 114.

[0203] The integrated unit implemented as a software functional module described above can be stored in a computer-readable storage medium. This software functional module, stored in a storage medium, includes several instructions to cause a computer device (which may be a personal computer, computer equipment, or network device, etc.) or processor to execute portions of the artificial intelligence-based eye-tracking anti-fraud method described in the various embodiments of this application.

[0204] If the modules / units integrated in electronic device 1 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware devices. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above.

[0205] The computer program includes computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory, and other memory.

[0206] Furthermore, the computer-readable storage medium may primarily include a stored program area and a stored data area, wherein the stored program area may store the operating system, an application program required for at least one function, etc.; and the stored data area may store data created based on the use of blockchain nodes, etc.

[0207] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0208] The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, in... Figure 3 The symbol is represented by only one arrow, but this does not indicate that there is only one bus or one type of bus. The bus is configured to enable communication between the memory 12 and at least one processor 13, etc.

[0209] This application also provides a computer-readable storage medium (not shown) storing computer-readable instructions, which are executed by a processor in an electronic device to implement the artificial intelligence-based eye-tracking anti-fraud method described in any of the above embodiments.

[0210] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0211] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0212] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0213] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices described in the specification may also be implemented by a single unit or device through software or hardware. Terms such as "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0214] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the spirit and scope of the technical solutions of this application.

Claims

1. A gaze-tracking anti-fraud method based on artificial intelligence, characterized in that, The method includes: Acquire eye-viewing video data to generate an eye-viewing sample trajectory set, which includes a positive eye-viewing sample trajectory set and a negative eye-viewing sample trajectory set; The process involves segmenting all gaze sample trajectories in the gaze sample trajectory set to obtain an initial trajectory fragment set. This initial trajectory fragment set includes a positive trajectory initial fragment set and a negative trajectory initial fragment set. The segmentation process includes: extracting low-frequency information from each gaze sample trajectory in the gaze sample trajectory set to obtain a low-frequency gaze sample trajectory; calculating the extreme points of the low-frequency gaze sample trajectory and segmenting the low-frequency gaze sample trajectory based on all extreme points to obtain multiple initial trajectory fragments; using all initial trajectory fragments corresponding to each gaze sample trajectory in the positive gaze sample trajectory set and the negative gaze sample trajectory set as the positive trajectory initial fragment set and the negative trajectory initial fragment set, respectively; and using the positive trajectory initial fragment set and the negative trajectory initial fragment set as the initial trajectory fragment set. The initial set of trajectory segments is filtered to obtain a set of valid positive trajectory segments and a set of valid negative trajectory segments; A gaze trajectory fraud detector is constructed based on the set of valid positive trajectory segments and the set of valid negative trajectory segments. Based on the eye trajectory fraud detector, anti-fraud detection is performed on the eye trajectory of the target to be detected to obtain the detection result.

2. The artificial intelligence-based eye-tracking anti-fraud method as described in claim 1, characterized in that, The step of acquiring gaze video data to generate a gaze sample trajectory set includes: Collect eye-tracking video data; According to a preset method, all gaze video data are labeled as positive and negative samples, and all gaze video data with positive sample labels are used as gaze positive sample datasets, and all gaze video data with negative sample labels are used as gaze negative sample datasets. Based on the gaze video data in the gaze positive sample dataset, gaze positive sample trajectories are generated, and all gaze positive sample trajectories are used as the gaze positive sample trajectory set. Based on the gaze video data in the gaze negative sample dataset, gaze negative sample trajectories are generated, and all gaze negative sample trajectories are used as a gaze negative sample trajectory set. The positive gaze sample trajectory set and the negative gaze sample trajectory set are used as the gaze sample trajectory set.

3. The artificial intelligence-based eye-tracking anti-fraud method as described in claim 1, characterized in that, The step of filtering the initial set of trajectory segments to obtain the set of valid positive trajectory segments and the set of valid negative trajectory segments includes: Calculate the segment similarity between the initial segment of the target positive trajectory and each initial segment of the positive trajectory in the set of initial segments of the positive trajectory, wherein the initial segment of the target positive trajectory is any one of the initial segments of the positive trajectory in the set of initial segments of the positive trajectory; Based on a preset threshold and the segment similarity, a set of similar segments corresponding to the initial segment of the target positive trajectory is obtained; By traversing each initial segment of the positive trajectory in the set of initial segments of the positive trajectory, a set of similar segments of the positive trajectory corresponding to each initial segment of the positive trajectory is obtained; Calculate the segment similarity between the initial segment of the target negative trajectory and each initial segment of the negative trajectory in the set of initial segments of the negative trajectory, wherein the initial segment of the target negative trajectory is any one of the initial segments of the negative trajectory; Based on a preset threshold and the segment similarity, a set of negative trajectory similar segments corresponding to the initial segment of the target negative trajectory is obtained; By traversing each initial negative trajectory segment in the set of initial negative trajectory segments, a set of similar negative trajectory segments corresponding to each initial negative trajectory segment is obtained; The effective segments of positive trajectories and the effective segments of negative trajectories are obtained by filtering all sets of similar segments of positive trajectories and all sets of similar segments of negative trajectories respectively.

4. The artificial intelligence-based eye-tracking anti-fraud method as described in claim 3, characterized in that, The step of filtering all positive trajectory similarity fragment sets and all negative trajectory similarity fragment sets to obtain the effective positive trajectory fragment set and the effective negative trajectory fragment set includes: Count the number of segments included in each positive trajectory similarity segment set and each negative trajectory similarity segment set; Sort all positive trajectory similarity fragment sets and all negative trajectory similarity fragment sets in descending order of the number of fragments to obtain positive trajectory sorting results and negative trajectory sorting results; Based on the positive trajectory sorting results and negative trajectory sorting results, all positive trajectory similar segment sets and all negative trajectory similar segment sets are filtered to obtain the same number of multiple positive trajectory valid segment sets and negative trajectory valid segment sets.

5. The artificial intelligence-based eye-tracking anti-fraud method as described in claim 1, characterized in that, The construction of the eye trajectory fraud detector based on the set of valid positive trajectory segments and the set of valid negative trajectory segments includes: Extract the longest common sequence of all segments in the set of valid segments for each positive trajectory; The longest common sequence of all valid segments of the positive trajectory is weighted and summed to obtain the positive trajectory reference template; Extract the longest common sequence of all segments in the set of valid segments for each negative trajectory; The longest common sequence of all valid segments of the negative trajectory is weighted and summed to obtain the negative trajectory reference template; The positive trajectory reference template and the negative trajectory reference template are used as eye trajectory fraud detectors.

6. The artificial intelligence-based eye-tracking anti-fraud method as described in claim 1, characterized in that, The process of obtaining detection results by performing anti-fraud detection on the target's eye trajectory based on the eye trajectory fraud detector includes: The similarity between the target's gaze trajectory to be detected and the positive trajectory reference template and the negative trajectory reference template in the gaze trajectory fraud detector is calculated to obtain the positive trajectory similarity and the negative trajectory similarity. If the positive trajectory similarity is greater than the negative trajectory similarity, then the target gaze trajectory to be detected has a fraud risk, and the detection result is a high risk of fraud. If the positive trajectory similarity is not greater than the negative trajectory similarity, then the target gaze trajectory to be detected does not pose a fraud risk, and the detection result is low risk of fraud.

7. An artificial intelligence-based eye-tracking anti-fraud device, characterized in that, The device is used to implement the artificial intelligence-based eye-tracking anti-fraud method as described in any one of claims 1 to 6, the device comprising: An acquisition unit is used to acquire eye gaze video data to generate an eye gaze sample trajectory set, the eye gaze sample trajectory set including an eye gaze positive sample trajectory set and an eye gaze negative sample trajectory set. The segmentation unit is used to segment all gaze sample trajectories in the gaze sample trajectory set to obtain an initial trajectory fragment set, which includes a positive trajectory initial fragment set and a negative trajectory initial fragment set. The filtering unit is used to filter the initial trajectory segment set to obtain a set of valid positive trajectory segments and a set of valid negative trajectory segments; The construction unit is used to construct a gaze trajectory fraud detector based on the set of valid positive trajectory segments and the set of valid negative trajectory segments; The detection unit is used to perform anti-fraud detection on the target eye trajectory to be detected based on the eye trajectory fraud detector to obtain the detection result.

8. An electronic device, characterized in that, The electronic device includes: Memory, which stores computer-readable instructions; and The processor executes computer-readable instructions stored in the memory to implement the AI-based eye-tracking anti-fraud method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the artificial intelligence-based eye-tracking anti-fraud method as described in any one of claims 1 to 6.

Images