A cross-browser device fingerprint feature generation method based on artificial intelligence technology

Abstract

The present invention discloses a method for generating cross-browser device fingerprint features based on artificial intelligence technology, the method comprising the following steps: preprocessing the collected fingerprint data; deriving features and expanding feature fields of fingerprint features; performing dimension reduction visualization and correlation matrix visualization operations on fingerprint data; calculating the feature discrimination and feature stability of candidate fingerprint features; screening the candidate fingerprint features; generating fingerprint features by splicing and encrypting the screened fingerprint features or by comparing the differences of fingerprint features. The present invention uses artificial intelligence to derive device collection features, thereby increasing the dimension of features, being more conducive to selecting better feature indicators, and improving the accuracy of device fingerprint recognition across browsers; at the same time, the present invention takes into account both feature stability and discrimination in feature selection.

Description

Technical Field

[0001] The present invention relates to the field of secure privacy computing technology, and in particular to a method for generating cross-browser device fingerprint features based on artificial intelligence technology. Background Art

[0002] Device fingerprint technology uses the characteristic attributes of a device to generate a unique identifier for each device. Traditional device fingerprint technology mostly selects features based on manual experience. Due to the size of the test sample, this selection method is highly accidental. More importantly, current devices will generate different device fingerprints on different browsers, and it is difficult to guarantee the accuracy of fingerprint technology in cross-browser fingerprint recognition.

[0003] In recent years, strategies for establishing models through machine learning methods for device uniqueness authentication have continued to emerge. However, most methods focus on model construction and rarely pay attention to feature selection. However, feature selection directly affects the performance of the model. Therefore, the use of artificial intelligence to select features is becoming increasingly important in the device fingerprint generation process.

[0004] However, existing feature engineering relies on the size of a given sample data volume, and the amount of data directly affects the stability of the screening features. In addition, browsers on the market are updated rapidly, and it is difficult to satisfy device fingerprint recognition on unknown browsers. The current device fingerprint generation technology has high discrimination but poor stability, that is, different fingerprint identifiers will be generated between different devices, but the same fingerprint identifier cannot be generated on different browsers on the same device.

[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0006] In view of the problems in the related art, the present invention proposes a cross-browser device fingerprint feature generation method based on artificial intelligence technology to overcome the above-mentioned technical problems existing in the existing related technology.

[0007] To this end, the specific technical solution adopted by the present invention is as follows:

[0008] A method for generating cross-browser device fingerprint features based on artificial intelligence technology, the method comprising the following steps:

[0009] Step S1, preprocessing the collected fingerprint data;

[0010] Step S2, deriving fingerprint features and expanding feature fields based on the preprocessed fingerprint data;

[0011] Step S3, performing dimension reduction visualization and correlation matrix visualization operations on the fingerprint data and generating candidate fingerprint features;

[0012] Step S4, using an adaptive method to calculate the feature discrimination and feature stability of the candidate fingerprint features respectively;

[0013] Step S5: Screening candidate fingerprint features according to business requirements and scenarios and in combination with feature discrimination and feature stability;

[0014] Step S6: Generate fingerprint features by combining and encrypting the selected fingerprint features or by comparing the differences of the fingerprint features.

[0015] Furthermore, the data preprocessing of the collected fingerprint data in step S1 includes performing missing value processing, outlier processing and data normalization processing on the fingerprint data.

[0016] Furthermore, the missing value processing adopts the method of zero filling, mean filling or maximum and minimum filling, the outlier processing adopts the method of three times variance rule or box plot analysis, and the data normalization processing adopts the method of maximum / minimum normalization or z-score normalization.

[0017] Furthermore, the feature derivation includes one-hot encoding, brute force crossover and synthetic features of the fingerprint features.

[0018] Furthermore, the dimension reduction and visualization operation of the fingerprint data in step S3 includes using a principal component analysis method to perform data distribution analysis, comparative analysis and statistical analysis on the fingerprint features after dimension reduction.

[0019] Furthermore, in step S3, the correlation matrix visualization operation on the fingerprint data includes using the Pearson correlation coefficient to calculate the correlation coefficient between different fingerprint features, and removing the features with high correlation and retaining one.

[0020] Furthermore, using traditional statistical methods and based on the idea of ​​controlling variables, we calculated the frequency distribution of different devices under the alternative fingerprint features under the same browser, and the feature discrimination calculation formula is as follows:

[0021]

[0022] Among them, u i represents the feature discrimination of the i-th feature, f k，i It represents the distribution of the values ​​of the i-th feature on different devices on the k-th browser. The unique function is a deduplication function that returns the number of different values ​​of the feature. n k It indicates how many devices there are on the k-th browser, i indicates the i-th feature, k indicates the k-th browser, K indicates the total number of browser types, and m indicates the total number of fingerprint features.

[0023] Furthermore, the feature stability of the candidate fingerprint features is calculated using traditional statistics and mask generation methods. The calculation formula is as follows:

[0024]

[0025] Among them, p i Indicates that the i-th feature calculates the fingerprint feature stability index based on the variance, f j，i It represents the distribution of the values ​​of the jth device on the i-th feature under all browser versions. The D function is a function for calculating the variance. j represents the jth device, J represents the total number of devices, i represents the i-th feature, and m represents the total number of fingerprint features.

[0026] Furthermore, the steps of generating the mask are as follows:

[0027] For any two different browsers, compare whether the feature values ​​of each device are the same in these two browsers. If they are the same, record them as 1, otherwise record them as 0;

[0028] For each device, the above steps will generate a 01 string with the length of the feature value, and calculate the number of 0s and 1s corresponding to each fingerprint feature;

[0029] If the number of 1s corresponding to each fingerprint feature is greater than the number of 0s, the mask corresponding to the fingerprint feature is 1, otherwise it is 0. At this time, the mask is generated.

[0030] Furthermore, after the device fingerprint is generated by the splicing and combination and encryption method in step S6, some test data is selected to calculate the proportion of different device fingerprints generated by different devices under the same browser version and the proportion of the same device fingerprint generated by the same device under two different browsers. Based on these two proportions, it is decided whether to further re-derive the fingerprint features.

[0031] The beneficial effects of the present invention are:

[0032] (1) The present invention adopts artificial intelligence to derive device collection features, thereby increasing the dimension of features, which is more conducive to selecting better feature indicators and improving the accuracy of device fingerprint recognition across browsers; at the same time, the present invention takes into account both feature stability and distinguishability in feature selection, and adopts a mask method in the process of calculating feature stability to help screen feature indicators with higher stability.

[0033] (2) The present invention combines artificial intelligence to screen fingerprint features, which minimizes human intervention compared to traditional fingerprint feature generation methods, and expands the feature field set through feature derivation, thereby better selecting fingerprint features.

[0034] (3) The present invention combines the statistical and artificial intelligence methods to construct fingerprint features by integrating the two indicators of feature stability and discrimination, thereby solving the problem of inconsistent generation of fingerprints across browsers. Moreover, by adopting the idea of ​​big data to calculate discrimination and stability, the accuracy of device fingerprints between different devices is improved, and the differences in device fingerprints on different browsers of the same device are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 It is a flowchart of a method for generating cross-browser device fingerprint features based on artificial intelligence technology according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0038] According to an embodiment of the present invention, a method for generating cross-browser device fingerprint features based on artificial intelligence technology is provided.

[0039] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the method for generating cross-browser device fingerprint features based on artificial intelligence technology in an embodiment of the present invention, the method comprises the following steps:

[0040] Step S1, preprocessing the collected fingerprint data;

[0041] In one embodiment, the data preprocessing of the collected fingerprint data in step S1 includes performing missing value processing, outlier processing and data normalization processing on the fingerprint data.

[0042] In one embodiment, the missing value processing adopts zero value filling, mean filling or maximum and minimum filling methods, the outlier processing adopts the 3-fold variance rule or box plot analysis method, and the data normalization processing adopts maximum / minimum normalization or z-score normalization method.

[0043] Step S2, deriving fingerprint features and expanding feature fields based on the preprocessed fingerprint data;

[0044] In one embodiment, the feature derivation includes one-hot encoding of fingerprint features, brute force crossover (combining multiple features through mathematical operations) and synthetic features (a process of mapping from a linear model to a nonlinear model by combining individual features).

[0045] Step S3, performing dimension reduction visualization and correlation matrix visualization operations on the fingerprint data and generating candidate fingerprint features;

[0046] In one embodiment, the dimension reduction and visualization operation of the fingerprint data in step S3 includes using a principal component analysis (PCA) method to perform data distribution analysis, comparative analysis and statistical analysis on the fingerprint features after dimension reduction.

[0047] In one embodiment, the correlation matrix visualization operation performed on the fingerprint data in step S3 includes using the Pearson correlation coefficient to calculate the correlation coefficient between different fingerprint features, and removing features with high correlation to keep one.

[0048] Specifically, by conducting data exploration on the processed data, analysts can more intuitively observe the distribution of different fingerprint features. At the same time, they can combine expert experience to further comprehensively evaluate the rationality and effectiveness of the fingerprint features derived in the previous step.

[0049] Step S4, using an adaptive method to calculate the feature discrimination and feature stability of the candidate fingerprint features respectively;

[0050] In one embodiment, a traditional statistical method is used, and based on the idea of ​​controlling variables, the frequency distribution of different devices under the candidate fingerprint features is calculated under the same browser, and the feature discrimination calculation formula is as follows:

[0051]

[0052] Among them, u i represents the feature discrimination of the i-th feature, f k，i It represents the distribution of the values ​​of the i-th feature on different devices on the k-th browser. The unique function is a deduplication function that returns the number of different values ​​of the feature. n k It indicates how many devices there are on the k-th browser, i indicates the i-th feature, k indicates the k-th browser, K indicates the total number of browser types, and m indicates the total number of fingerprint features.

[0053] In one embodiment, the feature stability of the candidate fingerprint features is calculated using traditional statistics and mask generation methods, and the calculation formula is as follows:

[0054]

[0055] Among them, p i Indicates that the i-th feature calculates the fingerprint feature stability index based on the variance, f j，i It represents the distribution of the values ​​of the jth device on the i-th feature under all browser versions. The D function is a function for calculating the variance. j represents the jth device, J represents the total number of devices, i represents the i-th feature, and m represents the total number of fingerprint features.

[0056] In one embodiment, the steps of generating the mask are as follows:

[0057] For any two different browsers, compare whether the feature values ​​of each device are the same in these two browsers. If they are the same, record them as 1, otherwise record them as 0;

[0058] For each device, the above steps will generate a 01 string with the length of the feature value, and calculate the number of 0s and 1s corresponding to each fingerprint feature;

[0059] If the number of 1s corresponding to each fingerprint feature is greater than the number of 0s, the mask corresponding to the fingerprint feature is 1, otherwise it is 0. At this time, the mask is generated.

[0060] Specifically, traditional statistics uses some stability indicators such as PSI, variance, etc. to output feature stability calculation results as one of the feature screening parameters; the mask is a set of 01 strings (1 represents retaining this feature, 0 represents discarding this feature) generated by brute force search of fingerprint features. Combining the mask can generate stability features more accurately;

[0061] In addition, the calculation of feature fingerprint stability combines the mask and the calculation of fingerprint feature stability index based on variance. First, the features with a mask of 1 are screened out. Secondly, the feature data values ​​are mapped to [0, 1] during the data normalization process, so p can be screened out from the features with a mask of 0. i Smaller fingerprint features.

[0062] Step S5: Screen candidate fingerprint features according to business needs and scenarios and in combination with feature discrimination and feature stability (the weights of the two can be adjusted according to expert experience. If the business needs are biased towards discrimination, the discrimination index can be screened more, otherwise the stability index can be screened more);

[0063] Step S6: Generate fingerprint features by encrypting the selected fingerprint features after splicing or by comparing the differences of fingerprint features (for example, when calculating the stability index based on the variance, the feature values ​​are different but very close).

[0064] In one embodiment, after the device fingerprint is generated by the splicing and combination and encryption method in step S6, some test data is selected to calculate the proportion of different device fingerprints generated by different devices under the same browser version and the proportion of the same device fingerprint generated by the same device under two different browsers. Based on these two proportions, it is decided whether to further re-derive the fingerprint features. Specifically, if the two proportions differ greatly, the fingerprint features are re-derived.

[0065] To sum up, with the help of the above-mentioned technical scheme of the present invention, the present invention adopts artificial intelligence to derive device collection features, thereby increasing the dimension of the features, which is more conducive to selecting better feature indicators and improving the accuracy of device fingerprint recognition across browsers; at the same time, the present invention takes into account the two dimensions of feature stability and distinctiveness in feature selection, and uses a mask method in the process of calculating feature stability to help screen feature indicators with higher stability.

[0066] In addition, the present invention combines artificial intelligence to screen fingerprint features, which minimizes human intervention compared to traditional fingerprint feature generation methods, and expands the feature field set through feature derivation, thereby better selecting fingerprint features.

[0067] In addition, the present invention combines the methods of statistics and artificial intelligence to integrate the two indicators of feature stability and discrimination to construct fingerprint features, thereby solving the problem of inconsistent generation of fingerprints across browsers; and by adopting the idea of ​​big data to calculate discrimination and stability, the accuracy of device fingerprints between different devices is improved, and the differences in device fingerprints on different browsers of the same device are reduced.

[0068] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for generating cross-browser device fingerprint features based on artificial intelligence technology, characterized in that: The method comprises the following steps: Step S1, preprocessing the collected fingerprint data; Step S2, deriving fingerprint features and expanding feature fields based on the preprocessed fingerprint data; Step S3, performing dimension reduction visualization and correlation matrix visualization operations on the fingerprint data and generating candidate fingerprint features; Step S4, using an adaptive method to calculate the feature discrimination and feature stability of the candidate fingerprint features respectively; Step S5: Screening candidate fingerprint features according to business requirements and scenarios and in combination with feature discrimination and feature stability; Step S6, generating fingerprint features by combining and encrypting the selected fingerprint features or by comparing the differences of the fingerprint features; The dimension reduction and visualization operation of the fingerprint data in step S3 includes using the principal component analysis method to perform data distribution analysis, comparative analysis and statistical analysis on the fingerprint features after dimension reduction; In step S3, the correlation matrix visualization operation of the fingerprint data includes using the Pearson correlation coefficient to calculate the correlation coefficient between different fingerprint features, and removing the features with high correlation and retaining one; Using traditional statistical methods and based on the idea of ​​controlling variables, we calculate the frequency distribution of different devices under the alternative fingerprint features under the same browser, and the feature discrimination calculation formula is as follows: Among them, u i represents the feature discrimination of the i-th feature, f k，i It represents the distribution of the values ​​of the i-th feature on different devices on the k-th browser. The unique function is a deduplication function that returns the number of different values ​​of the feature. n k represents the total number of devices on the k-th browser, i represents the i-th feature, k represents the k-th browser, K represents the total number of browser types, and m represents the total number of fingerprint features; The characteristic stability of the candidate fingerprint features is calculated using traditional statistics and mask generation methods. The calculation formula is as follows: Among them, p i Indicates that the i-th feature calculates the fingerprint feature stability index based on the variance, f j，i It represents the distribution of the values ​​of the jth device on the i-th feature under all browser versions. The D function is a function for calculating the variance. j represents the jth device, J represents the total number of devices, i represents the i-th feature, and m represents the total number of fingerprint features. The steps of mask generation are as follows: For any two different browsers, compare whether the feature values ​​of each device are the same in these two browsers. If they are the same, record them as 1, otherwise record them as 0; For each device, the above steps will generate a 01 string with the length of the feature value, and calculate the number of 0s and 1s corresponding to each fingerprint feature; If the number of 1s corresponding to each fingerprint feature is greater than the number of 0s, the mask corresponding to the fingerprint feature is 1, otherwise it is 0, and the mask is generated; The calculation of feature fingerprint stability combines masking and calculating fingerprint feature stability index based on variance. First, the features with mask 1 are filtered out, and then the features with mask 0 are filtered out. i Smaller fingerprint features.

2. According to the method for generating cross-browser device fingerprint features based on artificial intelligence technology according to claim 1, it is characterized in that: The data preprocessing of the collected fingerprint data in step S1 includes performing missing value processing, outlier processing and data normalization processing on the fingerprint data.

3. According to the method for generating cross-browser device fingerprint features based on artificial intelligence technology according to claim 2, it is characterized in that: The missing value processing adopts the method of zero filling, mean filling or maximum and minimum filling, the outlier processing adopts the method of three times variance rule or box plot analysis, and the data normalization processing adopts the method of maximum / minimum normalization or z-score normalization.

4. According to the method for generating cross-browser device fingerprint features based on artificial intelligence technology according to claim 1, it is characterized in that: The feature derivation includes one-hot encoding, brute force crossover and synthetic features of fingerprint features.

5. The method for generating cross-browser device fingerprint features based on artificial intelligence technology according to claim 1, characterized in that: After the device fingerprint is generated by the splicing and combination and encryption method in step S6, some test data is selected to calculate the proportion of different device fingerprints generated by different devices under the same browser version and the proportion of the same device fingerprint generated by the same device under two different browsers. Based on these two proportions, it is decided whether to further re-derive the fingerprint features.

Images