Method for authenticating a user identity based on a touch operation

Abstract

The application provides a method for authenticating user identity based on touch operation, comprising a training phase and an authentication phase; the training phase comprises: a touch interface generating a plurality of training touch parameters, and a processor generating a training heat map according to the training touch parameters; the authentication phase comprises: the touch interface generating a plurality of test touch parameters, the processor generating a test heat map according to the test touch parameters, the processor comparing the test heat map with the training heat map to generate an error map, and the processor generating one of an authentication pass signal and an authentication failure signal according to the error map; the application continuously and non-invasively monitors and authenticates the current user; the application is data-driven and adaptive, can adapt to the changing mode of touch operation of the user, and improves the frequency of user identity authentication.

Description

Technical Field

[0001] This invention relates to the field of user authentication technology, and in particular to a method for authenticating user identity based on touch operation. Background Technology

[0002] User authentication allows a computing device (such as a laptop or smartphone) to authenticate the identity of a person attempting to use the device. The authentication result is a binary pass or fail. Successful authentication means that the user is indeed the person he / she claims to be, while authentication failure means that the user is not the person he / she claims to be.

[0003] Passwords are a common mechanism for user authentication. However, if a password is stolen, the user who obtains the password gains control of the account. Even if the password is not stolen, when an authenticated user temporarily leaves the computing device they have logged into with their password, other unauthorized users can take the opportunity to access that computing device; this situation is called an insider attack.

[0004] When passwords are at risk of being stolen, biometric authentication can be used to address this issue. Biometric authentication methods compare a user's biometric features with records in a database. Typical biometric technologies include fingerprints, facial recognition, retinal scanning, iris scanning, and voice recognition. Because biometric authentication uses unique biometric features for authentication, it is difficult to copy or steal.

[0005] However, biometrics are highly sensitive and privacy-sensitive. A potential problem with biometric authentication is that systems can be used for functions beyond their original intent, a phenomenon known as function creep. For example, a biometric authentication system installed in the workplace to prevent general employees from accessing confidential areas could be used by system administrators to track the locations visited by individual employees without prior notification, thereby infringing on employees' privacy rights.

[0006] Furthermore, authenticating user identity using passwords or biometrics is typically a one-time process. Even if repeated authentication is required, it should be done at intervals; otherwise, it would cause additional inconvenience to users. In other words, existing user authentication methods lack continuity and non-intrusiveness. Summary of the Invention

[0007] In view of the shortcomings of the prior art described above, the purpose of this invention is to propose a method for authenticating user identity based on touch operation. This method achieves continuous and non-intrusive user authentication by observing the user's touch operation patterns. If the touch pattern of a new user differs significantly from that of an existing user, the new user will be identified as an imposter, and their access will be prohibited. The method proposed in this invention can be called TouchPrint, because the dynamic usage of touch operations can authenticate the user's identity like a fingerprint.

[0008] To achieve the above and other related objectives, the present invention provides a method for authenticating user identity based on touch operation, comprising: a training phase, including: a touch interface generating multiple training touch parameters; and a processor generating a training heatmap based on the training touch parameters; and an authentication phase, including: the touch interface generating multiple test touch parameters; the processor generating a test heatmap based on the test touch parameters; the processor comparing the test heatmap and the training heatmap to generate an error map; and the processor generating one of an authentication pass signal and an authentication failure signal based on the error map.

[0009] This invention provides a method for authenticating user identity based on touch operation, comprising: a training phase, including: a touch interface generating multiple training touch parameters; and a processor training a neural network model based on the training touch parameters; and an authentication phase, including: the touch interface generating multiple test touch parameters; the processor inputting the test touch parameters into the neural network model to generate a predicted value; and the processor calculating the error between the predicted value and an actual value to generate one of an authentication pass signal and an authentication failure signal.

[0010] This invention provides a method for authenticating user identity based on touch operation, comprising: a training phase, including: a touch interface generating multiple training touch parameters; and a processor generating a training heatmap and a training neural network model based on the training touch parameters; and an authentication phase, including: the touch interface generating multiple test touch parameters; the processor generating a test heatmap based on the test touch parameters; the processor comparing the test heatmap and the training heatmap to generate an error map; the processor calculating a first error value based on the error map; the processor inputting the test touch parameters into the neural network model to generate a predicted value; and the processor generating one of an authentication pass signal and an authentication failure signal based on the first error value and a second error value, wherein the second error value is associated with the predicted value and an actual value.

[0011] In summary, the method for authenticating user identity based on touch operation proposed in this invention has the following advantages:

[0012] 1. Continuously and non-intrusively monitor and authenticate current users;

[0013] 2. It adopts a data-driven approach and is adaptive, able to adapt to the ever-changing touch operation patterns of users; and

[0014] 3. Increase the frequency of user authentication, and perform authentication when existing user authentication mechanisms (such as passwords or biometrics) are not in use.

[0015] It is worth noting that the method proposed in this invention does not replace existing user authentication mechanisms, but rather supplements and enhances the security of existing mechanisms to prevent internal attacks. In other words, it can detect impersonators even when the computing device is unlocked.

[0016] The foregoing description of the disclosure and the following description of embodiments are intended to demonstrate and explain the spirit and principles of the present invention, and to provide a further explanation of the scope of the patent application of the present invention. Attached Figure Description

[0017] Figure 1 The flowchart shown is a first embodiment of the method for authenticating user identity based on touch operation according to the present invention;

[0018] Figure 2 The flowchart shown is a first embodiment of the training and certification phases of the present invention;

[0019] Figure 3 The image shows a location heatmap of two different users of the present invention based on mobile operations in one embodiment.

[0020] Figure 4 The image shows a positional heatmap of two different users of the present invention based on a pressing operation in one embodiment.

[0021] Figure 5 The image shows a positional heatmap of two different users of the present invention based on a scrolling operation in one embodiment.

[0022] Figure 6 The image shows a position heatmap and a velocity heatmap of the same user based on mobile operation in one embodiment of the present invention.

[0023] Figure 7 The flowchart shown is a second embodiment of the method for authenticating user identity based on touch operation according to the present invention;

[0024] Figure 8 The flowchart shown is a second embodiment of the training and certification phases of the present invention;

[0025] Figure 9 The flowchart shown is a third embodiment of the method for authenticating user identity based on touch operation according to the present invention;

[0026] Figure 10 The flowchart shown is a fourth embodiment of the method for authenticating user identity based on touch operation according to the present invention;

[0027] Figure 11 The flowchart shown is a representation of the training and certification phases of the present invention in the fourth embodiment.

[0028] Figure 12 The flowchart shown is a fifth embodiment of the method for authenticating user identity based on touch operation according to the present invention;

[0029] Figure 13 The flowchart shown is a representation of the training and certification phases of the present invention in the fifth embodiment.

[0030] Figure 14 The flowchart shown is a sixth embodiment of the method for authenticating user identity based on touch operation according to the present invention;

[0031] Figure 15 The flowchart shown is a method for authenticating user identity based on touch operation according to the seventh embodiment of the present invention;

[0032] Figure 16 The flowchart shown is a representation of the training and certification phases of the present invention in the seventh embodiment;

[0033] Figure 17 The flowchart shown is a method for authenticating user identity based on touch operation according to the eighth embodiment of the present invention;

[0034] Figure 18A and Figure 18B The flowcharts for the training and authentication phases of the present invention in the eighth embodiment are shown respectively; and

[0035] Figure 19 The flowchart shown is a method for authenticating user identity based on touch operation according to the ninth embodiment of the present invention.

[0036] Label Explanation

[0037] S1, S1', T1, T1', U1, U1': Training phase

[0038] S2, S2', T2, T2', U2, U2': Authentication Phase

[0039] S3, S3', T3, T3', U3, U3': Update Phase

[0040] S11, S13, S21, S23, S25, S27: Steps

[0041] S11', S12', S13', S21', S22', S23', S25', S27', S31, S32', S33', S34: Steps

[0042] T11, T13, T21, T23, T25: Steps

[0043] T11', T12', T13', T21', T22', T23', T25', T31, T32', T33': Steps

[0044] U11, U13, U21, U23, U24, U25, U26, U27: Steps

[0045] U11', U12', U13', U21', U22', U23', U24, U25', U26, U27': Steps

[0046] U31, U32', U33', U34: Steps Detailed Implementation

[0047] The following detailed description of the features and characteristics of the present invention in the embodiments is sufficient to enable anyone skilled in the art to understand the technical content of the present invention and implement it accordingly. Based on the disclosure of this specification, the scope of the patent applications, and the drawings, anyone skilled in the art can easily understand the related concepts and features of the present invention. The following embodiments further illustrate the viewpoints of the present invention in detail, but are not intended to limit the scope of the present invention in any way.

[0048] The present invention provides a method for authenticating user identity based on touch operation, comprising nine embodiments. The first, second, and third embodiments use heat maps established by touch operations as the authentication basis. The second and third embodiments, compared to the first embodiment, add an update mechanism for the authentication basis, but the execution order of the update mechanism differs between the second and third embodiments. The fourth, fifth, and sixth embodiments use neural network models established by touch operations as the authentication basis. The fifth and sixth embodiments, compared to the fourth embodiment, add an update mechanism for the authentication basis, but the execution order of the update mechanism differs between the fifth and sixth embodiments. The seventh, eighth, and ninth embodiments use both heat maps and neural networks as the authentication basis. The eighth and ninth embodiments, compared to the seventh embodiment, add an update mechanism for the authentication basis, but the execution order of the update mechanism differs between the eighth and ninth embodiments.

[0049] Figure 1The flowchart shown is a first embodiment of the method for authenticating user identity based on touch operation according to the present invention, wherein: touch operation is defined as any operation performed by a user in a specific area, and the type of touch operation may include at least one of moving, pressing (and subsequently releasing), and scrolling, but the present invention is not limited thereto. Figure 1 As shown, the method for authenticating user identity based on touch operation includes a training phase S1 and an authentication phase S2, and is applicable to computing devices with a touch interface and a processor. The computing device is, for example, a laptop or smartphone, and the touch interface is, for example, a physical touch panel or touch screen. In one embodiment, the touch interface may also consist of a physical device and a display interface, such as: a mouse and a screen displaying the mouse cursor, a motion sensor and a projector that projects the sensed motion onto a flat surface, a camera device and an application that converts the captured motion into a movement trajectory. However, the present invention is not limited to the above examples; any software or hardware device that allows a user to control the movement of a symbol (such as a cursor) within a specified range is applicable to the present invention.

[0050] Figure 2 This is shown as the training phase of the present invention (corresponding to...). Figure 1 S1) and certification phase (corresponding to Figure 1 S2) in the flowchart of the first embodiment, as shown below Figure 2 As shown, the training phase S1 includes steps S11 and S13, and the authentication phase S2 includes steps S21, S23, S25, and S27. The training phase S1 collects training data from multiple touch operations generated by a specific user through the touch interface and summarizes the behavioral patterns of that specific user. The authentication phase S2 collects test data from multiple touch operations generated by the current user through the touch interface and determines whether the behavioral patterns corresponding to these data are similar to the behavioral patterns generated in the training phase S1.

[0051] In the training phase S1, step S11 is that the touch interface generates multiple training touch parameters, step S13 is that the processor generates a training heatmap based on these training touch parameters, step S21 is that the touch interface generates multiple test touch parameters, step S23 is that the processor generates a test heatmap based on these test touch parameters, step S25 is that the processor compares the test heatmap and the training heatmap to generate an error map, and step S27 is that the processor generates one of an authentication pass signal and an authentication failure signal based on the error map.

[0052] In steps S11 and S21, each of the training and testing touch parameters includes: touch timing, touch location, and operation type. Through the touch interface driver or the operating system's software development kit (SDK), the processor can access the touch operation log file (Log) exemplified in Table 1 below.

[0053] Table 1: Touch operation log files

[0054]

[0055] In steps S13 and S23, each of the training heatmap and the test heatmap includes at least one of a position heatmap and a velocity heatmap. The position heatmap reflects the cumulative number of touch positions, and the processor generates multiple position heatmaps corresponding to different operation types, such as... Figure 3 , Figure 4 and Figure 5 As shown. Figure 3 The image shows location heatmaps of two different users based on mobile operations in one embodiment of the present invention. Figure 4 The image shows positional heatmaps of two different users of the present invention based on pressing operations in one embodiment. Figure 5 The image shows a positional heatmap of two different users of the present invention based on a scrolling operation in one embodiment.

[0056] Figure 6 Displayed as a position heatmap and velocity heatmap of a single user based on movement operations in one embodiment of the present invention, wherein the velocity heatmap reflects the direction and distance of the movement operation. In one embodiment, the processor first calculates multiple velocity vectors and then generates a velocity heatmap, wherein each velocity vector consists of two consecutive movement operations. Figure 6 In the velocity heatmap on the right, the horizontal axis represents the direction of the movement operation, in units such as degrees (radians); the vertical axis represents the distance of the movement operation, in units such as pixels. For example, if a movement operation takes 3 milliseconds to move from coordinates (0, 0) to coordinates (6, 12), then the velocity of this movement operation is 2√5 pixels per millisecond; and the X-axis component of this movement operation is 2 pixels, the Y-axis component is 4 pixels, and the angle of this movement operation is approximately 1.10714872 degrees (rad).

[0057] Step S25 primarily applies template matching techniques. Specifically, the processor compares the training heatmap and the test heatmap to generate an error map, where the training heatmap and the test heatmap used for comparison must be of the same type, for example, both are location heatmaps or both are velocity heatmaps. In a first example of step S25, the accuracy of the comparison is, for example, at least one of the pixel scale and the block scale. Pixel-scale comparison includes, for example, calculating the training heatmap and the test heatmap... Figure 2 The difference in grayscale values ​​between pixels at the same location; and the comparison of block scale, for example: calculating training heatmaps and test heatmaps. Figure 2 The structural similarity index measure (SSIM) is used to measure the structural similarity of the user. In the second example of step S25, the processor first divides the training heatmap and the test heatmap into multiple feature spaces, performs a rotation operation on each feature space, and then, using the method of the first example, randomly selects a feature space in the training heatmap and a feature space in the test heatmap for comparison. The second example can find the same touch pattern of the same user in different touch positions.

[0058] Based on Figure 2 In an extended embodiment, the processor also collects the currently executing processes on the computing device during the training phase S1 and the authentication phase S2, such as the application corresponding to the foreground window in the operating system (e.g., Microsoft Word, Google Chrome, etc.), and generates corresponding training heatmaps and test heatmaps based on the different processes. Therefore, when comparing the training heatmap and the test heatmap, in addition to the fact that their types must be the same, the corresponding processes must also be the same.

[0059] It is worth noting that, depending on the type of touch or the operating procedure, the processor may generate multiple training heatmaps in step S13; in step 23, the processor generates multiple test heatmaps corresponding to the multiple training heatmaps; therefore, in step S25, the processor generates multiple error maps, and the present invention does not limit the upper limit of the number of error maps.

[0060] In step S27, the processor calculates one or more error values ​​based on an error map, and then compares each error value with its corresponding threshold. If more than a specified number of error values ​​exceed the threshold, the processor generates an authentication failure signal; otherwise, if more than a specified number of error values ​​do not exceed the threshold, the processor generates an authentication success signal.

[0061] Figure 7The flowchart shown is a second embodiment of the method for authenticating user identity based on touch operation according to the present invention. Compared with the first embodiment, the second embodiment mainly adds an update stage S3'. Figure 8 The flowchart shown is a second embodiment of the training and authentication phases of the present invention. Figure 8 It can be seen that the training phase S1' of the second embodiment adds step S12' compared to the training phase S1 of the first embodiment, and the authentication phase S2' of the second embodiment adds step S22' compared to the authentication phase S2 of the first embodiment. The following only describes the newly added parts of the second embodiment; the steps that are the same in the second embodiment as in the first embodiment will not be repeated.

[0062] Step S12' involves the processor determining that the amount of training touch parameters collected exceeds a first threshold, and step S22' involves the processor determining that the amount of test touch parameters collected exceeds a test threshold. The amount of data collected can be at least one of a time interval (e.g., touch parameters collected within 72 hours) and a number of parameters (e.g., 100,000 touch parameters). After the collection of touch parameters begins in steps S11' and S21', the determination mechanisms in steps S12' and S22' must be satisfied before proceeding to steps S13' and S23', respectively.

[0063] Please refer to Figure 7 and Figure 8 In the second embodiment, the update phase S3' is located between the training phase S1' and the authentication phase S2'. Step S31' generates multiple new touch parameters for the touch interface, step S32' determines that the amount of these new touch parameters collected is greater than a second threshold, step S33' generates a new training heatmap based on these new touch parameters, and step S34' updates the training heatmap based on the new training heatmap. Basically, steps S31' to S33' in the update phase S3' are the same as steps S11' to S13' in the training phase S1'. In addition, in the second embodiment, the values ​​of the first threshold in step S12', the second threshold in step S32', and the test threshold in step S22' are not particularly limited.

[0064] In step S34', the processor calculates the new training heatmap and the difference between the training heatmaps. When the difference is less than the update threshold, the processor generates a new training heatmap (different from the training heatmap in step S1') based on the training parameters collected in step S11' and the new training parameters collected in step S31'. When the difference is not less than the update threshold, the processor replaces the original training heatmap generated in step S13' with the new training heatmap generated in step S33'.

[0065] Overall, since the user's touch pattern may change over time or with the current operating program, the second embodiment first collects a set of touch parameters to initialize a reference training heatmap, and then collects another set of touch parameters to update the original training heatmap.

[0066] Figure 9 The flowchart shown is a third embodiment of the method for authenticating user identity based on touch operation according to the present invention. The training phase S1' and authentication phase S2' of the third embodiment are basically the same as those of the second embodiment. In other words, the third embodiment is equivalent to adding steps S12' and S22' to the first embodiment, and adding an update phase S3' after the authentication phase S2'. In other words, the third embodiment is based on the second embodiment, but moves the execution order of the update phase S3' after the authentication phase S2'.

[0067] The third embodiment is applied as follows: A user generates their own training heatmap during the training phase S1'. The same user can then verify the accuracy of the training heatmap's judgment through the test heatmap during the authentication phase S2', and feed this accuracy back to the update phase S3' to correct the second threshold in step S32'. For example, if the accuracy in step S27' is less than a specific threshold during the authentication phase S2', the processor will increase the second threshold in step S32' to collect more new touch parameters in order to improve the accuracy when executing the authentication phase S2' next time.

[0068] In the second and third embodiments, the update phase S3' can also be performed periodically to update the touch parameters to reflect changes in the user's touch mode over time, wherein the update method is as follows:

[0069] TP = TP new ×w1+TP current ×w2

[0070] Where TP represents the updated touch parameters. new For the new touch parameters described in step S31', TP current The touch parameters described in step S11' are w1 and w2, which are weights.

[0071] Figure 10 The flowchart shown is a fourth embodiment of the method for authenticating user identity based on touch operation according to the present invention. Figure 11The flowchart shown is a training phase and an authentication phase of the present invention in the fourth embodiment. In step T11, the touch interface generates multiple training touch parameters. In step T13, the processor trains a neural network model based on these training touch parameters. In step T21, the touch interface generates multiple test touch parameters. In step T23, the processor inputs these test touch parameters into the neural network model to generate a predicted value. In step T25, the processor calculates the error between the predicted value and the actual value to generate one of an authentication pass signal and an authentication failure signal.

[0072] The main difference between the fourth embodiment and the first embodiment is step T13 in the training phase T1 and steps T23 and T25 in the authentication phase T2. Therefore, only the details of these different steps T13, T23 and T25 will be described below. As for the steps that are the same in the fourth embodiment and the first embodiment, they will not be described again.

[0073] In step T13, the processor converts multiple training touch parameters into time series, then uses these time series as input to train a neural network model, resulting in a prediction model for predicting subsequent time series. The neural network model can be, for example, a Long Short-Term Memory (LSTM) model. Furthermore, multiple prediction models can be trained according to touch parameters for different touch types.

[0074] Time series can be generated in ways including, but not limited to, the following three methods:

[0075] 1. The time series consists of the time and location of each move operation.

[0076] 2. The time series consists of a start time (or end time) and a velocity vector, where the velocity vector is calculated by the processor from two values ​​taken from multiple consecutive movement operations at fixed time intervals. Because the velocity vector is a two-dimensional vector, this type of time series is a multivariate statistical time series.

[0077] 3. The time series consists of one or more centroids for each training heatmap and the corresponding time point for each training heatmap. For example, multiple training touch parameters are divided into multiple groups according to time sequence, and a training heatmap is generated for each group of training touch parameters. Then, the centroid of each training heatmap is calculated, for example, using the K-means algorithm.

[0078] In step T23, the processor may input a set of test touch signals to the prediction model generated in step T13 to obtain a set of prediction values, which may include one or more test touch signals.

[0079] The first example of step T25 adopts the time series anomaly detection mechanism described in "H Nguyen, Kim Phuc Tran, S Thomassey, MHamad. Forecasting and Anomaly Detection approaches using LSTM and LSTMAutoencoder techniques with the applications in Supply Chain Management. International Journal of Information Management, Elsevier, 2020."

[0080] The anomaly detection mechanism is explained as follows: An LSTM model can predict the next touch parameters, such as position. If the actual position differs too much from the predicted position, the actual position is considered an anomaly. An autoencoder, such as an LSTM encoder or an image autoencoder, learns a norm as an authentication standard. The autoencoder acts as a feature extractor, regardless of whether the input is a general image, heatmap, time series, or multivariate statistical time series / heatmap. Using a position heatmap as an example, the norm represents the frequently touched areas of the touchpad and is represented in a low-dimensional or embedded space. In authentication phase T2, a test heatmap is input to the autoencoder and reconstructed. If the difference between the test heatmap and the reconstructed test heatmap is too large, the test heatmap is considered an anomaly.

[0081] A second example of step T25 is as follows: The processor obtains another set of test touch signals that are sequentially received after the set of test touch signals described in step T23, sets them as a set of actual values, and then calculates the error value between each of the actual values ​​and its corresponding predicted value. The processor compares each error value with its corresponding threshold. If more than a specified number of error values ​​are greater than the threshold, the processor generates an authentication failure signal; conversely, if more than a specified number of error values ​​are not greater than the threshold, the processor generates an authentication success signal.

[0082] Figure 12 The flowchart shown is a fifth embodiment of the method for authenticating user identity based on touch operation according to the present invention. Compared with the fourth embodiment, the fifth embodiment mainly adds an update stage T3'. Figure 13 The flowchart shown illustrates the training and certification phases of the present invention in the fifth embodiment. Figure 13It can be seen that the training phase T1' of the fifth embodiment adds step T12' compared to the training phase T1 of the fourth embodiment, and the authentication phase T2' of the fifth embodiment adds step T22' compared to the authentication phase T2 of the fourth embodiment. The steps identical to those in the fifth and fourth embodiments will not be repeated below. Furthermore, the update phase T3' of the fifth embodiment has a similar architecture to the update phase S3' of the second embodiment. The difference between the two is that step T33' involves the processor updating the neural network model based on the new touch parameters, while step S33' involves the processor updating the training heatmap based on the new touch parameters. In summary, these two embodiments update different authentication reference standards based on the new touch parameters.

[0083] Figure 14 The flowchart shown is a sixth embodiment of the method for authenticating user identity based on touch operation according to the present invention. The training phase T1' and authentication phase T2' of the sixth embodiment are basically the same as those of the fifth embodiment. In other words, the sixth embodiment is based on the fifth embodiment, but the execution order of the update phase T3' is moved after the authentication phase T2'.

[0084] Figure 15 The flowchart shown is a seventh embodiment of the method for authenticating user identity based on touch operation according to the present invention, including a training phase U1 and an authentication phase U2. Figure 16 The flowchart shown is a diagram of the training and authentication phases of the present invention in the seventh embodiment. The training phase U1 includes steps U11 and U13, and the authentication phase U2 includes steps U21, U23, U24, U25, U26, and U27.

[0085] Step U11 generates multiple training touch parameters for the touch interface; step U13 generates a training heatmap and a training neural network model based on these training touch parameters; step U21 generates multiple test touch parameters for the touch interface; step U23 generates a test heatmap based on these test touch parameters; step U24 compares the test heatmap and the training heatmap to generate an error map; step U25 calculates a first error value based on the error map; step U26 inputs these test touch parameters into the neural network model to generate a predicted value; and step U27 generates one of an authentication pass signal and an authentication failure signal based on the first error value and a second error value. As can be seen from the above, the seventh embodiment integrates the first and fourth embodiments, and uses both a training heatmap and a neural network model for user authentication.

[0086] Figure 17 The flowchart shown is a method for authenticating user identity based on touch operation according to the present invention in the eighth embodiment. Compared with the seventh embodiment, the eighth embodiment mainly adds an update stage U3'. Figure 18A and Figure 18BThe flowcharts shown in the eighth embodiment of the present invention are the training phase and the authentication phase, respectively. Figure 18A and Figure 18B It can be seen that the training phase U1' of the eighth embodiment adds step U12' compared to the training phase U1 of the seventh embodiment, and the authentication phase U2' of the eighth embodiment adds step U22' compared to the authentication phase U2 of the seventh embodiment. The same steps in the eighth embodiment and the seventh embodiment will not be repeated below. Furthermore, the eighth embodiment is equivalent to integrating the second and fifth embodiments; in other words, steps U23' and U24' are equivalent to steps S23' and S25', and step U26' is equivalent to T26'. Based on the integration requirements, the eighth embodiment adds steps U25' and U27'. In step U25', the processor calculates a first error value based on the error map. In step U27', the processor generates one of an authentication pass signal and an authentication failure signal based on the first error value and a second error value, wherein the second error value is associated with the predicted value obtained in step S26' and the actual value obtained from step U23'. Details regarding the predicted value, the actual value, and the second error value can be found in the error value calculation section of step T27. The processor, for example, calculates a weighted average of the first error value and the second error value, and determines whether this weighted average is greater than a threshold, thereby deciding whether to output an authentication pass signal or an authentication failure signal.

[0087] Furthermore, the update phase U3' of the eighth embodiment is equivalent to integrating S3' of the second embodiment and T3' of the fifth embodiment. In other words, in step U33', the processor not only generates a new training heatmap based on the new touch parameters, but also updates the neural network model based on the new touch parameters. In step U34', the processor updates the training heatmap based on the new training heatmap. In summary, the processor simultaneously employs two different certification reference standards.

[0088] Figure 19 The flowchart shown is from the ninth embodiment of the method for authenticating user identity based on touch operation, where the training phase U1' and authentication phase U2' are substantially the same as in the eighth embodiment. In other words, the ninth embodiment is based on the eighth embodiment, but the execution order of the update phase U3' is moved after the authentication phase U2'.

[0089] In the seventh, eighth and ninth embodiments, the present invention not only uses template matching technology to compare training heatmaps and test heatmaps, but also incorporates artificial intelligence (AI) technology to provide additional comparison standards for the user's unique touch recognition.

[0090] In summary, the method for authenticating user identity based on touch operation proposed in this invention has the following advantages:

[0091] 1. Continuously and non-intrusively monitor and authenticate current users;

[0092] 2. It adopts a data-driven approach and is adaptive, able to adapt to the ever-changing touch operation patterns of users; and

[0093] 3. Increase the frequency of user authentication, and perform authentication when existing user authentication mechanisms (such as passwords or biometrics) are not in use.

[0094] It is worth noting that the method proposed in this invention does not replace existing user authentication mechanisms, but rather supplements and enhances the security of existing mechanisms to prevent internal attacks. In other words, it can detect impersonators even when the computing device is unlocked.

[0095] While the present invention has been disclosed above with reference to the foregoing embodiments, it is not intended to limit the invention. Any modifications and refinements made without departing from the spirit and scope of the invention are within the scope of patent protection of the present invention. For the scope of protection defined by the present invention, please refer to the scope defined in the claims.

Claims

1. A method for authenticating user identity based on touch operation, characterized in that, include: The training phase includes: The touch interface generates multiple training touch parameters; and The processor generates a training heatmap and a training neural network model based on the training touch parameters; and The certification process includes: The touch interface generates multiple test touch parameters; The processor generates a test thermal map based on the test touch parameters; The processor compares the test heatmap and the training heatmap to generate an error map; The processor calculates a first error value based on the error map; The processor inputs the test touch parameters into the neural network model to generate a predicted value; and The processor generates one of an authentication pass signal and an authentication failure signal based on the first error value and a second error value, wherein the second error value is associated with the predicted value and an actual value; Both the training heatmap and the test heatmap include at least one of a position heatmap and a speed heatmap. The position heatmap is used to reflect the cumulative number of touch positions, and the speed heatmap is used to reflect the direction and distance of the movement operation.

2. The method for authenticating user identity based on touch operation according to claim 1, characterized in that, The training phase further includes: before the processor generates the training heatmap and trains the neural network model, the processor determines that the amount of training touch parameters collected is greater than a first threshold. The authentication phase further includes: before the processor generates the test heatmap and trains the neural network model, the processor determines that the amount of collected test touch parameters is greater than a test threshold; and The method further includes an update phase, which includes: The touch interface generates multiple new touch parameters; The processor determines that the amount of the new touch parameter collected is greater than a second threshold; The processor generates a new training heatmap and updates the neural network model based on the new touch parameters; and The processor updates the training heatmap based on the new training heatmap.

Images