Short message author identity authentication method and system in cloud computing scenario

Abstract

This invention discloses a method and system for verifying the identity of short message authors in a cloud computing scenario, belonging to the field of quantitative text application technology in cloud computing scenarios. The technical problem it addresses is how to verify the identity of short message authors in a cloud computing scenario. The method includes the following steps: For a long text from a target author, segments are created, and feature parameters for each segment are extracted using an n-gram feature extraction method. Based on the feature parameters of all segments, a feature standard for the target author is calculated. The supervised learning feature standard is used as a target threshold for the target author. For a long text from an author to be verified, the long text is segmented, and feature parameters for each segment are extracted using an n-gram feature extraction method. Based on the feature parameters of all segments, a feature standard for the target author is calculated. The feature standard for the author to be verified is compared with the target threshold for the target author, and the comparison result determines whether the author to be verified is the target author.

Description

Technical Field

[0001] This invention relates to the field of quantitative text application technology in cloud computing scenarios, specifically to a method and system for verifying the identity of short message authors in cloud computing scenarios. Background Technology

[0002] While existing methods can achieve high accuracy in identifying long texts, they still present certain challenges in identifying authors of short texts, especially when dealing with scenarios involving a large number of authors.

[0003] Cloud computing is a type of distributed computing that breaks down massive data processing programs into countless smaller programs via a network "cloud." These smaller programs are then processed and analyzed by a system composed of multiple servers to obtain results and return them to the user. Cloud computing services are divided into three layers: IaaS, PaaS, and SaaS, and have wide applications in daily life and production. These include online messaging services such as email, SMS, and microblogging provided by telecom operators and internet companies.

[0004] How to verify the identity of the author of short messages in a cloud computing scenario is a technical problem that needs to be solved. Summary of the Invention

[0005] The technical objective of this invention is to address the above-mentioned shortcomings by providing a method and system for verifying the identity of short message authors in a cloud computing scenario, thereby solving the technical problem of how to verify the identity of short message authors in a cloud computing scenario.

[0006] In a first aspect, the present invention provides a method for verifying the identity of short message authors in a cloud computing scenario, comprising the following steps:

[0007] Obtain at least one short message text from the target author, preprocess the at least one short message text, and then summarize it into a long text;

[0008] For a long text by a target author, the long text is sliced ​​and the feature parameters of each slice are extracted using the n-gram feature extraction method. The feature criteria of the target author are calculated based on the feature parameters of all slices.

[0009] The error rate of the slices is calculated based on the slices that fail to identify the correct author and the slices that identify the wrong author. The feature standard is then subjected to supervised learning based on the error rate, and the supervised learning feature standard is used as the target threshold for the target author.

[0010] Obtain at least one short message text from an author to be verified, and preprocess the at least one short message text and then summarize it into a long text.

[0011] For a long text by an author to be verified, the long text is sliced ​​and the feature parameters of each slice are extracted using the n-gram feature extraction method. The feature criteria of the author to be verified are calculated based on the feature parameters of all slices.

[0012] The feature criteria of the author to be verified are compared with the target threshold of the target author, and the comparison result is used to determine whether the author to be verified is the target author.

[0013] Preferably, the short message texts of the target author and the author to be verified are preprocessed using the following steps:

[0014] Delete all duplicate text messages;

[0015] The special format short message text is parsed separately and the message body is extracted. The special format short message text includes emails.

[0016] The text message is normalized by using filters. The normalization process includes unifying numbers, special symbols, encoding methods, and capitalization, as well as filtering spaces and punctuation.

[0017] Preferably, for long texts by the target author, the feature criteria of the target author are calculated through the following steps:

[0018] The long text is divided into two parts, namely the first part. Part Two ;

[0019] Part Two Dividing a text into p blocks into equal slices is represented as follows: , … ;

[0020] The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ;

[0021] Features are extracted from each slice using the n-gram feature extraction method, resulting in a set of feature parameters for each slice. slice The corresponding set of feature parameters ;

[0022] use express and The same feature parameters between The proportion of the total is denoted as ;

[0023] By traversal , … ,get The distribution of the numbers is analyzed, and their mean is calculated. With variance ;

[0024] Based on the above and Obtain an initial parameter As a characteristic criterion for the target author The calculation formula is = -( ).

[0025] Preferably, for long texts with authors to be verified, the feature criteria of the authors to be verified are calculated through the following steps:

[0026] The long text is divided into two parts, namely the first part. Part Two ;

[0027] Part Two Dividing a text into p blocks into equal slices is represented as follows: , … ;

[0028] The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ;

[0029] Features are extracted from each slice using the n-gram feature extraction method, resulting in a set of feature parameters for each slice. slice The corresponding set of feature parameters ;

[0030] use express and The same feature parameters between The proportion of the total is denoted as .

[0031] Preferably, the characteristic criteria of the author to be verified are compared with the target threshold of the target author, and the comparison result is used to determine whether the author to be verified is the target author, including the following steps:

[0032] when Greater than the results obtained from supervised learning When, it is considered satisfy Feature criteria, traversal , … This determines whether the author to be verified and the target author are the same author.

[0033] Secondly, the present invention provides a short message author authentication system in a cloud computing scenario, used to authenticate short message authors through the short message author authentication method in a cloud computing scenario as described in any one of the first aspects, the system comprising:

[0034] The text processing module is used to acquire at least one short message text from a target author, preprocess the at least one short message text and summarize it into a long text; and to acquire at least one short message text from an author to be verified, preprocess the at least one short message text and summarize it into a long text.

[0035] The first feature extraction module is applied to the long text of the target author. It is used to slice the long text and extract the feature parameters of each slice using the n-gram feature extraction method. Based on the feature parameters of all slices, the feature standard of the target author is calculated.

[0036] The feature-supervised learning module is used to calculate the error rate of a slice based on slices that fail to identify the correct author and slices that identify the wrong author, perform supervised learning on the feature standard based on the error rate, and use the supervised learning feature standard as the target threshold for the target author.

[0037] The second feature extraction module is applied to the long text of the author to be verified. It is used to slice the long text and extract the feature parameters of each slice using the n-gram feature extraction method. Based on the feature parameters of all slices, the feature standard of the author to be verified is calculated.

[0038] The comparison analysis module is used to compare the feature criteria of the author to be verified with the target threshold of the target author, and determine whether the author to be verified is the target author based on the comparison results.

[0039] Preferably, for both the target author's and the author's short message texts, the text processing module preprocesses the short message texts through the following steps:

[0040] Delete all duplicate text messages;

[0041] The special format short message text is parsed separately and the message body is extracted. The special format short message text includes emails.

[0042] The text message is normalized by using filters. The normalization process includes unifying numbers, special symbols, encoding methods, and capitalization, as well as filtering spaces and punctuation.

[0043] Preferably, for long texts by the target author, the first feature extraction module is used to calculate the feature criteria of the target author through the following steps:

[0044] The long text is divided into two parts, namely the first part. Part Two ;

[0045] Part Two Dividing a text into p blocks into equal slices is represented as follows: , … ;

[0046] The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ;

[0047] Features are extracted from each slice using the n-gram feature extraction method, resulting in a set of feature parameters for each slice. slice The corresponding set of feature parameters ;

[0048] use express and The same feature parameters between The proportion of the total is denoted as ;

[0049] By traversal , … ,get The distribution of the numbers is analyzed, and their mean is calculated. With variance ;

[0050] Based on the above and Obtain an initial parameter As a characteristic criterion for the target author The calculation formula is = -( ).

[0051] Preferably, for long texts by authors to be verified, the second feature extraction module calculates the feature criteria of the authors to be verified through the following steps:

[0052] The long text is divided into two parts, namely the first part. Part Two ;

[0053] Part Two Dividing a text into p blocks into equal slices is represented as follows: , … ;

[0054] The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ;

[0055] Features are extracted from each slice using the n-gram feature extraction method, resulting in a set of feature parameters for each slice. slice The corresponding set of feature parameters ;

[0056] use express and The same feature parameters between The proportion of the total is denoted as .

[0057] Preferably, the comparison analysis module is used to perform the following:

[0058] when Greater than the results obtained from supervised learning When, it is considered satisfy Feature criteria, traversal , … This determines whether the author to be verified and the target author are the same author.

[0059] The short message author authentication method and system in the cloud computing scenario of the present invention have the following advantages:

[0060] 1. The long texts of the target author and the author to be verified are sliced ​​and n-gram features are extracted separately. The feature standard of the target author is formed based on the n-gram features of the target author and optimized through supervised learning. The feature standard of the author to be verified is formed based on the n-gram features of the author to be verified. The feature standard of the author to be verified is compared with the feature standard of the target author after supervised learning. The comparison result determines whether the author to be verified is the target author. This realizes the identity verification of the author of short text messages and improves the efficiency and accuracy of verification.

[0061] 2. Applicable to scenarios with a large number of authors. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] The invention will be further described below with reference to the accompanying drawings.

[0064] Figure 1 This is a flowchart of a short message author authentication method in a cloud computing scenario, as shown in Example 1. Detailed Implementation

[0065] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments are not intended to limit the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0066] This invention provides a method and system for authenticating short message authors in a cloud computing scenario, which addresses the technical problem of how to authenticate short message authors in a cloud computing scenario.

[0067] Example 1:

[0068] Quantitative stylistics (or style measurement) refers to quantifying and visualizing the stylistic features exhibited by an author in a text, which can be extracted and used to identify the author of online texts. In the past decade, author identification analysis of both physical and electronic texts has had a wide range of applications. Author identification analysis can be conducted from three different perspectives: author attribution, author verification, and author profiling. Author attribution refers to determining the most likely author of a target text; author verification includes checking whether the target text was written by a specific individual; and author profiling refers to determining the characteristics of anonymous text authors (such as gender, age, and ethnicity).

[0069] Using quantitative stylistic methods to analyze an author's language style and writing characteristics can enable author verification. This invention provides a method for verifying the identity of short message authors in a cloud computing scenario, such as... Figure 1 As shown, it includes the following steps:

[0070] S100. Obtain at least one short message text from the target author, and preprocess the at least one short message text to summarize it into a long text.

[0071] To obtain the same structured text data and improve classification accuracy, step S100 involves processing the acquired short message text from the target author. Specifically, this includes the following steps:

[0072] (1) Delete all duplicate short message texts;

[0073] (2) Parse the short message text in special format separately and extract the message body. The short message text in special format includes emails.

[0074] (3) Normalize the short message text through a filter. The normalization process includes unifying numbers, special symbols, encoding methods and capitalization, as well as filtering spaces and punctuation.

[0075] S200. For a long text by the target author, the long text is sliced ​​and the feature parameters of each slice are extracted using the n-gram feature extraction method. The feature criteria of the target author are calculated based on the feature parameters of all slices.

[0076] Step S200 involves extracting n-gram features. Feature extraction is the targeted extraction of various features from the text, such as vocabulary, syntax, structure, specific content, and special style markers. Over a thousand stylistic features have been identified and used in text, along with a wide variety of analysis methods. Analyzing a large number of features does not necessarily provide optimal results, as some features offer little predictive information. This embodiment identifies and retains only the most recognizable features. Due to the noise resistance and effectiveness of n-gram features, and because online short messages are unstructured text, this embodiment reduces the number of n-gram features to one, compared to methods used in traditional text analysis that involve calculating the frequency of n-grams in a given text.

[0077] In a specific implementation, for long texts by target authors, the characteristic criteria of the target authors are calculated through the following steps:

[0078] (1) Divide the long text into two parts, namely the first part. Part Two ;

[0079] (2) The second part Dividing a text into p blocks into equal slices is represented as follows: , … ;

[0080] (3) The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ;

[0081] (4) Extract features from each slice using the n-gram feature extraction method to obtain the feature parameter set corresponding to each slice. slice The corresponding set of feature parameters ;

[0082] (5) Use express and The same feature parameters between The proportion of the total is denoted as ;

[0083] (6) By traversal , … ,get The distribution of the numbers is analyzed, and their mean is calculated. With variance ;

[0084] (7) Based on the above and Obtain an initial parameter As a characteristic criterion for the target author The calculation formula is = -( ).

[0085] S300: Calculate the error rate of the slices based on the slices that failed to identify the correct author and the slices that identified the wrong author; perform supervised learning on the feature criteria based on the error rate; and use the supervised learning feature criteria as the target threshold for the target author.

[0086] Step S300 performs supervised learning on the feature criteria of the target author obtained in step S200. Specifically, the implementation is as follows:

[0087] (1) Train the slides and record the number of slides that fail to identify the correct author as . By traversing , … get Value, calculate error rate ;

[0088] The number of slices with incorrect authors identified is denoted as By traversing , … get Value, calculate error rate ;

[0089] (2) According to and value pairs Training is performed, and the variables are... The initial value is set to 1, when Greater than hour, = - ,when Less than hour, = + When both the correct author and the incorrect author are identified, / 10, traverse from here , … until Less than 0.0001, the final result is The value serves as the final feature criterion for the target author.

[0090] S400. Obtain at least one short message text from an author to be verified, and preprocess the at least one short message text to summarize it into a long text.

[0091] To obtain the same structured text data and improve classification accuracy, step S400 involves processing the acquired short message text of the author to be verified. Specifically, this includes the following steps:

[0092] (1) Delete all duplicate short message texts;

[0093] (2) Parse the short message text in special format separately and extract the message body. The short message text in special format includes emails.

[0094] (3) Normalize the short message text through a filter. The normalization process includes unifying numbers, special symbols, encoding methods and capitalization, as well as filtering spaces and punctuation.

[0095] S500. For a long text of an author to be verified, the long text is sliced ​​and the feature parameters of each slice are extracted using the n-gram feature extraction method. The feature criteria of the author to be verified are calculated based on the feature parameters of all slices.

[0096] Step S500 extracts features from the long text of the author to be verified. Specifically, this includes the following steps:

[0097] (1) Divide the long text into two parts, namely the first part. Part Two ;

[0098] (2) The second part Equal slices Block text, represented as , … ;

[0099] (3) The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ;

[0100] (4) Extract features from each slice using the n-gram feature extraction method to obtain the feature parameter set corresponding to each slice. slice The corresponding set of feature parameters ;

[0101] (5) Use express and The same feature parameters between The proportion of the total is denoted as .

[0102] S600. The feature criteria of the author to be verified are compared with the target threshold of the target author, and the comparison result is used to determine whether the author to be verified is the target author.

[0103] In this embodiment, when Greater than the results obtained from supervised learning When, it is considered satisfy The feature criteria are traversed through step S600. , … This determines whether the author to be verified and the target author are the same author.

[0104] The method in this embodiment combines n-gram feature extraction and supervised learning to verify the author identity of short messages in cloud computing scenarios. This method can be applied to verifying the author identity of online messages such as emails, text messages, or microblogs.

[0105] In the specific implementation process, multiple target authors can be selected. For each target author, steps S100-S300 are executed to calculate the target threshold for each target author, and steps S400-S500 are executed to calculate the feature criteria of the author to be verified. Then, step S600 is used to compare the feature criteria of the author to be verified with the target threshold of each target author to confirm whether the author to be verified is one of the above target authors.

[0106] Example 2:

[0107] This invention discloses a short message author authentication system in a cloud computing scenario, comprising a text processing module, a feature extraction module, a feature supervised learning module, and a comparison analysis module. The system can execute the methods disclosed in the embodiments.

[0108] The text processing module is used to obtain at least one short message text from a target author, preprocess the at least one short message text, and summarize it into a long text; and to obtain at least one short message text from an author to be verified, preprocess the at least one short message text, and summarize it into a long text.

[0109] To obtain the same structured text data and improve classification accuracy, the text message text is preprocessed by a text processing module. In this embodiment, the text processing module operates as follows:

[0110] (1) Delete all duplicate short message texts;

[0111] (2) Parse the short message text in special format separately and extract the message body. The short message text in special format includes emails.

[0112] (3) Normalize the short message text through a filter. The normalization process includes unifying numbers, special symbols, encoding methods and capitalization, as well as filtering spaces and punctuation.

[0113] The feature extraction module is used to slice the long text of the target author and extract feature parameters for each slice using the n-gram feature extraction method. Based on the feature parameters of all slices, the feature standard of the target author is calculated. It is also used to slice the long text of the author to be verified and extract feature parameters for each slice using the n-gram feature extraction method. Based on the feature parameters of all slices, the feature standard of the author to be verified is calculated.

[0114] The feature extraction module extracts features from the long texts of the target author and the author to be verified, respectively.

[0115] For long texts from the target author, the feature extraction module performs the following:

[0116] (1) Divide the long text into two parts, namely the first part. Part Two ;

[0117] (2) The second part Dividing a text into p blocks into equal slices is represented as follows: , … ;

[0118] (3) The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ;

[0119] (4) Extract features from each slice using the n-gram feature extraction method to obtain the feature parameter set corresponding to each slice. slice The corresponding set of feature parameters ;

[0120] (5) Use express and The same feature parameters between The proportion of the total is denoted as ;

[0121] (6) By traversal , … ,get The distribution of the numbers is analyzed, and their mean is calculated. With variance ;

[0122] (7) Based on the above and Obtain an initial parameter As a characteristic criterion for the target author The calculation formula is = -( ).

[0123] As can be seen, the feature extraction module provides a standard feature value for the target author.

[0124] For long texts by authors to be verified, the feature extraction module performs the following:

[0125] (1) Divide the long text into two parts, namely the first part. Part Two ;

[0126] (2) The second part Equal slices Block text, represented as , … ;

[0127] (3) The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ;

[0128] (4) Extract features from each slice using the n-gram feature extraction method to obtain the feature parameter set corresponding to each slice. slice The corresponding set of feature parameters ;

[0129] (5) Use express and The same feature parameters between The proportion of the total is denoted as .

[0130] The feature-supervised learning module is used to calculate the error rate of slices based on slices that fail to identify the correct author and slices that identify the wrong author, perform supervised learning on the feature criteria based on the error rate, and use the supervised learning feature criteria as the target threshold for the target author.

[0131] The feature-supervised learning module is used to supervise the learning of the target author's feature criteria obtained through the feature extraction module. Specifically, this module performs the following:

[0132] (1) Train the slides and record the number of slides that fail to identify the correct author as . By traversing , … get Value, calculate error rate ;

[0133] The number of slices with incorrect authors identified is denoted as By traversing , … get Value, calculate error rate ;

[0134] (2) According to and value pairs Training is performed, and the variables are... The initial value is set to 1, when Greater than hour, = - ,when Less than hour, = + When both the correct author and the incorrect author are identified, / 10, traverse from here , … until Less than 0.0001, the final result is The value serves as the final feature criterion for the target author.

[0135] The comparison analysis module is used to compare the feature criteria of the author to be verified with the target threshold of the target author, and determine whether the author to be verified is the target author based on the comparison results.

[0136] In a specific implementation, the comparison and analysis module is used to determine whether the author to be verified is the target author based on the following rules: when Greater than the results obtained from supervised learning When, it is considered satisfy Feature criteria, traversal , … This determines whether the author to be verified and the target author are the same author.

[0137] The present invention has been shown and described in detail above with reference to the accompanying drawings and preferred embodiments. However, the present invention is not limited to these disclosed embodiments. Based on the above multiple embodiments, those skilled in the art will know that more embodiments of the present invention can be obtained by combining the code review methods in the different embodiments above. These embodiments are also within the protection scope of the present invention.

Claims

1. A method for verifying the identity of short message authors in a cloud computing scenario, characterized in that, Includes the following steps: Obtain at least one short message text from the target author, preprocess the at least one short message text, and then summarize it into a long text; For a long text by a target author, the long text is sliced ​​and the feature parameters of each slice are extracted using the n-gram feature extraction method. The feature criteria of the target author are calculated based on the feature parameters of all slices. The error rate of the slices is calculated based on the slices that fail to identify the correct author and the slices that identify the wrong author. The feature standard is then subjected to supervised learning based on the error rate, and the supervised learning feature standard is used as the target threshold for the target author. Obtain at least one short message text from an author to be verified, and preprocess the at least one short message text and then summarize it into a long text. For a long text by an author to be verified, the long text is sliced ​​and the feature parameters of each slice are extracted using the n-gram feature extraction method. The feature criteria of the author to be verified are calculated based on the feature parameters of all slices. The feature criteria of the author to be verified are compared with the target threshold of the target author, and the comparison result is used to determine whether the author to be verified is the target author; For long texts by the target author, the feature criteria of the target author are calculated through the following steps: The long text is divided into two parts, namely the first part. Part Two ; Part Two Dividing a text into p blocks into equal slices is represented as follows: , … ; The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ; Features are extracted from each slice using the n-gram feature extraction method, resulting in a set of feature parameters for each slice. slice The corresponding set of feature parameters ; use express and The same feature parameters between The proportion of the total is denoted as ; By traversal , … ,get The distribution of the numbers and their mean values ​​are calculated. With variance ; Based on the above and Obtain an initial parameter As a characteristic criterion for the target author The calculation formula is = -( ).

2. The short message author authentication method in a cloud computing scenario according to claim 1, characterized in that, For both the target author's and the author's text messages, the text messages are preprocessed using the following steps: Delete all duplicate text messages; The special format short message text is parsed separately and the message body is extracted. The special format short message text includes emails. The text message is normalized by using filters. The normalization process includes unifying numbers, special symbols, encoding methods, and capitalization, as well as filtering spaces and punctuation.

3. The short message author authentication method in a cloud computing scenario according to claim 1, characterized in that, The error rate of the slices is calculated based on the slices that failed to identify the correct author and the slices that identified the wrong author. Supervised learning of the feature criteria is then performed based on the error rate, including the following steps: The slices are trained, and the number of slices that fail to identify the correct author is recorded as follows: By traversing , … get Value, calculate error rate ; The number of slices with incorrect authors identified is denoted as By traversing , … get Value, calculate error rate ; according to and value pairs Training is performed, and the variables are... The initial value is set to 1, when Greater than hour, = - ,when Less than hour, = + When both the correct author and the incorrect author are identified, / 10, traverse from here , … until Less than 0.0001, the final result is The value serves as the final feature criterion for the target author.

4. The short message author authentication method in a cloud computing scenario according to claim 3, characterized in that, For long texts with authors to be verified, the characteristic criteria of the authors to be verified are calculated through the following steps: The long text is divided into two parts, namely the first part. Part Two ; Part Two Dividing a text into p blocks into equal slices is represented as follows: , … ; The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ; Features are extracted from each slice using the n-gram feature extraction method, resulting in a set of feature parameters for each slice. slice The corresponding set of feature parameters ; use express and The same feature parameters between The proportion of the total is denoted as .

5. The short message author authentication method in a cloud computing scenario according to claim 4, characterized in that, The process of comparing the feature criteria of the author to be verified with the target threshold of the target author, and determining whether the author to be verified is the target author based on the comparison result, includes the following steps: when Greater than the results obtained from supervised learning When, it is considered satisfy Feature criteria, traversal , … This determines whether the author to be verified and the target author are the same author.

6. A short message author authentication system in a cloud computing scenario, characterized in that, The system includes: The text processing module is used to acquire at least one short message text from a target author, preprocess the at least one short message text and summarize it into a long text; and to acquire at least one short message text from an author to be verified, preprocess the at least one short message text and summarize it into a long text. The feature extraction module is used to slice the long text of the target author and extract feature parameters for each slice using the n-gram feature extraction method, and calculate the feature standard of the target author based on the feature parameters of all slices; and is also used to slice the long text of the author to be verified, extract feature parameters for each slice using the n-gram feature extraction method, and calculate the feature standard of the author to be verified based on the feature parameters of all slices. The feature-supervised learning module is used to calculate the error rate of a slice based on slices that fail to identify the correct author and slices that identify the wrong author, perform supervised learning on the feature standard based on the error rate, and use the supervised learning feature standard as the target threshold for the target author. The comparison analysis module is used to compare the feature criteria of the author to be verified with the target threshold of the target author, and determine whether the author to be verified is the target author based on the comparison results. For long texts by the target author, the feature extraction module calculates the feature criteria of the target author through the following steps: The long text is divided into two parts, namely the first part. Part Two ; Part Two Dividing a text into p blocks into equal slices is represented as follows: , … ; The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ; Features are extracted from each slice using the n-gram feature extraction method, resulting in a set of feature parameters for each slice. slice The corresponding set of feature parameters ; use express and The same feature parameters between The proportion of the total is denoted as ; By traversal , … ,get The distribution of the numbers and their mean values ​​are calculated. With variance ; Based on the above and Obtain an initial parameter As a characteristic criterion for the target author The calculation formula is = -( ).

7. The short message author authentication system in a cloud computing scenario according to claim 6, characterized in that, For both the target author's and the author's short message texts, the text processing module preprocesses the short message texts through the following steps: Delete all duplicate text messages; The special format short message text is parsed separately and the message body is extracted. The special format short message text includes emails. The text message is normalized by using filters. The normalization process includes unifying numbers, special symbols, encoding methods, and capitalization, as well as filtering spaces and punctuation. For the long text of the author to be verified, the feature extraction module is used to calculate the feature criteria of the author to be verified through the following steps: The long text is divided into two parts, namely the first part. Part Two ; Part Two Dividing a text into p blocks into equal slices is represented as follows: , … ; The first part is extracted using the n-gram feature extraction method. Feature extraction is performed to obtain a set of feature parameters. ; Features are extracted from each slice using the n-gram feature extraction method, resulting in a set of feature parameters for each slice. slice The corresponding set of feature parameters ; use express and The same feature parameters between The proportion of the total is denoted as .

8. The short message author authentication system in a cloud computing scenario according to claim 7, characterized in that, The feature-supervised learning module is used to perform: The slices are trained, and the number of slices that fail to identify the correct author is recorded as follows: By traversing , … get Value, calculate error rate ; The number of slices with incorrect authors identified is denoted as By traversing , … get Value, calculate error rate ; according to and value pairs Training is performed, and the variables are... The initial value is set to 1, when Greater than hour, = - ,when Less than hour, = + When both the correct author and the incorrect author are identified, / 10, traverse from here , … until Less than 0.0001, the final result is The value serves as the final feature criterion for the target author; The comparison analysis module is used to compare the feature criteria of the author to be verified with the target threshold of the target author. Greater than the results obtained from supervised learning When, it is considered satisfy Feature criteria, traversal , … This determines whether the author to be verified and the target author are the same author.

Images