Video fraud detection method and device and related equipment

Abstract

The invention belongs to the technical field of financial science and technology, provides a video fraud detection method and device and related equipment, aims to solve the problem of inaccurate video fraud detection in the traditional technology, and determines a video corresponding to a user by responding to a video fraud detection instruction. The method comprises the following steps: acquiring a video, extracting combination features between image frames corresponding to video frames contained in the video and corresponding audio frames, and then predicting prediction probability delay distribution corresponding to a next video frame according to a preset probability prediction model, all the combination features and a corresponding time sequence and the combination feature corresponding to a previous video frame. And judging whether the video has a video fraud risk or not according to the predicted probability delay distribution, so that the accuracy of video fraud detection can be improved, for example, in financial science and technology, for credit business in financial business, whether the user is compliant or not and whether fraud exists or not can be detected through the video provided by the user, and the user experience is improved. The accuracy of video fraud detection of credit business can be improved.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and in particular to a video fraud detection method, apparatus, computer equipment and computer-readable storage medium. Background Art

[0002] Video is an important multimedia information carrier and is widely used in financial scenarios including but not limited to identity verification, transaction review, and risk assessment. For example, in bank credit business, users are generally required to upload videos, and the bank system verifies users based on the videos to detect whether the users are compliant or have committed fraud.

[0003] However, with the development of deep learning technology, videos are also facing the risk of being tampered with or forged. For example, fraudulent videos can be generated using face-changing technology, voice synthesis technology, editing and splicing technology, etc. These videos may cause huge losses and harm to financial institutions and users. Therefore, how to effectively detect and prevent video fraud is an important and urgent issue in the financial field.

[0004] In traditional technologies, recognition algorithms based on face recognition or face action recognition are generally used to verify users through video. However, these recognition algorithms often require a large amount of labeled data or expert knowledge, and are difficult to adapt to the ever-changing and increasing fraud methods, resulting in the risk of video fraud. Therefore, how to effectively detect the ever-changing video fraud has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The present application provides a video fraud detection method, apparatus, computer device and computer-readable storage medium, which can solve the technical problem of inaccurate video fraud detection in traditional technologies including but not limited to the field of financial technology.

[0006] In a first aspect, the present application provides a video fraud detection method, comprising: responding to a video fraud detection instruction, determining a video corresponding to a user; based on a preset video signal delay estimation model, determining a combination feature corresponding to a combination between an image frame corresponding to each video frame of the video and its corresponding audio frame, and obtaining a plurality of combination features of the video based on a time sequence; based on a preset probability prediction model and all the combination features and their corresponding time sequence, and according to the combination feature corresponding to the previous video frame, predicting a probability delay distribution corresponding to the delay between an image frame corresponding to the next video frame and its corresponding audio frame, and obtaining a predicted probability delay distribution; determining whether the predicted probability delay distribution meets a preset probability delay distribution condition; and determining that video fraud exists in the video when the predicted probability delay distribution meets the preset probability delay distribution condition.

[0007] In a second aspect, the present application provides a video fraud detection device, comprising: a first determination unit, for responding to an instruction for video fraud detection and determining a video corresponding to a user; a second determination unit, for determining, based on a preset video signal delay estimation model, a combination feature corresponding to a combination between an image frame corresponding to each video frame of the video and its corresponding audio frame, to obtain a plurality of combination features of the video based on a time sequence; a first prediction unit, for predicting, based on a preset probability prediction model and all the combination features and their corresponding time sequence, and according to the combination features corresponding to a previous video frame, a probability delay distribution corresponding to the delay between an image frame corresponding to a next video frame and its corresponding audio frame, to obtain a predicted probability delay distribution; a first judgment unit, for judging whether the predicted probability delay distribution satisfies a preset probability delay distribution condition; and a first judgment unit, for judging whether video fraud exists in the video when the predicted probability delay distribution satisfies the preset probability delay distribution condition.

[0008] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the video fraud detection method when executing the computer program.

[0009] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor performs the steps of the video fraud detection method.

[0010] The present application provides a video fraud detection method, apparatus, computer equipment and computer-readable storage medium. The method determines the video corresponding to the user by responding to a video fraud detection instruction, and extracts the combination features between the image frame corresponding to the video frame contained in the video and its corresponding audio frame, and then predicts the predicted probability delay distribution corresponding to the next video frame based on a preset probability prediction model and all the combination features and their corresponding time sequence, and according to the combination features corresponding to the previous video frame, and then judges whether the video has a video fraud risk based on the predicted probability delay distribution. Since in the process of video fraud detection, the combination features of the image frame and the audio frame corresponding to the video frame and the delay probability distribution of the previous video frame are fully utilized to predict the probability delay distribution of the next video frame, and the fraudulent nature of the video is judged based on the predicted probability delay distribution, video fraud detection can be achieved based on the video itself and by fully utilizing the commonality of delay in video transmission, and the accuracy of video fraud detection can be improved. For example, in financial technology, for credit business in financial business, it is possible to detect whether the user is compliant and whether there is fraud through the video provided by the user, and the accuracy of video fraud detection can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. 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.

[0012] Figure 1 A schematic diagram of a process flow of a video fraud detection method provided in an embodiment of the present application;

[0013] Figure 2 A schematic diagram of a probability delay distribution prediction process of a video fraud detection method provided in an embodiment of the present application;

[0014] Figure 3 A schematic diagram of a first sub-process of the video fraud detection method provided in an embodiment of the present application;

[0015] Figure 4 A schematic diagram of a second sub-process of the video fraud detection method provided in an embodiment of the present application;

[0016] Figure 5 A schematic diagram of a third sub-process of the video fraud detection method provided in an embodiment of the present application;

[0017] Figure 6 A schematic block diagram of a video fraud detection device provided in an embodiment of the present application;

[0018] Figure 7 A schematic block diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0021] An embodiment of the present application provides a video fraud detection method, which can be applied to but not limited to computer devices such as laptops, desktop computers, servers, and is used in video fraud detection based on financial services in the field of financial technology, including but not limited to.

[0022] In the face of technical problems in traditional technologies such as inaccurate video fraud detection in financial services, the inventors proposed a video fraud detection method according to an embodiment of the present application. The core idea of ​​the embodiment of the present application is: in the process of video fraud detection, the combined characteristics of the image frame and the audio frame corresponding to the video frame and the delay probability distribution of the previous video frame are fully utilized to predict the probability delay distribution of the next video frame, and the fraudulent nature of the video is judged based on the predicted probability delay distribution. Video fraud detection can be achieved based on the video itself and by fully utilizing the common delay characteristics of video transmission, and the accuracy of video fraud detection can be improved. For example, in financial technology, for credit business in financial business, it is possible to detect whether the user is compliant and whether there is fraud through the video provided by the user, thereby improving the accuracy of video fraud detection.

[0023] Some embodiments of the present application are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0024] See also Figure 1 and Figure 2 , Figure 1 A schematic diagram of a video fraud detection method provided in an embodiment of the present application. Figure 2 A schematic diagram of the probability delay distribution prediction process of the video fraud detection method provided in the embodiment of the present application. Figure 1 and Figure 2 As shown, the method includes but is not limited to the following steps S11-S16:

[0025] S11, responding to the video fraud detection instruction and determining the video corresponding to the user.

[0026] Explanatory note, in the case of video fraud detection, the video that needs to be detected is determined. The video is generally the video corresponding to the user. For example, the user's video is collected and obtained, or the user is prompted to collect and upload the required video, and then the video that needs to be detected is obtained to determine the video that needs to be detected.

[0027] For example, in the field of financial technology, for financial services including but not limited to credit services, when users need to handle related credit services, the corresponding videos of the users are collected and obtained, and the collected videos are tested for video fraud risks to determine whether the users are compliant and whether there is fraud.

[0028] S12. Based on a preset video signal delay estimation model, determine the combination features corresponding to the combination between the image frame corresponding to each video frame of the video and its corresponding audio frame, and obtain a plurality of combination features of the video based on the time sequence.

[0029] Explanatory, please continue to see Figure 2 , a video signal delay estimation model is pre-set, that is, a preset video signal delay estimation model, the preset video signal delay estimation model represents a model for estimating video signal delay by identifying relevant video signal delay features, and the preset video signal delay estimation model includes but is not limited to the following three parts: 1) an image feature extraction module for extracting image features corresponding to image frames of an input video; 2) an audio feature extraction module for extracting audio features corresponding to audio frames of an input video; 3) an image-audio combination feature module for determining the combination features corresponding to the combination features between the image frames of the input video and the corresponding audio frames, and using preset video training samples to train the preset video signal delay estimation model, so that the preset video signal delay estimation model can extract the combination features corresponding to the combination features between the image frames corresponding to the video frames contained in the video and the corresponding audio frames, wherein the preset video training samples include the input preset video training samples and their corresponding labels, and the preset video training samples include but are not limited to the synchronization of the image frames corresponding to the video frames and their corresponding audio frames and their corresponding Combined features, and the image frame corresponding to the video frame is asynchronous with the corresponding audio frame and its corresponding combined features, thereby, the preset video signal delay estimation model can identify the delay features of the relevant video signal corresponding to the video frame of the input video to estimate the relevant information of the video signal delay. At the same time, the preset video signal delay estimation model can also identify the non-delay features of the relevant video signal corresponding to the video frame of the input video to estimate the relevant information of the video signal non-delay, that is, the preset video signal delay estimation model can also identify the non-delay features of the relevant video signal corresponding to the video frame of the input video to estimate the relevant information of the video signal synchronization. The delay features corresponding to the asynchrony of the relevant video signal corresponding to the video frame of the input video and the non-delay features corresponding to the synchronization of the relevant video signal corresponding to the video frame are like the front and back of a coin. When the delay features corresponding to the asynchrony of the relevant video signal corresponding to the video frame of the input video can be identified, the non-delay features corresponding to the synchronization of the relevant video signal corresponding to the video frame of the input video can also be identified from the opposite angle. It should be noted that, in the scenario of video fraud detection, it is generally preferred to detect the asynchrony of the relevant video signals corresponding to the video frames of the input video (i.e., the delay information of the relevant video signals corresponding to the video frames) to perform video fraud detection. Therefore, the above model is named the relevant delay estimation model. However, it does not mean that the preset video signal delay estimation model can only be used to estimate the video signal delay, but can also identify and estimate the non-delay features corresponding to the synchronization of the relevant video signals corresponding to the video frames of the input video.

[0030] According to the setting of the above-mentioned preset video signal delay estimation model, based on the preset video signal delay estimation model including but not limited to the Transformer of 8 layers of Block, the combination features corresponding to the combination between the image frame corresponding to each video frame of the video and its corresponding audio frame are determined. The video generally includes several video frames, and the several video frames correspond to the time sequence corresponding to the duration of the video. Thus, several combination features of the video based on the time sequence are obtained. The image frame contained in the video frame corresponding to the combination feature and its corresponding audio frame may be asynchronous or synchronous.

[0031] Illustratively, in the field of financial technology, for financial services including but not limited to credit services, when a user needs to handle related credit services, the video corresponding to the user is collected and obtained, and then based on a preset video signal delay estimation model, the combination features corresponding to the combination of the image frame corresponding to each video frame of the video corresponding to the credit business user and its corresponding audio frame are determined to obtain a number of combination features of the video based on time sequence.

[0032] S13. Based on a preset probability prediction model and all the combined features and their corresponding time sequence, and according to the combined features corresponding to the previous video frame, predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame, and obtain the predicted probability delay distribution.

[0033] Explanatoryly, a probability prediction model is pre-set, i.e., a preset probability prediction model. The preset probability prediction model represents a probability prediction model that can predict the probability delay distribution of the next video frame based on the information of the previous video frame of the video. The preset probability prediction model is an autoregressive model (English: Autoregressive model, referred to as AR model). The preset probability prediction model can use a Transformer including but not limited to a 6-layer GRU+6-layer Block.

[0034] Please continue reading Figure 2As set by the preset probability prediction model above, based on the preset probability prediction model, and based on all the combined features and their corresponding time sequence, and according to the combined features corresponding to the previous video frame, the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame is predicted to obtain the predicted probability delay distribution, wherein, since the combined features actually describe the features corresponding to the combination between the image frame corresponding to the video frame and its corresponding audio frame, the combined features correspond to the delay probability distribution of the video frame, and thus, the above-mentioned prediction of the probability delay distribution of the next video frame is actually based on the delay probability distribution of the previous video frame to predict the probability delay distribution of the next video frame, wherein the delay probability distribution describes the probability distribution of the delay, which refers to the probability distribution of the delay of the video frames contained in the video within a certain time period, and the probability delay distribution describes the distribution of the probability delay, which usually refers to the probability distribution law followed by the delay of the video frame in a random process, and this distribution describes the probability density or probability accumulation of the delay of the video frame at different values.

[0035] S14, determining whether the predicted probability delay distribution meets a preset probability delay distribution condition;

[0036] S15. When the predicted probability delay distribution meets a preset probability delay distribution condition, determining that video fraud exists in the video;

[0037] S16: When the predicted probability delay distribution does not satisfy a preset probability delay distribution condition, determine that there is no video fraud in the video.

[0038] Explanatory, the probability delay distribution condition is pre-set, i.e., the preset probability delay distribution condition. The preset probability delay distribution condition indicates the condition that the probability delay distribution should meet when the preset video is fraudulent. The preset probability delay distribution condition is divided into two situations, including but not limited to the following: 1) When the video contains several video frames and the image frames contained in the video frames are not synchronized with the corresponding audio frames, the larger the number of all the unsynchronized video frames is, the greater the possibility that the video is abnormal and the greater the risk of fraud in the video is; 2) When the video contains several video frames and the image frames contained in the video frames are synchronized with the corresponding audio frames, the smaller the number of all the synchronized video frames is, the greater the possibility that the video is abnormal and the greater the risk of fraud in the video is. The above two situations are similar to the front and back of a coin. Although the contents are different, the essence is the same. The preset probability delay distribution condition can be set from at least one of the above two situations, and both can achieve the same purpose in different ways.

[0039] Based on the specific settings of the above-mentioned preset probability delay distribution conditions, it is determined whether the predicted probability delay distribution meets the preset probability delay distribution conditions. For example, when the preset probability delay distribution conditions indicate that the image frames included in the video frames are not synchronized with the corresponding audio frames, and the number of all synchronized video frames is greater than the preset number, it is determined that the video is abnormal and there is a risk of fraud in the video. Alternatively, when the preset probability delay distribution conditions indicate that the image frames included in the video frames are synchronized with the corresponding audio frames, and the number of all synchronized video frames is less than the preset number, it is determined that the video is abnormal and there is a risk of fraud in the video. Furthermore, when the predicted probability delay distribution meets the preset probability delay distribution conditions, it is determined that the video is subject to video fraud. Furthermore, when the predicted probability delay distribution does not meet the preset probability delay distribution conditions, it is determined that the video is not subject to video fraud.

[0040] For example, in financial technology, for identity authentication, transaction review, risk assessment and other services involved in credit business in financial services, whether the predicted probability delay distribution corresponding to the video frames contained in the video provided by the credit user meets the preset probability delay distribution conditions can be used to judge whether the video has a fraud risk, and then to detect whether the user is compliant and whether there is fraud. This can improve the accuracy of video fraud detection for credit users, thereby conducting effective video fraud detection, improving the efficiency of video fraud detection, and improving the security of credit business.

[0041] In the embodiment of the present application, by responding to the instruction of video fraud detection, the video corresponding to the user is determined, and based on the preset image and audio combined feature extraction model, the combined features corresponding to the combination between the image frame corresponding to each video frame of the video and its corresponding audio frame are extracted to obtain a plurality of combined features based on the time sequence of the video, and based on the preset probability prediction model and all the combined features and their corresponding time sequence, and according to the combined features corresponding to the previous video frame, the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame is predicted to obtain the predicted probability delay distribution, and then whether the video has a video fraud risk is judged according to whether the predicted probability delay distribution meets the preset probability delay distribution condition. Since in the process of video fraud detection, the combined features of the image frame and the audio frame corresponding to the video frame and the delay probability distribution of the previous video frame are fully utilized to predict the probability delay distribution of the next video frame, and the fraud judgment of the video is performed according to the predicted probability delay distribution, video fraud detection can be realized based on the video itself and by fully utilizing the commonality of delays of different video transmissions, and the accuracy of video fraud detection can be improved. For example, in financial technology, for credit business in financial business, it is possible to detect whether the user is compliant and whether there is fraud through the video provided by the user, and the accuracy of video fraud detection can be improved.

[0042] In one embodiment, based on a preset video signal delay estimation model, a combination feature corresponding to the combination between an image frame corresponding to each video frame of the video and its corresponding audio frame is determined to obtain several combination features of the video based on time sequence, including:

[0043] Determining a discrete delay feature corresponding to the video based on a preset video signal delay estimation model;

[0044] Determining a delay probability distribution corresponding to a video frame included in the video according to the discrete delay feature;

[0045] According to the delay probability distribution, a combination feature corresponding to the combination between the image frame corresponding to the video frame and its corresponding audio frame is determined, and a plurality of combination features based on time sequence corresponding to a plurality of video frames contained in the video are obtained.

[0046] Explanatoryly, as described above, a video signal delay estimation model is preset, and based on the preset video signal delay estimation model, a continuous video signal is converted into a discrete signal. Generally, the continuous video signal is sampled according to video frames to discretize the video signal to obtain a discrete signal, and based on the discrete signal, a discrete delay feature corresponding to the video is determined. The discrete delay feature represents a set of frame numbers of delay between an image frame and different audio frames.

[0047] When there is a delay between the image frame corresponding to the video frame and its corresponding audio frame, the corresponding feature, namely the delay feature, will definitely be exhibited. When the continuous video signal is discretized, it is correspondingly called a discrete delay feature. For the discrete delay feature, there is a corresponding delay probability distribution to represent the probability distribution corresponding to the delay of the video frame. Therefore, according to the discrete delay feature, the delay probability distribution corresponding to the video frame contained in the video is determined. The delay probability distribution represents the probability distribution of the delay exhibited by the discrete delay feature corresponding to the video frame. Then, according to the delay probability distribution, the combination feature corresponding to the combination between the image frame corresponding to the video frame and its corresponding audio frame is determined, and several combination features based on time sequence corresponding to several video frames contained in the video are obtained. The combination feature includes two situations: 1) the combination feature exhibited when there is a delay between the image frame corresponding to the video frame and its corresponding audio frame; 2) the combination feature exhibited when there is no delay between the image frame corresponding to the video frame and its corresponding audio frame. However, this situation is a special case where the delay is zero.

[0048] In an embodiment of the present application, a discrete delay feature corresponding to a video is determined, and based on the discrete delay feature, a delay probability distribution corresponding to a video frame contained in the video is determined, and then based on the delay probability distribution, a combined feature corresponding to the video frame is determined to obtain a combined feature corresponding to the video frame contained in the video, and then based on the combined feature, a probability delay distribution corresponding to the next video frame is predicted, and then based on the predicted probability delay distribution, it is determined whether the video has a video fraud risk, thereby realizing video fraud detection based on the video itself and making full use of the delay commonalities of different video transmissions, thereby improving the accuracy of video fraud detection.

[0049] In one embodiment, see Figure 3 , Figure 3 This is a schematic diagram of the first sub-process of the video fraud detection method provided in the embodiment of the present application. Figure 3 As shown, in this embodiment, judging whether the predicted probability delay distribution meets the preset probability delay distribution condition includes:

[0050] S31, judging whether an image frame of a corresponding video frame and its corresponding audio frame are asynchronous according to the predicted probability delay distribution;

[0051] S32: when the image frame of the video frame is synchronized with the corresponding audio frame, the asynchronous ratio of the video frame corresponding to the asynchronous image frame and the corresponding audio frame in all the video frames is not counted;

[0052] S33, when the image frame of the video frame is asynchronous with the corresponding audio frame, counting the asynchronous ratio of the video frames corresponding to the asynchronous image frame and the corresponding audio frame in all video frames;

[0053] S34, determining whether the asynchronous ratio meets a preset asynchronous ratio threshold condition;

[0054] S35. When the asynchronous ratio satisfies a preset asynchronous ratio threshold condition, determining that the predicted probability delay distribution satisfies a preset probability delay distribution condition;

[0055] S36: When the asynchronous ratio does not satisfy a preset asynchronous ratio threshold condition, it is not determined whether the predicted probability delay distribution satisfies a preset probability delay distribution condition.

[0056] Explanatoryly, an asynchronous ratio threshold condition is pre-set, that is, a preset asynchronous ratio threshold condition, which indicates a pre-set ratio threshold between an image frame corresponding to a video frame and its corresponding audio frame under asynchronous conditions. The above ratio threshold indicates that when the ratio of asynchronous video frames in all video frames is greater than or equal to the ratio threshold, it indicates that video fraud exists in the video. At the same time, when the ratio of asynchronous video frames in all video frames is less than the above ratio threshold, it cannot be determined that video fraud exists in the video.

[0057] Based on the setting of the above-mentioned preset asynchronous ratio threshold condition, it is determined whether the image frame of the corresponding video frame and its corresponding audio frame are asynchronous according to the predicted probability delay distribution. When the image frame of the video frame and its corresponding audio frame are synchronized, the predicted probability delay distribution should be zero. When the image frame of the video frame and its corresponding audio frame are not synchronized (i.e., asynchronous), the predicted probability delay distribution should not be zero.

[0058] When the image frame of the video frame is synchronized with its corresponding audio frame, that is, when the predicted probability delay distribution is zero, it means that the video signal has no time lag and no time delay, that is, zero delay. In this case, the asynchronous ratio of the video frames corresponding to the asynchronous image frames and their corresponding audio frames in all video frames is not counted. Similarly, when the image frame of the video frame is not synchronized with its corresponding audio frame, that is, asynchronous, the asynchronous ratio of the video frames corresponding to the asynchronous image frames and their corresponding audio frames in all video frames is counted. The asynchronous ratio represents the ratio of asynchronous video frames to all video frames.

[0059] Then determine whether the asynchronous ratio meets the preset asynchronous ratio threshold condition. If the asynchronous ratio meets the preset asynchronous ratio threshold condition, determine that the predicted probability delay distribution meets the preset probability delay distribution condition, and determine that the video contains video fraud. If the asynchronous ratio does not meet the preset asynchronous ratio threshold condition, do not determine that the predicted probability delay distribution meets the preset probability delay distribution condition, and do not determine that the video contains video fraud. In this case, the video may or may not contain fraud.

[0060] The embodiments of the present application determine whether a video contains fraud from the perspective of video frame asynchrony by setting a preset asynchronous ratio threshold condition. This can be based on the video itself and fully utilize the transmission information of image frames and their corresponding audio frames corresponding to different video frames contained in the video to detect video fraud. This can improve the accuracy of video fraud detection. For example, in financial technology, for credit business in financial business, it can detect whether the user is compliant and whether there is fraud through the images and audio contained in the video provided by the user, which can improve the accuracy of video fraud detection.

[0061] In one embodiment, determining whether the asynchronous ratio satisfies a preset asynchronous ratio threshold condition includes:

[0062] Determine whether the asynchronous ratio is greater than or equal to a preset first asynchronous ratio threshold;

[0063] When the asynchronous ratio is greater than or equal to a preset first asynchronous ratio threshold, determining that the asynchronous ratio satisfies a preset asynchronous ratio threshold condition;

[0064] When the asynchronous ratio is less than a preset first asynchronous ratio threshold, it is not determined that the asynchronous ratio satisfies the preset asynchronous ratio threshold condition.

[0065] Explanatoryly, a first asynchronous ratio threshold is preset, that is, the first asynchronous ratio threshold is preset, and the preset first asynchronous ratio threshold represents a threshold of the ratio of asynchronous video frames in all video frames. When the ratio of asynchronous video frames in all video frames is greater than or equal to the preset first asynchronous ratio threshold, it indicates that video fraud exists in the video. At the same time, when the ratio of asynchronous video frames in all video frames is less than the preset first asynchronous ratio threshold, it cannot be determined that there is fraud in the video, nor can it be determined that there is no fraud in the video.

[0066] Based on the setting of the above-mentioned preset first asynchronous ratio threshold, it is determined whether the asynchronous ratio is greater than or equal to the preset first asynchronous ratio threshold. When the asynchronous ratio is greater than or equal to the preset first asynchronous ratio threshold, it is determined that the asynchronous ratio meets the preset asynchronous ratio threshold condition, indicating that the proportion of asynchronous video frames in all video frames is large, and then it is determined that the video is fraudulent. When the asynchronous ratio is less than the preset first asynchronous ratio threshold, it is not determined that the asynchronous ratio meets the preset asynchronous ratio threshold condition, and then it is not determined that the video is fraudulent. In this case, the video may be fraudulent or may not be fraudulent.

[0067] In the embodiments of the present application, a first asynchronous ratio threshold is preset and compared with the first asynchronous ratio threshold to determine whether the video contains fraud. When the asynchronous ratio is greater than or equal to the first asynchronous ratio threshold, the asynchronous ratio is determined to satisfy the preset asynchronous ratio threshold condition, and then the video is determined to contain fraud. This allows for accurate judgment of video fraud and improves the accuracy of video fraud detection. For example, in financial technology, for credit business in financial services, it is possible to detect whether the user is compliant and whether there is fraud through the images and audio contained in the video provided by the user, thereby improving the accuracy of video fraud detection.

[0068] In one embodiment, see Figure 4 , Figure 4 This is a schematic diagram of the second sub-flow of the video fraud detection method provided in the embodiment of the present application. Figure 4As shown, in this embodiment, judging whether the asynchronous ratio meets the preset asynchronous ratio threshold condition further includes:

[0069] S41. When the asynchronous ratio is less than a preset first asynchronous ratio threshold, determine whether the asynchronous ratio is greater than or equal to a preset second asynchronous ratio threshold, wherein the preset second asynchronous ratio threshold is less than the preset first asynchronous ratio threshold;

[0070] S42: when the asynchronous ratio is less than a preset second asynchronous ratio threshold, determining that there is no video fraud in the video;

[0071] S43: extracting, when the asynchronous ratio is greater than or equal to a preset second asynchronous ratio threshold, an image frame corresponding to each video frame of the video and its corresponding audio frame based on a preset video signal delay estimation model;

[0072] S44, fusing the image frame, the audio frame and the combined feature to obtain a target comprehensive feature corresponding to the video frame;

[0073] S45, based on the preset probability prediction model, all the target comprehensive features and their corresponding time sequence, and according to the target comprehensive features corresponding to the previous video frame, re-predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame, and obtain a new predicted probability delay distribution;

[0074] S46. Re-determine whether the video contains video fraud based on the new predicted probability delay distribution.

[0075] Explanatoryly, a second asynchronous ratio threshold is preset, that is, a second asynchronous ratio threshold is preset, and the preset second asynchronous ratio threshold is smaller than the preset first asynchronous ratio threshold, and the preset second asynchronous ratio threshold means the following: when the ratio of asynchronous video frames in all video frames is smaller than the threshold, it can be determined that the video does not have fraud, and when the ratio of asynchronous video frames in all video frames is greater than or equal to the threshold, it cannot be determined that the video does have fraud, that is, the preset second asynchronous ratio threshold and the preset first asynchronous ratio threshold form an interval {preset second asynchronous ratio threshold, preset first asynchronous ratio threshold}, and it cannot be determined whether the video does have fraud within the above interval, but it can be determined that the video does have fraud outside the upper limit of the interval, and it can be determined that the video does not have fraud below the lower limit of the interval.

[0076] Based on the setting of the above-mentioned preset second asynchronous ratio threshold, when the asynchronous ratio is less than the preset first asynchronous ratio threshold, it is determined whether the asynchronous ratio is greater than or equal to the preset second asynchronous ratio threshold, and the preset second asynchronous ratio threshold is less than the preset first asynchronous ratio threshold. When the asynchronous ratio is less than the preset second asynchronous ratio threshold, it indicates that the proportion of asynchronous video frames in all video frames is small, and it is determined that there is no video fraud in the video. Moreover, when the asynchronous ratio is greater than or equal to the preset second asynchronous ratio threshold, based on the preset video signal delay estimation model, the image frames corresponding to each video frame of the video and their corresponding audio frames are extracted. The image frames corresponding to the video frames contained in the video can be extracted using the preset video signal delay estimation model including but not limited to the Open-Clip image Encoder model to extract the image features of each frame as a vector including but not limited to 2048 dimensions, and the audio frames corresponding to each video frame of the video are extracted including but not limited to MFCC features (Mel Frequency Cepstrum in English). Coefficient, Mel frequency cepstrum coefficient), fuse the image frame, audio frame and combined features, which can be achieved by concatenating the vectors corresponding to the image frame, audio frame and combined features into a new vector for fusion, to obtain the target comprehensive features corresponding to the video frame, and then re-predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame based on the preset probability prediction model and all the target comprehensive features and their corresponding time sequence, and according to the target comprehensive features corresponding to the previous video frame, re-predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame, to obtain a new predicted probability delay distribution, and re-judge whether the video involves video fraud based on the new predicted probability delay distribution, to cycle the above process of determining whether there is fraud.

[0077] In an embodiment of the present application, a preset second asynchronous ratio threshold is set, and the asynchronous ratio is compared with the preset second asynchronous ratio threshold to determine whether the video is fraudulent. When the asynchronous ratio is greater than or equal to the preset second asynchronous ratio threshold, and the asynchronous ratio is less than the preset first asynchronous ratio threshold, the image frame and audio frame corresponding to the video frame are fused with the combined features, and more dimensional information of the image frame is used to re-determine whether the video is fraudulent. Although the efficiency is slightly reduced compared to using the fused features alone to judge video fraud, the accuracy of video fraud judgment can be improved, thereby comprehensively predicting different situations of probability delay distribution, making corresponding judgments on video fraud, and improving the accuracy of video fraud detection. For example, in financial technology, for credit business in financial business, the image and audio contained in the video provided by the user can be used to detect whether the user is compliant and whether there is fraud, which can improve the accuracy of video fraud detection and thereby reduce the fraud risk of credit business, including the legitimate interests of all parties.

[0078] In one embodiment, see Figure 5 , Figure 5 This is a schematic diagram of the third sub-flow of the video fraud detection method provided in the embodiment of the present application. Figure 5 As shown, in this embodiment, judging whether the predicted probability delay distribution meets the preset probability delay distribution condition includes:

[0079] S51, judging whether an image frame of a corresponding video frame and its corresponding audio frame are synchronized according to the predicted probability delay distribution;

[0080] S52: when the image frame of the video frame and the corresponding audio frame are asynchronous, the synchronization ratio of the video frame corresponding to the synchronized image frame and the corresponding audio frame in all the video frames is not counted;

[0081] S53, when the image frame of the video frame is synchronized with the corresponding audio frame, counting the synchronization ratio of the video frame corresponding to the synchronized image frame and the corresponding audio frame in all video frames;

[0082] S54, determining whether the synchronization ratio meets a preset synchronization ratio threshold condition;

[0083] S55. When the synchronization ratio satisfies a preset synchronization ratio threshold condition, determining that the predicted probability delay distribution satisfies a preset probability delay distribution condition;

[0084] S56: When the synchronization ratio does not satisfy a preset synchronization ratio threshold condition, it is not determined that the predicted probability delay distribution satisfies a preset probability delay distribution condition.

[0085] Explanatoryly, a synchronization ratio threshold condition is pre-set, that is, a preset synchronization ratio threshold condition, which indicates a pre-set ratio threshold of an image frame corresponding to a video frame and its corresponding audio frame in a synchronized state. The above ratio threshold indicates that when the ratio of synchronized video frames in all video frames is less than or equal to the ratio threshold, it indicates that video fraud exists in the video. At the same time, when the ratio of synchronized video frames in all video frames is greater than the above ratio threshold, it cannot be determined that video fraud exists in the video.

[0086] Based on the setting of the above-mentioned preset synchronization ratio threshold condition, according to the predicted probability delay distribution, it is determined whether the image frame of the corresponding video frame and its corresponding audio frame are synchronized. When the image frame of the video frame and its corresponding audio frame are synchronized, the predicted probability delay distribution should be zero. When the image frame of the video frame and its corresponding audio frame are not synchronized (i.e., asynchronous), the predicted probability delay distribution should not be zero.

[0087] When the image frame of the video frame is synchronized with its corresponding audio frame, that is, when the predicted probability delay distribution is zero, it means that the video signal has no time lag and no time delay, that is, zero delay. In this case, the synchronization ratio of the synchronized image frame and the video frame corresponding to its corresponding audio frame in all video frames is counted. Similarly, when the image frame of the video frame is not synchronized with its corresponding audio frame, that is, asynchronous, the synchronization ratio of the synchronized image frame and the video frame corresponding to its corresponding audio frame in all video frames is not counted. The synchronization ratio represents the ratio of synchronized video frames to all video frames.

[0088] Then determine whether the synchronization ratio meets the preset synchronization ratio threshold condition. If the synchronization ratio meets the preset synchronization ratio threshold condition, determine that the predicted probability delay distribution meets the preset probability delay distribution condition, and determine that the video contains video fraud. If the synchronization ratio does not meet the preset synchronization ratio threshold condition, do not determine that the predicted probability delay distribution meets the preset probability delay distribution condition, and do not determine that the video contains video fraud. In this case, the video may or may not contain fraud.

[0089] It should be noted that judging whether a video is fraudulent from the perspective of video frame synchronization and from the perspective of video frame asynchrony are similar to the front and back of a coin. Therefore, fraud detection is performed on the video from the perspective of video frame synchronization. In addition to the technical solutions described in the above-mentioned embodiments of the present application, the concepts and principles of the technical solutions related to the preset second asynchronous ratio threshold described in the embodiments of the present application from the perspective of video frame asynchrony, including but not limited to, can also be applied to fraud detection of the video from the perspective of video frame synchronization. Based on the above-mentioned technical solutions from the perspective of video frame asynchrony, various equivalent adaptive modifications or replacements can be easily conceived from the perspective of video frame synchronization, and these modifications or replacements should all be covered within the scope of protection of the present application.

[0090] The embodiments of the present application determine whether a video contains fraud from the perspective of video frame synchronization by setting a preset synchronization ratio threshold condition. This can be based on the video itself and fully utilize the transmission information of image frames and their corresponding audio frames corresponding to different video frames contained in the video to implement video fraud detection. This can improve the accuracy of video fraud detection. For example, in financial technology, for credit business in financial business, it can detect whether the user is compliant and whether there is fraud through the images and audio contained in the video provided by the user. This can improve the accuracy of video fraud detection.

[0091] It should be noted that the video fraud detection method described in the above embodiments can recombine the technical features contained in different embodiments as needed to obtain a combined implementation plan, but all of them are within the scope of protection required by this application.

[0092] See also Figure 6 , Figure 6 Schematic block diagram of a video fraud detection device provided in an embodiment of the present application. Corresponding to the above-mentioned video fraud detection method, an embodiment of the present application also provides a video fraud detection device. Figure 6 As shown, the video fraud detection device includes a unit for executing the above-mentioned video fraud detection method, and the video fraud detection device can be configured in a computer device. Figure 6 The video fraud detection device 60 includes a first determination unit 61, a second determination unit 62, a first prediction unit 63, a first judgment unit 64 and a first determination unit 65.

[0093] The first determination unit 61 is used to respond to the video fraud detection instruction and determine the video corresponding to the user;

[0094] A second determining unit 62 is used to determine, based on a preset video signal delay estimation model, a combination feature corresponding to a combination between an image frame corresponding to each video frame of the video and its corresponding audio frame, to obtain a plurality of combination features of the video based on a time sequence;

[0095] A first prediction unit 63 is used to predict a probability delay distribution corresponding to a delay between an image frame corresponding to a next video frame and its corresponding audio frame based on a preset probability prediction model and all the combined features and their corresponding time sequence, and according to the combined features corresponding to the previous video frame, to obtain a predicted probability delay distribution;

[0096] A first judging unit 64 is used to judge whether the predicted probability delay distribution meets a preset probability delay distribution condition;

[0097] The first determination unit 65 is configured to determine whether video fraud exists in the video when the predicted probability delay distribution satisfies a preset probability delay distribution condition.

[0098] In one embodiment, the second determining unit 62 includes:

[0099] A first determining subunit, configured to determine a discrete delay feature corresponding to the video based on a preset video signal delay estimation model;

[0100] A second determining subunit, configured to determine a delay probability distribution corresponding to video frames included in the video according to the discrete delay feature;

[0101] The third determination subunit is used to determine the combination features corresponding to the combination between the image frame corresponding to the video frame and its corresponding audio frame according to the delay probability distribution, and obtain several combination features based on time sequence corresponding to several video frames contained in the video.

[0102] In one embodiment, the first determining unit 64 includes:

[0103] A first judgment subunit is used to judge whether an image frame of a corresponding video frame and its corresponding audio frame are asynchronous according to the predicted probability delay distribution;

[0104] A first statistics subunit is used for counting the asynchronous ratio of the video frames corresponding to the asynchronous image frames and their corresponding audio frames in all the video frames when the image frames of the video frames are asynchronous with respect to the corresponding audio frames;

[0105] A second judgment subunit is used to judge whether the asynchronous ratio meets a preset asynchronous ratio threshold condition;

[0106] The first determination subunit is configured to determine, when the asynchronous ratio satisfies a preset asynchronous ratio threshold condition, that the predicted probability delay distribution satisfies a preset probability delay distribution condition.

[0107] In one embodiment, the second determination subunit includes:

[0108] A third judgment subunit is used to judge whether the asynchronous ratio is greater than or equal to a preset first asynchronous ratio threshold;

[0109] The second determination subunit is configured to determine that the asynchronous ratio satisfies a preset asynchronous ratio threshold condition when the asynchronous ratio is greater than or equal to a preset first asynchronous ratio threshold.

[0110] In one embodiment, the second determination subunit further includes:

[0111] a fourth judgment subunit, configured to judge whether the asynchronous ratio is greater than or equal to a preset second asynchronous ratio threshold value when the asynchronous ratio is less than a preset first asynchronous ratio threshold value, wherein the preset second asynchronous ratio threshold value is less than the preset first asynchronous ratio threshold value;

[0112] A first extraction subunit is used to extract, when the asynchronous ratio is greater than or equal to a preset second asynchronous ratio threshold, an image frame corresponding to each video frame of the video and its corresponding audio frame based on a preset video signal delay estimation model;

[0113] A first fusion subunit is used to fuse the image frame, the audio frame and the combined feature to obtain a target comprehensive feature corresponding to the video frame;

[0114] A first prediction subunit is used to re-predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame based on the preset probability prediction model, all the target comprehensive features and their corresponding time sequence, and according to the target comprehensive features corresponding to the previous video frame, to obtain a new predicted probability delay distribution;

[0115] The re-judgment subunit is used to re-judgment whether the video contains video fraud according to the new predicted probability delay distribution.

[0116] In one embodiment, the second determination subunit further includes:

[0117] The third determination subunit is configured to determine that there is no video fraud in the video when the asynchronous ratio is less than a preset second asynchronous ratio threshold.

[0118] In one embodiment, the first determining unit 64 includes:

[0119] A fifth judgment subunit, configured to judge whether an image frame of a corresponding video frame is synchronized with its corresponding audio frame according to the predicted probability delay distribution;

[0120] A second statistical subunit is used for counting the synchronization ratio of the synchronized image frame and the video frame corresponding to the corresponding audio frame in all video frames when the image frame of the video frame is synchronized with the corresponding audio frame;

[0121] A sixth judgment subunit, used to judge whether the synchronization ratio meets a preset synchronization ratio threshold condition;

[0122] The fourth determination subunit is configured to determine, when the synchronization ratio satisfies a preset synchronization ratio threshold condition, that the predicted probability delay distribution satisfies a preset probability delay distribution condition.

[0123] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned video fraud detection device and each unit can refer to the corresponding description in the aforementioned method embodiment, and for the convenience and brevity of description, it will not be repeated here.

[0124] At the same time, the division and connection method of each unit in the above-mentioned video fraud detection device is only used for illustration. In other embodiments, the video fraud detection device can be divided into different units as needed, and the units in the video fraud detection device can also adopt different connection orders and methods to complete all or part of the functions of the above-mentioned video fraud detection device.

[0125] The video fraud detection device can be implemented in the form of a computer program. Figure 7 Runs on the computer device shown.

[0126] See also Figure 7 , Figure 7 1 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device 500 may be a computer device such as a desktop computer or a server, or may be a component or part in other devices.

[0127] See also Figure 7 The computer device 500 includes a processor 502, a memory and a network interface 505 connected via a system bus 501, wherein the memory may include a non-volatile storage medium 503 and an internal memory 504, and the memory may also be a volatile storage medium.

[0128] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, the processor 502 can execute the above-mentioned video fraud detection method.

[0129] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0130] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute the above-mentioned video fraud detection method.

[0131] The network interface 505 is used to communicate with other devices over the network. Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the present application scheme, and does not constitute a limitation on the computer device 500 to which the present application scheme is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. For example, in some embodiments, the computer device may only include a memory and a processor. In such an embodiment, the structure and function of the memory and the processor are the same as those of the present application scheme. Figure 7 The embodiments shown are consistent and will not be described again here.

[0132] Among them, the processor 502 is used to run a computer program 5032 stored in the memory to implement the following steps: respond to the instruction of video fraud detection and determine the video corresponding to the user; based on a preset video signal delay estimation model, determine the combination features corresponding to the combination between the image frame corresponding to each video frame of the video and its corresponding audio frame, and obtain a number of combination features of the video based on the time sequence; based on a preset probability prediction model and all the combination features and their corresponding time sequence, and according to the combination features corresponding to the previous video frame, predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame, and obtain the predicted probability delay distribution; determine whether the predicted probability delay distribution meets the preset probability delay distribution conditions; if the predicted probability delay distribution meets the preset probability delay distribution conditions, determine that the video contains video fraud.

[0133] In one embodiment, when the processor 502 determines the combination features corresponding to the combination between the image frame corresponding to each video frame of the video and its corresponding audio frame based on the preset video signal delay estimation model, and obtains a plurality of combination features of the video based on the time sequence, the processor 502 specifically implements the following steps:

[0134] Determining a discrete delay feature corresponding to the video based on a preset video signal delay estimation model;

[0135] Determining a delay probability distribution corresponding to a video frame included in the video according to the discrete delay feature;

[0136] According to the delay probability distribution, a combination feature corresponding to the combination between the image frame corresponding to the video frame and its corresponding audio frame is determined, and a plurality of combination features based on time sequence corresponding to a plurality of video frames contained in the video are obtained.

[0137] In one embodiment, when the processor 502 determines whether the predicted probability delay distribution meets the preset probability delay distribution condition, it specifically implements the following steps:

[0138] Determining whether an image frame of a corresponding video frame and its corresponding audio frame are asynchronous according to the predicted probability delay distribution;

[0139] In the case where the image frame of the video frame is asynchronous with the corresponding audio frame, counting the asynchronous ratio of the video frame corresponding to the asynchronous image frame and the corresponding audio frame in all the video frames;

[0140] Determine whether the asynchronous ratio meets a preset asynchronous ratio threshold condition;

[0141] In a case where the asynchronous ratio satisfies a preset asynchronous ratio threshold condition, it is determined that the predicted probability delay distribution satisfies a preset probability delay distribution condition.

[0142] In one embodiment, when the processor 502 determines whether the asynchronous ratio meets a preset asynchronous ratio threshold condition, the processor 502 specifically implements the following steps:

[0143] Determine whether the asynchronous ratio is greater than or equal to a preset first asynchronous ratio threshold;

[0144] When the asynchronous ratio is greater than or equal to a preset first asynchronous ratio threshold, it is determined that the asynchronous ratio meets a preset asynchronous ratio threshold condition.

[0145] In one embodiment, when the processor 502 determines whether the asynchronous ratio meets a preset asynchronous ratio threshold condition, it further specifically implements the following steps:

[0146] In the case where the asynchronous ratio is less than a preset first asynchronous ratio threshold, determining whether the asynchronous ratio is greater than or equal to a preset second asynchronous ratio threshold, wherein the preset second asynchronous ratio threshold is less than the preset first asynchronous ratio threshold;

[0147] When the asynchronous ratio is greater than or equal to a preset second asynchronous ratio threshold, extracting an image frame corresponding to each video frame of the video and its corresponding audio frame based on a preset video signal delay estimation model;

[0148] The image frame, the audio frame and the combined feature are fused to obtain a target comprehensive feature corresponding to the video frame;

[0149] Based on the preset probability prediction model, all the target comprehensive features and their corresponding time sequence, and according to the target comprehensive features corresponding to the previous video frame, re-predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame, and obtain a new predicted probability delay distribution;

[0150] Re-determine whether the video contains video fraud based on the new predicted probability delay distribution.

[0151] In one embodiment, when the processor 502 determines whether the asynchronous ratio meets a preset asynchronous ratio threshold condition, it further specifically implements the following steps:

[0152] When the asynchronous ratio is less than a preset second asynchronous ratio threshold, it is determined that there is no video fraud in the video.

[0153] In one embodiment, when the processor 502 determines whether the predicted probability delay distribution meets the preset probability delay distribution condition, it specifically implements the following steps:

[0154] Determining whether an image frame of a corresponding video frame and its corresponding audio frame are synchronized according to the predicted probability delay distribution;

[0155] In the case where the image frame of the video frame is synchronized with the corresponding audio frame, counting the synchronization ratio of the video frame corresponding to the synchronized image frame and the corresponding audio frame in all video frames;

[0156] Determining whether the synchronization ratio meets a preset synchronization ratio threshold condition;

[0157] In a case where the synchronization ratio satisfies a preset synchronization ratio threshold condition, it is determined that the predicted probability delay distribution satisfies a preset probability delay distribution condition.

[0158] It should be understood that in the embodiment of the present application, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0159] Those skilled in the art will appreciate that all or part of the processes in the method for implementing the above embodiment can be completed by a computer program, which can be stored in a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiment of the above method.

[0160] Therefore, the present application also provides a computer-readable storage medium. The computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, and the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the following steps:

[0161] A computer program product, when running on a computer, enables the computer to execute the steps of the video fraud detection method described in the above embodiments.

[0162] The computer-readable storage medium may be an internal storage unit of the aforementioned device, such as a hard disk or memory of the device. The computer-readable storage medium may also be an external storage device of the device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of the device and an external storage device.

[0163] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0164] The storage medium is a physical, non-transient storage medium, for example, it can be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk, etc., which can store computer programs.

[0165] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0166] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0167] The steps in the method of the embodiment of the present application can be adjusted in order, combined and deleted according to actual needs. The units in the device of the embodiment of the present application can be combined, divided and deleted according to actual needs. In addition, the functional units in the various embodiments of the present 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.

[0168] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for an electronic device (which can be a personal computer, terminal, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application.

[0169] The software tools or components not provided by our company that appear in the embodiments of this application are merely examples and do not represent actual use.

[0170] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A video fraud detection method, characterized in that: include: In response to the video fraud detection instruction, determine the video corresponding to the user; Based on a preset video signal delay estimation model, determining a combination feature corresponding to a combination between an image frame corresponding to each video frame of the video and its corresponding audio frame, and obtaining a plurality of combination features of the video based on a time sequence; Based on a preset probability prediction model and all the combined features and their corresponding time sequence, and according to the combined features corresponding to the previous video frame, predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame, and obtain the predicted probability delay distribution; Determining whether the predicted probability delay distribution meets a preset probability delay distribution condition; When the predicted probability delay distribution satisfies a preset probability delay distribution condition, it is determined that video fraud exists in the video.

2. The video fraud detection method according to claim 1, characterized in that: Based on a preset video signal delay estimation model, a combination feature corresponding to a combination between an image frame corresponding to each video frame of the video and its corresponding audio frame is determined to obtain a plurality of combination features of the video based on a time sequence, including: Determining a discrete delay feature corresponding to the video based on a preset video signal delay estimation model; Determining a delay probability distribution corresponding to a video frame included in the video according to the discrete delay feature; According to the delay probability distribution, a combination feature corresponding to the combination between the image frame corresponding to the video frame and its corresponding audio frame is determined, and a plurality of combination features based on time sequence corresponding to a plurality of video frames contained in the video are obtained.

3. The video fraud detection method according to claim 1 or 2, characterized in that: Determining whether the predicted probability delay distribution meets a preset probability delay distribution condition includes: Determining whether an image frame of a corresponding video frame and its corresponding audio frame are asynchronous according to the predicted probability delay distribution; In the case where the image frame of the video frame is asynchronous with the corresponding audio frame, counting the asynchronous ratio of the video frame corresponding to the asynchronous image frame and the corresponding audio frame in all the video frames; Determine whether the asynchronous ratio meets a preset asynchronous ratio threshold condition; In a case where the asynchronous ratio satisfies a preset asynchronous ratio threshold condition, it is determined that the predicted probability delay distribution satisfies a preset probability delay distribution condition.

4. The video fraud detection method according to claim 3, characterized in that: Determining whether the asynchronous ratio meets a preset asynchronous ratio threshold condition includes: Determine whether the asynchronous ratio is greater than or equal to a preset first asynchronous ratio threshold; When the asynchronous ratio is greater than or equal to a preset first asynchronous ratio threshold, it is determined that the asynchronous ratio meets a preset asynchronous ratio threshold condition.

5. The video fraud detection method according to claim 4, characterized in that: Determining whether the asynchronous ratio meets a preset asynchronous ratio threshold condition further includes: In the case where the asynchronous ratio is less than a preset first asynchronous ratio threshold, determining whether the asynchronous ratio is greater than or equal to a preset second asynchronous ratio threshold, wherein the preset second asynchronous ratio threshold is less than the preset first asynchronous ratio threshold; When the asynchronous ratio is greater than or equal to a preset second asynchronous ratio threshold, extracting an image frame corresponding to each video frame of the video and its corresponding audio frame based on a preset video signal delay estimation model; Fusing the image frame, the audio frame and the combined feature to obtain a target comprehensive feature corresponding to the video frame; Based on the preset probability prediction model, all the target comprehensive features and their corresponding time sequence, and according to the target comprehensive features corresponding to the previous video frame, re-predict the probability delay distribution corresponding to the delay between the image frame corresponding to the next video frame and its corresponding audio frame, and obtain a new predicted probability delay distribution; Re-determine whether the video contains video fraud based on the new predicted probability delay distribution.

6. The video fraud detection method according to claim 5, characterized in that: Determining whether the asynchronous ratio meets a preset asynchronous ratio threshold condition further includes: When the asynchronous ratio is less than a preset second asynchronous ratio threshold, it is determined that there is no video fraud in the video.

7. The video fraud detection method according to claim 1 or 2, characterized in that: Determining whether the predicted probability delay distribution meets a preset probability delay distribution condition includes: Determining whether an image frame of a corresponding video frame and its corresponding audio frame are synchronized according to the predicted probability delay distribution; In the case where the image frame of the video frame is synchronized with the corresponding audio frame, counting the synchronization ratio of the video frame corresponding to the synchronized image frame and the corresponding audio frame in all video frames; Determining whether the synchronization ratio meets a preset synchronization ratio threshold condition; In a case where the synchronization ratio satisfies a preset synchronization ratio threshold condition, it is determined that the predicted probability delay distribution satisfies a preset probability delay distribution condition.

8. A video fraud detection device, characterized in that: include: A first determination unit, configured to respond to a video fraud detection instruction and determine a video corresponding to a user; A second determining unit is used to determine, based on a preset video signal delay estimation model, a combination feature corresponding to a combination between an image frame corresponding to each video frame of the video and its corresponding audio frame, to obtain a plurality of combination features of the video based on a time sequence; A first prediction unit is used to predict a probability delay distribution corresponding to a delay between an image frame corresponding to a next video frame and its corresponding audio frame based on a preset probability prediction model and all the combined features and their corresponding time sequence, and according to the combined features corresponding to a previous video frame, to obtain a predicted probability delay distribution; A first judging unit, configured to judge whether the predicted probability delay distribution satisfies a preset probability delay distribution condition; The first determination unit is configured to determine whether video fraud exists in the video when the predicted probability delay distribution satisfies a preset probability delay distribution condition.

9. A computer device, characterized in that: The computer device comprises a memory and a processor connected to the memory; the memory is used to store a computer program; the processor is used to run the computer program to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 can be implemented.

Images