Fraud prediction method, prediction apparatus, mobile device, and storage medium

Abstract

Embodiments of the present application provide a fraud prediction method, a prediction apparatus, a mobile device, and a storage medium. The fraud prediction method is applied to a mobile device, and comprises: acquiring, from the mobile device, content to be detected; and determining a fraud probability of said content on the basis of a fraud prediction model. The fraud prediction method provided in the embodiments of the present application can predict the fraud probability by means of the fraud prediction model, thereby improving the intelligence level of fraud prediction.

Description

Fraud prediction methods, prediction devices, mobile devices, and storage media

[0001] This application claims priority to Chinese Patent Application No. 202411565131.4, filed on November 5, 2024, entitled "A Fraud Prediction Method, Prediction Device, Mobile Device and Storage Medium", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This specification relates to the field of computer security technology, specifically to a fraud prediction method, prediction device, mobile device, and storage medium. Background Technology

[0003] With the rapid development of information technology, telecommunications fraud is constantly evolving and upgrading. Telecommunications fraud not only causes economic losses to victims but also affects social harmony and stability. Therefore, combating telecommunications fraud is of paramount importance.

[0004] In related technologies, telecommunications operators can reduce the spread of fraudulent information by filtering and blocking suspicious calls and text messages. However, this method relies on blacklists and whitelists, which are usually outdated. Furthermore, fraud that is not conducted via phone or text message, such as in-app scams or website scams, is often difficult to detect. Technical issues

[0005] Scams that are not conducted via telephone or SMS, such as those within apps or websites, are often more difficult to detect. Technical solutions

[0006] This application provides a fraud prediction method, prediction device, mobile device, and storage medium, which can predict the probability of fraud through a fraud prediction model, thereby improving the intelligence level of fraud prediction.

[0007] In a first aspect, embodiments of this application provide a fraud prediction method applied to a mobile device, comprising: acquiring content to be detected in the mobile device; and determining the fraud probability of the content to be detected based on a fraud prediction model.

[0008] In some implementations, obtaining the content to be detected in the mobile device includes: obtaining the display interface of the mobile device at a preset frequency, and obtaining the content to be detected in the display interface.

[0009] In some implementations, the fraud prediction model includes a text recognition model and a content recognition model; determining the fraud probability of the content to be detected based on the fraud prediction model includes: performing text recognition on the content to be detected based on the text recognition model to obtain the text to be detected, and performing content recognition on the text to be detected based on the content recognition model to obtain the fraud probability.

[0010] In some implementations, the content recognition model includes a first recognition model and a second recognition model, wherein the computational complexity of the first recognition model is less than that of the second recognition model.

[0011] In some implementations, the step of performing content recognition on the text to be detected based on the content recognition model to obtain the fraud probability includes: obtaining a first fraud probability of the text to be detected based on the first recognition model; and using the first fraud probability as the fraud probability when the first fraud probability is less than a first threshold or greater than a second threshold, or obtaining the fraud probability of the text to be detected based on the second recognition model when the first fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold.

[0012] In some implementations, the fraud prediction model includes a first prediction model and a second prediction model, wherein the computational cost of the first prediction model is less than that of the second prediction model.

[0013] In some implementations, determining the fraud probability of the content to be detected based on the fraud prediction model includes: obtaining a second fraud probability of the content to be detected based on the first prediction model; and using the second fraud probability as the fraud probability when the second fraud probability is less than a first threshold or greater than a second threshold, or obtaining the fraud probability of the content to be detected based on the second prediction model when the second fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold.

[0014] Secondly, embodiments of this application provide a prediction device, the device comprising:

[0015] The acquisition module is used to acquire the content to be detected in the mobile device; and

[0016] The determination module is used to determine the probability of fraud of the content to be detected based on the fraud prediction model.

[0017] Thirdly, embodiments of this application provide a mobile device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the operations described in the first aspect.

[0018] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the operations described in the first aspect. Beneficial effects

[0019] This application provides a fraud prediction method, a prediction device, a mobile device, and a storage medium. The fraud prediction method, applied to a mobile device, includes: acquiring content to be detected in the mobile device; and determining the probability of fraud in the content to be detected based on a fraud prediction model. The beneficial effect of this application is that it improves the intelligence level of fraud prediction by predicting the probability of fraud through a fraud prediction model. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 shows a flowchart illustrating a fraud prediction method provided by some embodiments of this application.

[0022] Figure 2 shows a schematic flowchart of a process for determining the probability of fraud provided by some embodiments of this application.

[0023] Figure 3 illustrates a flowchart of multi-level content recognition provided by some embodiments of this application.

[0024] Figure 4 illustrates a flowchart of a multi-level fraud prediction process provided by some embodiments of this application.

[0025] Figure 5 shows a schematic diagram of the structure of a prediction device provided in some embodiments of this application.

[0026] Figure 6 shows a schematic diagram of the structure of a mobile device provided in some embodiments of this application.

[0027] Implementation methods of this application

[0028] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0029] Although the flowcharts show a logical order, in some cases the operations shown or described may be performed in a different order than that shown in the figures.

[0030] Figure 1 shows a flowchart illustrating a fraud prediction method provided in some embodiments of this application. This fraud prediction method can be applied to mobile devices, including, but not limited to, smartphones, tablets, smartwatches, game consoles, or similar devices. The specific flow of the fraud prediction method can be as follows:

[0031] S101. Obtain the content to be detected from the mobile device.

[0032] The content to be detected can be an object used to determine the probability of fraud. The specific form of the content to be detected can be various, such as images, text, or voice.

[0033] In some embodiments, the content to be detected may be an image. For example, the fraud prediction method may acquire images displayed on the mobile device's screen. Alternatively, the fraud method may acquire images from MMS or rich media communications received by the mobile device.

[0034] In some embodiments, the content to be detected may be text. For example, the fraud prediction method may acquire text messages received by the mobile device. Alternatively, the fraud prediction method may acquire chat text from instant messaging or social media software on the mobile device.

[0035] In some embodiments, the content to be detected may be voice. For example, the fraud prediction method may acquire voice recordings from a mobile device during telephone communications. Alternatively, the fraud prediction method may acquire voice recordings from instant messaging or social media software on a mobile device during real-time voice communications or voice messages.

[0036] S102. Determine the probability of fraud in the content to be detected based on the fraud prediction model.

[0037] The fraud prediction model can be a pre-trained machine learning model or a deep learning model. Before step S102, the fraud prediction method may further include fine-tuning the pre-trained model to obtain a pre-trained fraud prediction model. The fraud prediction model may include one or more models. For example, different models or the same model can be used to obtain fraud probabilities for different types of content to be detected. For example, one model or multiple models can be used to obtain fraud probabilities for the same type of content to be detected. The training method of the fraud prediction model will be described later in conjunction with specific embodiments.

[0038] Fraud probability can be used to evaluate the likelihood that the content to be detected contains fraudulent information. A higher fraud probability indicates a higher probability of containing fraudulent information, while a lower probability indicates a lower probability. Fraud probability can be a continuously distributed probability value, for example, any probability value within the range of 0% to 100%. It can also be a discretely distributed probability rating, for example, divided into high, medium, and low ratings. Furthermore, fraud probability can be used to determine whether something is a scam; for example, a high fraud probability indicates a scam, while a low fraud probability indicates no scam.

[0039] In some embodiments, after obtaining the fraud probability, corresponding actions can be performed based on the fraud probability. For example, the corresponding action could be to generate alerts with different risk levels based on the fraud probability. When the fraud probability is in the first probability range, the risk level is low, and only a pop-up alert is issued. When the fraud probability is in the second probability range, a confirmation dialog box pops up, requiring the user to confirm before closing the dialog box. When the fraud probability is in the third probability range, the risk level is high, and the user needs to answer security questions to remove usage restrictions. Of course, the probability range or alert method can be preset by the fraud prediction method or set by the user. For example, the corresponding action could be to block messages based on the fraud probability. Another example is to restrict user behavior based on the fraud probability, such as fund transfers.

[0040] The fraud prediction methods provided in some embodiments of this application can be applied to mobile devices to acquire content to be detected in the mobile device and determine the probability of fraud of the content to be detected based on a fraud prediction model. The fraud prediction methods provided in some embodiments of this application improve the intelligence level of fraud prediction by predicting the probability of fraud through a fraud prediction model.

[0041] As mentioned above, the content to be detected can be in the form of an image. In some embodiments, obtaining the content to be detected in the mobile device in operation S101 may include:

[0042] The display interface of the mobile device is acquired at a preset frequency, and the content to be detected in the display interface is also acquired.

[0043] The acquisition of the mobile device's display interface can be done at a preset frequency when the mobile device's screen is on. Acquiring the content to be detected within the display interface can involve acquiring an image of the entire display interface or an image of a portion of the display interface. For example, in some embodiments, the fraud prediction method can determine the location of the message pop-up based on user settings or the mobile device's model, and acquire an image of the message pop-up area.

[0044] The preset frequency can be set in the fraud prediction method or set by the user. The preset frequency can be a fixed frequency or an adjustable frequency. For example, in some embodiments, the fraud prediction method can adjust the preset frequency according to the phone's battery level or power consumption.

[0045] In some embodiments, the fraud prediction model for predicting fraud in image-based content to be detected may include a text recognition model and a content recognition model. Figure 2 illustrates a flowchart of determining fraud probability provided by some embodiments of this application. Determining the fraud probability of the content to be detected based on the fraud prediction model may include:

[0046] S201. Based on the character recognition model, perform text recognition on the content to be detected to obtain the text to be detected.

[0047] The content to be detected can be an image, and the text recognition model can be used to identify the text in the image. There can be one or more text recognition models.

[0048] In some embodiments, the text recognition model can be a cascaded model of a text detection model and a character recognition model. The text detection model can detect text regions in the image to be detected, and then the text region image is input into the character recognition model to output the characters in the text region image, i.e., the detected text. The text detection model can use general object detection models, such as the YOLO series models, the SSD model, the Faster R-CNN model, or improved models based on these models. The text detection model can also be a domain-specific detection model, such as the EAST detection model (Efficient and Accurate Scene Text Detector) or the CTPN model (Column Proposal Networks).

[0049] In some embodiments, the text recognition model can be an end-to-end text recognition model that can directly output the text to be detected based on the input image to be detected.

[0050] S202. Based on the content recognition model, perform content recognition on the text to be detected to obtain the fraud probability.

[0051] Content recognition models can be used to analyze and understand the text to be detected in order to obtain the probability of fraud.

[0052] In some embodiments, the content recognition model can be a natural language processing model. The natural language processing model can be a bag-of-words (BoW) model, an N-gram model, a hidden Markov model (HMM), a recurrent neural network (RNN), or a Transformer model.

[0053] To achieve better prediction results in fraud prediction tasks, the content recognition pre-trained model can be fine-tuned using data from the fraud prediction domain. To do this, training and testing sample sets can be constructed, where each sample is a piece of text content, and each sample corresponds to a label indicating whether fraudulent information exists in the sample. For example, if fraudulent information exists in the sample, the corresponding label is 1; if no fraudulent content exists in the sample, the corresponding label is 0.

[0054] In some embodiments, the text in character form needs to be converted into feature vectors before being input into the content recognition pre-training model. The model parameters are continuously adjusted based on the difference between the output of the training samples after passing through the content recognition pre-training model and the labels, in order to obtain a well-trained content recognition model.

[0055] The training process for the content recognition model can be performed on a server. The trained content recognition model can then be stored on a mobile device or in the cloud. Based on this model, the text to be detected or its feature vector can be input into the model, which then outputs the probability of fraud.

[0056] The fraud prediction method provided in some embodiments of this application, when predicting the fraud probability of content to be detected in the form of an image, first extracts the text in the image to be detected, and then performs content recognition on the extracted text to determine the fraud probability.

[0057] To improve the speed of fraud prediction, in some embodiments, multi-level recognition models can be used to predict fraud probabilities. For example, the content recognition model may include a first recognition model and a second recognition model, wherein the computational cost of the first recognition model is less than that of the second recognition model. For the same input data and the same computing device, the time spent using the first recognition model for inference is less than the time spent using the second recognition model for inference. Furthermore, at least on the test sample set, the accuracy or recall rate of the first recognition model is lower than that of the second recognition model.

[0058] Figure 3 illustrates a flowchart of multi-level content recognition provided by some embodiments of this application. Specifically, the process of performing content recognition on the text to be detected based on a content recognition model to obtain the fraud probability may include:

[0059] S301. Obtain the first fraud probability of the text to be detected based on the first recognition model.

[0060] The first fraud probability can be used to make a preliminary judgment on whether the text to be detected contains fraudulent information, and based on the result of this preliminary judgment, it can be determined whether a second recognition model is needed for further prediction.

[0061] S302. When the first fraud probability is less than the first threshold or greater than the second threshold, the first fraud probability shall be used as the fraud probability.

[0062] When the first fraud probability is low or high, the result obtained by the first identification model can be considered relatively reliable, that is, the first fraud probability can be used as the fraud probability.

[0063] S303. When the first fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold, the fraud probability of the text to be detected is obtained based on the second recognition model.

[0064] When the first fraud probability is in the middle range, it can be considered that the first recognition model has difficulty in judging whether the text to be detected contains fraudulent information. At this time, the second recognition model can be used to make further predictions and obtain the fraud probability.

[0065] It should be noted that the first recognition model and the second recognition model can be stored together on the mobile device; or the first recognition model and the second recognition model can be stored together on the cloud device; or the first recognition model can be stored on the mobile device and the second recognition model can be stored on the cloud device.

[0066] The fraud prediction method provided in some embodiments of this application can improve the prediction speed of fraud probability by using a multi-level content recognition model. When the results obtained using a lightweight first recognition model are relatively reliable, only a few cases require a second recognition model for secondary judgment.

[0067] In other embodiments, the fraud prediction model that performs fraud prediction on the image-based content to be detected can be an end-to-end image semantic understanding model. The image semantic understanding model can receive input in image form and analyze and understand whether fraudulent information exists in the input image to be detected.

[0068] Similarly, to achieve better prediction results, training and testing sample sets can be constructed. Each sample is an image, and each sample can correspond to a label indicating whether fraudulent information exists in the sample. For example, if fraudulent information exists in the sample, the label is 1; if no fraudulent content exists, the label is 0. Furthermore, if fraudulent information exists in the sample, the label could be "What type of fraud is occurring in the image?"; if no fraudulent content exists, the label could be "Nothing is happening." The model parameters are continuously adjusted based on the difference between the output of the image semantic understanding model after processing the training samples and the labels, resulting in a well-trained image semantic understanding model.

[0069] Similarly, the training process for the image semantic understanding model can be performed on a server. The trained model can be stored on a mobile device or in the cloud. Based on this model, an image to be detected can be input into it, and the model will output the probability of fraud.

[0070] Understandably, for other forms of content to be detected, the fraud prediction model can be another model that matches the input format. For example, for text-based content, the content recognition model described above can be used as a fraud prediction model to predict the probability of fraud. For example, for speech-based content, a speech recognition model can be used as a fraud prediction model to predict the probability of fraud.

[0071] Similar to two-stage (text recognition + content recognition) fraud prediction, single-stage fraud prediction can also use multi-level fraud prediction models to predict fraud probabilities. In some embodiments, the fraud prediction model may include a first prediction model and a second prediction model, wherein the computational cost of the first prediction model is less than that of the second prediction model. For the same input data and the same computing device, the time spent using the first prediction model for inference is less than the time spent using the second prediction model for inference. Furthermore, at least on the test sample set, the accuracy or recall rate of the first prediction model is lower than that of the second prediction model.

[0072] Figure 4 illustrates a flowchart of a multi-level fraud prediction process provided by some embodiments of this application. Specifically, determining the fraud probability of the content to be detected based on the fraud prediction model may include:

[0073] S401. Obtain the second fraud probability of the content to be detected based on the first prediction model.

[0074] The first prediction model can be a model that predicts the probability of fraud for the content to be detected, such as images, text, or audio. The second fraud probability can be used to make a preliminary judgment on whether the content to be detected contains fraudulent information, and based on the result of this preliminary judgment, decide whether to use the second prediction model for further prediction.

[0075] S402. When the second fraud probability is less than the first threshold or greater than the second threshold, the second fraud probability shall be used as the fraud probability.

[0076] When the second fraud probability is small or large, the result obtained by the first prediction model can be considered more reliable, that is, the second fraud probability can be used as the fraud probability.

[0077] S403. When the second fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold, the fraud probability of the content to be detected is obtained based on the second prediction model.

[0078] When the second fraud probability is in the middle range, it can be considered that the first prediction model has difficulty in judging whether the content to be detected contains fraudulent information. At this time, the second prediction model can be used to make further predictions and obtain the fraud probability.

[0079] It should be noted that the first prediction model and the second prediction model can be stored together on the mobile device; or the first prediction model and the second prediction model can be stored together on the cloud device; or the first prediction model can be stored on the mobile device and the second prediction model can be stored on the cloud device.

[0080] The fraud prediction method provided in some embodiments of this application can improve the speed of fraud probability prediction by using a multi-level fraud prediction model. When the results obtained using a lightweight first prediction model are relatively reliable, only a few cases require a second prediction model for secondary judgment.

[0081] Figure 5 shows a schematic diagram of the structure of a prediction device provided in some embodiments of this application. The prediction device 500 can be applied to a mobile device and may specifically include:

[0082] The acquisition module 501 can be used to acquire the content to be detected in the mobile device.

[0083] The determination module 502 can be used to determine the probability of fraud of the content to be detected based on the fraud prediction model.

[0084] In some embodiments of this application, the acquisition module 501 first acquires the content to be detected from the mobile device. Next, the determination module 502 determines the probability of fraud in the content to be detected based on a fraud prediction model. The prediction device 500 provided in some embodiments of this application can predict the probability of fraud through a fraud prediction model, thereby improving the intelligence level of fraud prediction.

[0085] In some embodiments, the acquisition module 501 may include a first acquisition unit and a second acquisition unit. The first acquisition unit may be used to acquire the display interface of the mobile device at a preset frequency. The second acquisition unit may be used to acquire the content to be detected in the display interface.

[0086] In some embodiments, the determining module 502 may include a text recognition unit and a content recognition unit. The text recognition unit can be used to perform text recognition on the content to be detected based on a character recognition model to obtain the text to be detected. The content recognition unit can be used to perform content recognition on the text to be detected based on a content recognition model to obtain the fraud probability.

[0087] In some embodiments, the content recognition unit may include a first recognition subunit and a second recognition subunit. The first recognition subunit may be used to obtain a first fraud probability of the text to be detected based on a first recognition model. The second recognition subunit may be used to use the first fraud probability as the fraud probability when the first fraud probability is less than a first threshold or greater than a second threshold, or to obtain the fraud probability of the text to be detected based on a second recognition model when the first fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold.

[0088] In some embodiments, the determining module 502 may include a first prediction unit and a second prediction unit. The first prediction unit may be used to obtain a second fraud probability of the content to be detected based on a first prediction model. The second prediction unit may be used to use the first fraud probability as the fraud probability when the second fraud probability is less than a first threshold or greater than a second threshold, or to obtain the fraud probability of the content to be detected based on a second prediction model when the second fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold.

[0089] Furthermore, this application also provides a mobile device. Figure 6 shows a schematic diagram of the structure of a mobile device provided in some embodiments of this application. The mobile device 600 includes, but is not limited to, smartphones, tablets, smartwatches, game consoles, or similar devices. Specifically:

[0090] The mobile device 600 may include components such as a processor 601 with one or more processing cores, and a memory 602 with one or more computer-readable storage media. Those skilled in the art will understand that the structure of the mobile device 600 shown in FIG. 6 does not constitute a limitation on the mobile device 600, and it may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0091] The processor 601 is the control center of the mobile device 600. It connects various parts of the mobile device 600 via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 602, and by calling data stored in the memory 602, it performs various functions and processes data of the mobile device 600, thereby providing overall monitoring of the mobile device 600. Optionally, the processor 601 may include one or more processing cores; preferably, the processor 601 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 601.

[0092] The memory 602 can be used to store software programs and modules. The processor 601 executes various functional applications and data processing by running the software programs and modules stored in the memory 602. The memory 602 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile device 600, etc. In addition, the memory 602 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 602 may also include a memory controller to provide the processor 601 with access to the memory 602.

[0093] Specifically, in this embodiment, the processor 601 in the mobile device 600 will load the executable files corresponding to the processes of one or more applications into the memory 602 according to the following instructions, and the processor 601 will run the applications stored in the memory 602, thereby realizing the operation in any fraud prediction method provided in this application embodiment.

[0094] The specific process of the mobile device 600 performing the fraud prediction method can be seen in the descriptions in Figures 2 to 5, and will not be repeated here.

[0095] Those skilled in the art will understand that all or part of the operations in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0096] Therefore, this application provides a computer-readable storage medium storing a computer program that can be loaded by a processor to perform operations in any of the fraud prediction methods provided in this application.

[0097] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0098] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0099] Since the instructions stored in the computer-readable storage medium can execute the operations in any of the fraud prediction methods provided in this application, the beneficial effects that any of the fraud prediction methods provided in this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0100] It should be noted that the fraud prediction method provided in this application strictly adheres to the principle of user privacy protection during the data acquisition and prediction process. Before collecting any personal information, a clear and detailed explanation will be provided to the user, including but not limited to the data type, purpose of collection, scope of use, and retention period. The user must indicate that they have read and understood the above information and voluntarily authorize the data collection and specific use purposes by clearly and affirmatively performing actions (such as clicking the confirmation button or turning on the function switch).

[0101] The above provides a detailed description of a fraud prediction method, prediction device, mobile device, and storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A fraud prediction method applied to mobile devices, wherein, include: Obtain the content to be detected from the mobile device; The probability of fraud in the content to be detected is determined based on a fraud prediction model.

2. The method as described in claim 1, wherein, The step of obtaining the content to be detected in the mobile device includes: The display interface of the mobile device is acquired at a preset frequency, and the content to be detected in the display interface is acquired.

3. The method as described in claim 2, wherein, The fraud prediction model includes a text recognition model and a content recognition model; The process of determining the fraud probability of the content to be detected based on the fraud prediction model includes: Based on the aforementioned character recognition model, text recognition is performed on the content to be detected to obtain the text to be detected. Based on the content recognition model, the text to be detected is subjected to content recognition to obtain the fraud probability.

4. The method of claim 3, wherein, The content recognition model includes a first recognition model and a second recognition model, wherein the computational complexity of the first recognition model is less than that of the second recognition model.

5. The method of claim 4, wherein, The step of performing content recognition on the text to be detected based on the content recognition model to obtain the fraud probability includes: The first fraud probability of the text to be detected is obtained based on the first recognition model; When the first fraud probability is less than the first threshold or greater than the second threshold, the first fraud probability is used as the fraud probability; or when the first fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold, the fraud probability of the text to be detected is obtained based on the second recognition model.

6. The method of claim 1, wherein, The fraud prediction model includes a first prediction model and a second prediction model, wherein the computational complexity of the first prediction model is less than that of the second prediction model.

7. The method of claim 6, wherein, The process of determining the fraud probability of the content to be detected based on the fraud prediction model includes: The second fraud probability of the content to be detected is obtained based on the first prediction model; When the second fraud probability is less than the first threshold or greater than the second threshold, the second fraud probability is used as the fraud probability; or when the second fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold, the fraud probability of the content to be detected is obtained based on the second prediction model.

8. The method of claim 2, wherein, The step of acquiring the content to be detected in the display interface includes: acquiring an image of the entire display interface, or acquiring an image of a partial area of ​​the display interface.

9. The method of claim 1, wherein, The step of obtaining the content to be detected in the mobile device further includes: obtaining the text message received by the mobile device, or obtaining the voice of the mobile device during telephone communication.

10. The method of claim 1, wherein, After determining the fraud probability of the content to be detected based on the fraud prediction model, the method further includes: generating alerts with different risk levels based on the fraud probability.

11. The method of claim 6, wherein, The time spent in reasoning using the first prediction model for the same input data is less than the time spent in reasoning using the second prediction model.

12. The method of claim 1, wherein, The fraud probability is used to evaluate the likelihood that the content to be detected contains fraudulent information; the higher the fraud probability, the higher the likelihood that the content to be detected contains fraudulent information; the lower the fraud probability, the lower the likelihood that the content to be detected contains fraudulent information.

13. A prediction device, wherein, The device includes: The acquisition module is used to acquire the content to be detected in the mobile device; The determination module is used to determine the probability of fraud of the content to be detected based on the fraud prediction model.

14. The apparatus of claim 13, wherein, The acquisition module includes a first acquisition unit and a second acquisition unit; wherein, the first acquisition unit is used to acquire the display interface of the mobile device at a preset frequency; and the second acquisition unit is used to acquire the content to be detected in the display interface.

15. The apparatus of claim 14, wherein, The determining module includes a text recognition unit and a content recognition unit; wherein, the text recognition unit is used to perform text recognition on the content to be detected based on the text recognition model to obtain the text to be detected; the content recognition unit is used to perform content recognition on the text to be detected based on the content recognition model to obtain the fraud probability.

16. The apparatus of claim 15, wherein, The content recognition model includes a first recognition model and a second recognition model, wherein the computational complexity of the first recognition model is less than that of the second recognition model.

17. The apparatus of claim 16, wherein, The content recognition unit includes a first recognition subunit and a second recognition subunit; wherein, the first recognition subunit is used to obtain a first fraud probability of the text to be detected based on the first recognition model; the second recognition subunit is used to use the first fraud probability as the fraud probability when the first fraud probability is less than a first threshold or greater than a second threshold, or to obtain the fraud probability of the text to be detected based on the second recognition model when the first fraud probability is greater than or equal to the first threshold and less than or equal to the second threshold.

18. The apparatus of claim 13, wherein, The fraud prediction model includes a first prediction model and a second prediction model, wherein the computational complexity of the first prediction model is less than that of the second prediction model.

19. A mobile device, wherein, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, performs the operations described in any one of claims 1-12.

20. A computer-readable storage medium, wherein, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the operations described in any one of claims 1-12.

Images