=Paper=
{{Paper
|id=Vol-3187/paper6
|storemode=property
|title=Machine Training for Intelligent Analysis of Text for the Identification of the Author
|pdfUrl=https://ceur-ws.org/Vol-3187/paper6.pdf
|volume=Vol-3187
|authors=Nadiia Pasieka,Vasyl Sheketa,Myroslava Kulynych,Yulia Romanyshyn,Svitlana Chupakhina,Olha Khytrova
|dblpUrl=https://dblp.org/rec/conf/cpits/PasiekaSKRCK21
}}
==Machine Training for Intelligent Analysis of Text for the Identification of the Author==
https://ceur-ws.org/Vol-3187/paper6.pdf
Machine Training for Intelligent Analysis of Text
for the Identification of the Author
Nadiia Pasieka1, Vasyl Sheketa2, Myroslava Kulynych3, Yulia Romanyshyn2,
Svitlana Chupakhina1, and Olha Khytrova4
1
Vasyl Stefanyk Precarpathian National University, Ivano-Frankivsk, 76000, Ukraine
2
National Tech. University of Oil & Gas, Ivano-Frankivsk, 76068, Ukraine
3
Ukraine Academy of Printing, Lviv, 79020, Ukraine
4
Chernivtsi Trade and Economic Institute Kiev National University of Trade and Economics, Ukraine
Abstract
The continuous development of information technology has led to an increasing danger and
critical cyberattacks, which have recently developed and penetrated unimpeded in various
institutions that have a sophisticated infrastructure of information technology use. Based on
the analysis of the last three years, there have been critical cases of cybercrime around the
world, primarily involving significant leaks of critical information, the spread of fake
messages, cyberbullying, and cloud-based cryptojacking. As a result, scientific research has
sprung up around the world to unambiguously identify the cybercriminal. For this purpose,
various agencies have improvised innovative methods to combat this vice, as well as on the
possibility of bringing the perpetrators to justice, in connection with such critical cybersecurity
issues. As one option to effectively address this problem, the Forensic Writer Identification
system, which works on the principles of stylometry is being considered. Indeed, the
intellectual analysis of text for belonging to one or another author is a complex technological
task, at its base uses artificial intelligence technology to identify, protect, recognize, create,
extract and document digital evidence, which can then be used as evidence of wrongdoing
regarding social media users or simply to analyze critical data. Thus, the main goal of this study
is to examine in detail the capabilities of Forensic Writer Identification technology to analyze
the tweets of different users around the world and unequivocally and apply it to reduce the
search time for criminals by providing the police with the most accurate methodology. As well
as to compare the accuracy of different methodologies. Conducted analytical research behind
a logical literature review that examines the most important methods of text analysis. The study
used texts from various Twitter users for intelligent analysis. Various online and offline
databases were used to expedite the study, and information systems were used to efficiently
search for relevant scholarly results. As systems analysts have recently emphasized computer
methods for rapid analysis of digital text in order to establish authorship, the results presented
are very encouraging. Thus, this research provides a general framework and rationale for the
use of text and author identification methods. This article reviews current research methods
and software applications, and touches on the issues of evaluating the performance of such
research. Various research strategies for digital text research are summarized, and a more
detailed description of two combined methods is presented. Thus, through the use of textures,
algorithms, and polygraphs, new technologies are beginning to show valuable levels of
performance. Nevertheless, the use of combined methods to analyze text for its identity will
play a vital role in future technologies. In this regard, the goal of formulating a project proposal
is to create an analytical analysis system that automatically recognizes authorship of all aspects
of technology on a global scale, which may partially solve the problems of modern cybercrime.
Keywords 1
Machine learning, tweet analysis, cyberbullying, cybercrime, text analysis, stylometry.
CPITS-II-2021: Cybersecurity Providing in Information and Telecommunication Systems, October 26, 2021, Kyiv, Ukraine
MAIL: pasyekanm@gmail.com (N. Pasieka); vasylsheketa@gmail.com (V. Sheketa); kumyr@ukr.net (M. Kulynych);
yulromanyshyn@gmail.com (Y. Romanyshyn); cvitlana2706@gmail.com (S. Chupakhina); olga_hitrova@ukr.net (O. Khytrova)
ORCID: 0000-0002-4824-2370 (N. Pasieka); 0000-0002-1318-4895 (V. Sheketa); 0000-0002-9271-7855 (M. Kulynych); 0000-0001-7231-
8040 (Y. Romanyshyn); 0000-0003-1274-0826 (S. Chupakhina); 0000-0003-2253-4356 (O. Khytrova)
©️ 2022 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
58
1. Introduction
The permanent process, the development of innovations in digital communications, such as network
technologies (social networking platforms), SMS, online forums, blog posts and emails, have defined the
initial conditions for a faster as well as effective transmission of digital information through an innovative
intelligent IT-infrastructure. Under certain conditions, the author of texts can remain incognito, but
unfortunately, this status can also lead to various cybersecurity problems on all available social networks.
Digital forensics of anonymous texts for authorship is the process of using information technology, namely
the storage, extraction and transmission of data from various technical platforms to identify evidence and
combine it into useful information that can be effectively used to solve network problems like cyberbullying
around the world. Thus, effective digital, authentication and attribution of a specific author of an online
source text is a critical tool in combating cybercrime. Unfortunately, in our study, the length of the original
text is limited to a few steps, which complicates the process of effective testing. [1, 2, 4, 6] Thus, the goal
of our study is to process the differentiation of triggering messages from Twitter that are limited to one
hundred and forty characters in length. In our study we consider and evaluate the basic stylometric signs of
the digital text belonging to a particular author, for this purpose we use abstract methods for examination,
as well as explicit signs for the social network Twitter, such as URLs, hashtags, responses or statements.
[3, 5, 11, 16, 28] Subsequently, such analytical approaches are able to achieve an optimum accuracy of
about eighty percent in determining text authorship. Thus, with the increasing number of cybercrimes
against social network users around the world, there is a critical need to identify the author of digital text
with an acceptable and adequate quality of author identification, there is an important issue for scientific
expertise in order to effectively combat cybercrime, such as cyberbullying, emasculative notes and
deception on the Internet by various criminals.
Of course, the anonymity provided by phones with ready-made SIM cards, public networks, public Wi-
Fi zones and decentralized network resources, for example such as Tor, an application that hides the
personal attributes of the user, can make the task of identifying online customers much more difficult. Thus,
sometimes the content of a separately published message text becomes the only effective information to
identify its creator. At the same time, it is also worth noting that the real existence of online text messaging
clients sometimes turns out to be quite different from what they appear to be and how they look on the
Internet. [7, 9, 12, 24, 36] However, it is necessary to emphasize the permanent processes that change the
nature as well as the results of this and are constantly evolving. In addition, in various publications
circulating in the Internet space there is an example of a media organization which in its activities has
allegedly created and used pseudonymous virtual personages to conduct a coordinated campaign of
disinformation through the network media systems. Along with this, it is also expected that some of these
tasks to determine the authorship of text messages in order to stop illegal actions will be supported by state
structures.
To correctly identify an anonymous author from a published text message (which looks like a
cyberbullying) through the analytical process of authorship identification in social and public networks is
presented in the model Figure 1.
The permanent development of the Internet offers more and more opportunities for cybercriminals to
secretly spread their malicious deeds, such as phishing, cyberbullying, fraud and spam. To this end, an
original study has been proposed to determine the authorship of such digital works on the Internet (e.g.,
discussion comments, analysis of text messages, tweets, SMS) The use of scientific phonetics provides a
special mathematical apparatus for the examination of such digital text messages based on a certain
linguistic evidence of belonging to a particular author Thus, online identity is ways of orthographic and
phonetic composition of original digital text, and an This study also involves computer analysis of text
messages for combining styles or style metrics from an archive of such malformed messages. [8, 13, 18,
40] The study solves several different problems, namely the automated text message extraction analysis, in
order to determine and identify the real author of the analyzed content. Thus, the largest subjects of storage
and generation of digital text messages make considerable efforts involving data analysts. [10, 17, 31]
However, the use of applications without advanced artificial intelligence (AI) functions, does not provide
sufficient information to unambiguously identify the authorship of digital text messages, both real and
59
suspected cybercriminals. The Internet provides a useful environment (arena) for cybercriminals to covertly
carry out their intended activities, such as phishing, fraud and spam.
Figure 1: A model for identifying the author of a text message
Thus, the relevance of original research on anonymous digital text messages such as, discussions,
comments on them tweets. Therefore, the qualitative determination of the authorship of anonymous text
messages directed against users in the Internet environment and received serious direction in the analytical
analysis of the available information. Scientific phonetics interrogates experts who investigate linguistic
turns (unknown or authorized works used in cybercrimes) during the preliminary analysis of digital text
messages. In addition, the analytical examination of digital text messages, which malicious actors tend to
obscure problems with online content, are called creation checks. [14, 15, 21] Essentially, online digital
text message originality checks are an examination of voice or computational attributes that may be
recorded by a combination of known or unknown software systems. In addition, the process of digital text
message author identification can also include checking the composition style or style accents in the
investigated signature. [19, 22] It should also be emphasized that there are various challenges to
implementing this process, for example, manual extraction and automatic text analysis to identify the true
author of a digital text message has become a significant challenge for various specialized units. Based on
the paradigm that most of the specialized units that are engaged in deep analysis of big data, use computer
systems that do not have innovative capabilities with elements of artificial intelligence, and therefore cannot
cope with the task of identifying the author of a digital text message, both real and perceived.
In addition, various projected applications around the world lack basic artificial intelligence algorithms
for analyzing digital text messages for anonymous author identification, which can solve some delicate
60
problems, such as language attribute recognition in electronic communication services. Internet short
messages, such as the signature set in Twitter characters, have some attributes that make the origin of the
location test comparison, and moreover, the official content of the artistic function is written as follows.
[19, 26, 32, 37] It should be emphasized that the web digital text message is short or long in most cases,
indicating that a specific language complex measure that depends on the number of words in the content
may not be appropriate.
In addition, some computer applications may not recognize parts of speech. In fact, part-of-speech
tagging classifies the parts of speech for each word in online content, taking into account the quality of the
word and the specific circumstances in which it appears. In addition, in most cyberattacks, cybercriminals
use false and fraudulent methods to gain access to various Internet and Web technologies, manipulating
user information without user authorization. Undoubtedly, the constant growth in the use of information
and communication systems and technological equipment has led to an increase in global cybercrime and
conflicts between governments and criminals. [20, 23, 35, 41] Ultimately, in order to correctly identify an
anonymous user, a significant volume of banned digital text messages must be processed for analysis.
However, cybercrime is generating more and more new methods and techniques for its unlawful acts and
as a result there are more and more cyberattacks on various Internet sites. [25, 27] As a rule, cybercriminals,
for their unlawful deeds, use mass distribution in social networks specially programmed false accounts,
which they use as:
- а means to legally carry out cybercrime business;
- organizing the embedding of special information flows with the purpose of discrediting both
organizations and private users of social networks;
- unlawful possibility of mass theft of personal data of social network users;
- artificial creation of certain conditions for deterioration of trust in social networks;
- the artificial creation of fake news feeds and fake votes (ratings);
- artificial creation of problems for social marketing.
To achieve the task of identifying an anonymous author of a digital text message we used methods of
computer visualization (for graphical representation of social networks), methods of collecting personal
data from various Internet services with analytical analysis, mathematical methods of visual pattern
recognition, methods of piece intelligence, methods of hierarchical clustering, neural network methods,
fuzzy clustering methods (to conduct clustering Web users into groups with the same attributes), as well as
simulation modeling (to analyze the results obtained).
2. Analysis Literature
In fact, some of the nastiest digital text messages posted in the Internet environment and on various
social media platforms can be used in criminal investigations of content authorship, especially through
the method of forensic phonetic identification of anonymous authors. Unfortunately, many authors of
such illegal online content messages still remain out of reach. Despite the fact that on some digital
platforms where web technologies are hosted, the problem of identifying the anonymous author of a
particular digital textual content can be quite a challenge, since the length of digital text messages on
the Internet has limitations on a certain number of characters. [29, 30] Consequently, the goal of a
responsible data analyst is to identify the anonymous sender of a tweet or the owner of a virtual Twitter
account, where a tweet digital text message can be up to two hundred and eighty characters long.
Therefore, it became a matter of developing an algorithm to evaluate the key metrics used for initial
attribution, which can identify the key characteristics of each cybercriminal's compositional style.
Specifically, research data analysts applied stylistic measurement techniques to various digital text
messages for natural language processing (NLP), including specifying an anonymous author, author
profile, author verification, detecting changes in a particular text style, and grouping different tweets.
[22, 34, 39] Obviously, stylistic algorithms were used to identify the anonymous author of the digital
textual content or the blog of a given tweet from the so-called list of investigated authors in order to
achieve a positive result at the center of this study. The aforementioned methods were used to test
61
different data sets, and finally, ideas for future work were considered. In fact, this research topic has
important implications for the application of data analysis tools to help reduce cybercrime worldwide.
Obviously, with offline and online digital text messages, it is not easy to track any direct intra-
species handwritten creations to identify or verify an anonymous author. Nevertheless, many research
papers have been published on author-specific technology for identifying an anonymous author from
digital text messages, which has become a basic tool for big data analytics. Therefore, they have been
used to use stylistic measurement techniques to analyze social media texts for this article. In addition,
online social networks (OSN), such as Facebook, LinkedIn and Twitter, provide increasingly new
factors for triggering anonymous attribution. [33, 38–42] These online tools are believed to provide
effective and quick virtual linking methods that help anonymous social media users commit
cybercrimes. Essentially, users can use screen names, aliases, or VPNs in these places, and other users
may not provide the correct credentials for the record. In fact, to obtain meaningful test results, the data
scientist or analyst should use an information validation method based on reliable information that
contains known, validated models and key points applicable to the test being performed. Will be most
accurate as shown in Figure 2 below. In addition, to compare multiple literature articles, the report
should consider the following aspects.
Figure 2: Model of the information-analytical system for identifying an anonymous author
3. Analytical identification versus verification
In essence, semantic language properties are focused on obtaining information about the creators of
digital textual content in the field of stylistics. In addition, most big data analysts use a variety of stylistic
identity verification methods to identify anonymous customers based on their digital text messages. This
study details the digital content originality verification methods used for sequential validation using
unstructured online partitions based on message content. The online file is then split into continuous grids
of short messages, and (sequential) confirmations are selected in these grids to separate true and self-
proclaimed practices. To achieve the exact goal of the study, analysts are encouraged to analyze content
containing one hundred forty, two hundred eighty, and five hundred characters based on the proposed
algorithm, which is designed to extract the digital text message in the study. In addition, the list of features
includes traditional digital text selection such as morphology, syntax, explicit selection, as well as
innovative selection created in the study of n-grams. In addition, the proposed method includes a number
of methods to circumvent the emerging problems associated with unbalanced sets of big data, and uses the
obtained information and the analyzed information as component selection methods and support vector
mechanisms (SVMs) for grouping digital text messages. Experimental evaluation of the proposed method
and text message parsing algorithm based on Enron and Twitter emails gave promising results: according
to the syntax of the algorithm used, the EER is approximately 23.86%. In addition, the clustering of the
analyzed digital text message can be obtained by the formula (1) shown in Figure 3 below, where A and λ
62
(sigma) represent the normalization of the big data set, SK represents the Gaussian distribution of the data
set, and Du represents the u-th digital text message.
𝑛0 𝑛0
1 1 1 (𝑆 − 𝑆𝑘 )
𝑃𝑡,𝑢 (𝑆) = ∑ 𝐺(𝑆𝑘 ) = ∑ 𝑒𝑥𝑝 [1 ] (1)
𝐴
𝑘=1
𝐴 √2𝜋 𝑘=1
2
4. Methodologies for analyzing digital text messages
As one of the criteria for identifying digital text messages or tweets, the research uses stylistic
methods to accurately perform the post-creation task of identifying anonymous authors of content. In
this section, we will comprehensively describe the main methods that can be used to automatically
identify an anonymous author across multiple Twitter accounts. Essentially, the desired methodology
involves the selection of software that uses stylometric, semantic algorithms and the fleeting parts of
authoring mechanisms.
Stylometric mechanisms for identifying an anonymous author by content.
Of course, different analysts use stylistic variables to identify the anonymous authorship of amateur
bloggers on social media as one way to tackle the growing problem of cybercrime. In addition, the
stylistic analysis of the corpus is usually one of the main methods of identifying the author of a Twitter
post. Usually their consideration depends on the field of application. For example, the source attribution
method and other short online posts include basic style tags such as good news, document links, specific
HTML tags, etc., or bizarre style tags such as misspellings. Since our research area is abstract writing,
we decided to include a wider range of style markers. The stylistic variables commonly used in this
study are specific aspects such as punctuation in tweets and n-gram text arrays.
In fact, the corpora used in this study usually consists of multiple digital text messages from several
anonymous authors. Like any corpus-based method, corpus configuration is strictly considered to
produce reasonable research results. Regarding source attribution, the characteristics of the corpus (text
type, language, time, case) affect the accuracy of attribution. In addition, the ideal consensus
recommendation is based on the summary of the author's company, taking into account the narrowest
possible language range. In addition, we assume that we have received a set of predetermined posts or
tweets as information in online media, and each method will generate a similarity score for each pair of
accounts analyzed. In fact, the corpus used in this study usually consists of multiple posts by multiple
authors. As with every model-based method, the configuration of the model is considered from a critical
point of view to produce the most reasonable research results. Regarding the attribution of provenance,
model characteristics (text type, language, time, case) affect the accuracy of attribution. In addition, it
has been clearly suggested that it is ideal to accumulate a corpus of digital text messages based on
anonymous authors, who have considered the narrowest possible language range. In addition, we
assume that we have obtained a set of time-stamped posts or digital text messages (tweets) as
information in online media. Therefore, each suggested determination method will generate a similarity
score for each pair of accounts analyzed.
In addition, this technology is used in conjunction with the spry information mining program that is
essential for text extraction, and the check also corresponds to the four important stages of tweet
recognition, such as preprocessing different tweets, extracting tweet features, and comparing the same
tweets And to determine the identity of the author.
Logically speaking, the stylistic research of tweets is similar to the stylistic research of different
types of short messages, such as online discussion posts or online text conversations. It is worth noting
that they are random and comparative in design and syntax. For our tests, a complete feature list has
been developed, taking style data into account in any case, and assuming that creators unknowingly
follow a certain design and are predictable in their decisions. In the analysis process, the algorithm will
use iterative techniques to search the array tree and identify various characteristic vocabulary features-
this will help produce the most accurate results.
In addition, preprocessing tweets involves deleting suspicious tweets from some authors. In fact, the
proposed tweet preprocessing technique relies on the association of slang words with other matching
words to check the importance of slang words and their presumed interpretation. We use n-gram to
detect connections and condition arbitrary fields to check the meaning of slang terms. However, an
63
important issue in this field is data matching about hype, relevance, emoji, folklore classification, and
slang.
Recognizing the characteristics of tweets involves analyzing pre-suggested tweets to determine the
total number of words in a particular tweet, the special characters used, the consonants and vowels, the
total number of characters, and the regularity of the use of certain words or phrases. In fact, recognizing
appropriate selections for learning opinions in tweets remains an open area of research, as text-ordering
methods face the problem of parsimony and part-of-speech (POS) labeling strategies bombard the lack
of linguistic construction of tweets. The character-based selection, namely n-gram characters, is now
very mature because they do not contain any language.
Logically speaking, the analyst must find content that matches the content stored in the database in
order to use feature matching techniques to detect similarities when possible. In addition, the feature
evaluation algorithm of selected tweets can be used to directly identify friendships and find evaluations.
For example, decision-making evaluation is an important area, and presumption checking of tweets has
been widely used in the past few years.
Lexical
This is the technique of Twitter sentiment investigation. In fact, it attempts to quantify the diversity
of popular assessments of retail signs. The first is a vocabulary-based strategy, which uses word
references and semantic scores of words to calculate the final endpoint of a tweet and incorporate
grammatical feature tags. In addition, with the rapid spread of interpersonal organizations, Weibo
applications, and gatherings, its main role is developing significantly.
Structural
Basically, Twitter is a Weibo platform that spreads about 450 million tweets every day. It requires
a structured algorithm to accurately analyze all data. In addition, it also solves the important information
hotspots of disease fighting and control in local areas. This research investigates the structural algorithm
methods used to separate and analyze Twitter information, including the attributes and
representativeness of the information; information sources, access, and cost; inspection methods;
information boards and cleanliness; standardized measurements; and examinations.
Syntactic
In essence, the language used on Twitter has some peculiarities, such as the use of hashtags or client
references. Therefore, data preprocessing technology will use syntactic algorithms or data structures to
speed up the analysis process. In order to improve the effectiveness of language preparation techniques
(such as morphological restoration and syntactic analysis), we have performed some standardization
steps. We eliminate #images, all @ notices, and connect and perform lowercase conversion. Similarly,
if a vowel is repeated multiple times in a word, we reduce it to a single event, and reduce the different
back-to-back accent marks to a single event. Finally, we lemmatize the standardized content.
N-grams
With the advancement of global Web technology, the Internet has been used as a hot spot to get
news about the latest developments. Recently, Twitter is probably the most famous online media
platform that allows public customers to share news. This stage develops rapidly, especially among
young people who may be affected by data from mysterious sources. In this way, foreseeing the
credibility of information on Twitter becomes a need, especially in the event of a crisis. This paper
proposes an arrangement model based on managed AI strategy and word-based N-gram analysis to
naturally rank Twitter messages as sounds instead of entities. Applied and specifically analyzed five
different management characterization programs: Linear Support Vector Machine (LSVM), Logistic
Regression (LR), Random Forest (RF), Naive Bayes (NB) and K-Nearest Neighbor (KNN). This exam
explores two descriptions (TF and TF-IDF) and different word N-gram ranges. For model preparation
and testing, a 10-crease cross-approval was performed on two data sets in different dialects (such as
English). In addition, the application of syntactic concepts will also be used to relieve the sentence
structure about stress and the grammatical form-part of speech. According to the theory, punctuation is
an important guide to characterize restrictions and make tweets or posts meaningful by dividing
paragraphs into sentences and dividing each sentence into different tags, such as exclamation marks,
quotations, and periods. The arrangement of basic accents includes single sentences, commas, periods,
colons, semicolons, question marks, and shout signs. In addition, according to the Unicode design with
code and specific grammar, a series of notable emphasis is made using various symbols, such as (!, ").
The function is subject-specific and captures the creator’s style in various themes These highlights
64
include the number of sentences per square of text, the number of normal characters, words and
sentences in a square text, and the number of normal sentences beginning with uppercase and lowercase.
Text Representation
Logically, for each style variable, the data analyst creates a loop vector, where each measurement is
related to an alternative component. In addition, in most experiments, data analysts will select different
combinations of factors, and then treat each report as a link to a repeating vector related to these factors.
Basically, some of the important features include the total number of characters in each tweet, the total
number of words in each sentence in the tweet, the frequency of words in the tweet, the total number of
common words in the dictionary, emoji in the tweet and its appearance in the tweet. In the meaning of
Google and various online applications, trend words are for specific reasons-maybe a day or a week or
so. Fundamentally, the main period of Twitter content confirmation will be pre-measurement of
different tweets from certain clients, where untrusted parts will be removed from content written by
others, for example, other web-based media-Twitter-customers Tweets forwarded by the end.
Many etymological characters have been recommended for origin checking, for example, the
determination of specific words and syntactic structure. In sharp contrast to topic-based content
arrangements (the essential issue is a set of texts), key arrangements are expanded by adding
vocabulary, syntax, and application explicit focus. Such a mix can better convey the creator's style. In
this way, the expert will use the created calculations to coordinate the highlights of the indistinguishable
content characters. This is done after the component is discovered, and it is compared with the data sets
of different authors to find similarities. It is worth emphasizing that the grouping model includes various
configuration files created independently for a single customer.
5. Results
Thus, analyzing the results obtained by using different models to determine the anonymous
authorship of digital text messages and machine learning to automate this process, we found that the
best model is the model of increasing the gradient with an accuracy of 37.43%. We can state that the
probability of detection is low, but we can also note that this figure can be directly related to the nature
of the digital text messages processed. Since these anonymous authors can change on bot farms, and it
is very likely that some authors are constantly changing agents, that is why the prediction is so low.
Which will ultimately affect the accuracy of the proposed model.
Due to the cumbersome nature of analytical calculations, we present only the general values of
accuracy of the models considered by the anonymous author by numerical analysis of text messages,
which was for: Gradient Boosting 37.43%; Decision Tree 23.17%; Bayesian Network 19.89%.
As the results of the different models show, they are in line with our expectations, along with
demonstrating the significant potential of different machine learning algorithms in identifying
anonymous social media authors. And some differences in the percentage of detection for each
algorithm successfully shows us that machine learning algorithms can be used to extract useful
information from significant amounts of digital text message data and predict the author of a particular
textual content. This could play a huge role in the future in the fight against cybercrime, cyberbullying,
and identity theft from social media users.
6. Conclusions
The main purpose of the study is to automate the identification of the author of a digital message
using various analytics technologies. To identify the anonymous author using the recognition of his text
messages on the social network Twitter to minimize and prevent cybercrime. In fact, the science of data
analytics has activated the development of a unique stylistic evaluation model for digital text messages
that has the ability to use computer analysis of unique linguistic and stylistic "fingerprints" for affiliation
and author identification. The study attempted to test the proposed concept of digital text message
authorship by testing and evaluating the accuracy of the proposed model, the principle approaches that
were tried to test, proves that they can significantly help to automate these mechanisms for the relevant
structures. The results of the study clearly delineate the research topics. In addition, the evidence
presented in the study subsequently confirmed that the proposed model is an acceptable mathematical
65
model for the process of author identification using stylometry technology. In fact, empirical research
on determining the author of a digital text message has answered almost all of the questions posed and
achieved the basic goals of being reviewed and investigated through computer dialogue and analytic
analysis. However, there is still room for improvement in method models and algorithms, which
requires additional research to achieve better results that will have a positive impact on reducing
cybercrime. Due to the limited amount of data sets used in the proposed model, the ones used in this
study are not highly accurate. These datasets consist of popular Twitter users who are likely to find a
way to continue generating digital text messages or pay penalties to the organizers and continue
participating in social media. Future research will explore other models, methods, and algorithms to
obtain more robust metrics, and will greatly expand the data set that is now being accumulated on
chimera resources. This research is also a successful starting point for further research, and applying
models and methods to digital text messaging data significantly reduces the potential for cybercrime,
which can be useful for various specialized units.
7. References
[1] Adak, C., Chaudhuri, B. B., & Blumenstein, M. (2019). An empirical study on writer identification
and verification from intra-variable individual handwriting. IEEE Access, 7, 24738-24758.
doi: 10.1109/access.2019.2899908
[2] Agile Business Consortium Limited. (2021). Agilebusiness.org. agilebusiness.org.
https://www.agilebusiness.org/page/ProjectFramework_06_Process
[3] Alonso-Fernandez, F., Belvisi, N. M., Hernandez-Diaz, K., Muhammad, N., & Bigun, J. (2020).
Writer Identification Using Microblogging Texts for Social Media Forensics. International Journal
of Recent Trends in Engineering and Research, pp. 1-22
[4] Analysis of Stylometric variables in long and short texts. Procedia - Social and Behavioral
Sciences, 95, 604-611. Retrieved from https://doi.org/10.1016/j.sbspro.2013.10.688
[5] B. Durnyak, B. H. O. Tymchenko, O. Tymchenko and D. Anastasiya, "Research of image
processing methods in publishing output systems," 2018 XIV-th International Conference on
Perspective Technologies and Methods in MEMS Design (MEMSTECH), 2018, pp. 178-181, doi:
10.1109/MEMSTECH.2018.8365728.
[6] B. Durnyak, B. Havrysh, O. Tymchenko, M. Zelyanovsky, O. O. Tymchenko and O. Khamula,
"Intelligent System for Sensor Wireless Network Access: Modeling Methods of Network
Construction," 2018 IEEE 4th International Symposium on Wireless Systems within the
International Conferences on Intelligent Data Acquisition and Advanced Computing Systems
(IDAACS-SWS), 2018, pp. 93-97, doi: 10.1109/IDAACS-SWS.2018.8525792.
[7] B. Durnyak, O. Tymchenko, O. Tymchenko and B. Havrysh, "Applying the Neuronetchic
Methodology to Text Images for Their Recognition," 2018 IEEE Second International Conference
on Data Stream Mining & Processing (DSMP), 2018, pp. 584-589, doi:
10.1109/DSMP.2018.8478482.
[8] Belvisi, N. M., Muhammad, N., & Alonso-Fernandez, F. (2020). Forensic authorship analysis of
Microblogging texts using N-grams and Stylometric features. 2020 8th International Workshop on
Biometrics and Forensics (IWBF), 1-6. Retrieved from https://arxiv.org/pdf/2003.11545.pdf
[9] Dronyuk I., Nazarkevych M., Fedevych O. (2016) Synthesis of Noise-Like Signal Based on Ateb-
Functions. In: Vishnevsky V., Kozyrev D. (eds) Distributed Computer and Communication
Networks. DCCN 2015. Communications in Computer and Information Science, vol 601.
Springer, Cham https://doi.org/10.1007/978-3-319-30843-2_14
[10] Gomez Adorno, H. M., Rios, G., Posadas Durán, J. P., Sidorov, G., & Sierra, G. (2018).
Stylometry-based approach for detecting writing style changes in literary texts. Computación y
Sistemas, 22(1). Retrieved from http://www.scielo.org.mx/scielo.php?script=sci_arttext&pid
[11] Joshi, S., & Deshpande, D. (2018). Twitter Sentiment Analysis System. International Journal of
Computer Applications (IJCA), 180(47), 35-39. Retrieved from https://arxiv.org/ftp/arxiv
[12] Killian, A., Brounstein, T., Skryzalin, J., & Garcia, D. (2019). Stylometric and Temporal
Techniques for Social Media Account Resolution. Sandia National Laboratories Journal, 1-8.
66
https://www.osti.gov/servlets/purl/1456316 Kramer, S. (2020, June 9). Tracking coronavirus
disinformation on Twitter.
[13] Kula, S., Choraś, M., Kozik, R., Ksieniewicz, P., & Woźniak, M. (2020). Sentiment Analysis for
Fake News Detection by Means of Neural Networks. Springer, Cham. Retrieved from
https://link.springer.com/chapter/10.1007/978-3-030- 50423-6_49
[14] M. Pasyeka, V. Sheketa, N. Pasieka, S. Chupakhina and I. Dronyuk, "System Analysis of Caching
Requests on Network Computing Nodes," 2019 3rd International Conference on Advanced
Information and Communications Technologies (AICT), Lviv, Ukraine, 2019, pp. 216-222, doi:
10.1109/AIACT.2019.8847909.
[15] Mariya Nazarkevych, Andrii Marchuk, Lesia Vysochan, Yaroslav Voznyi, Hanna Nazarkevych
and Anzhela Kuza Ateb-Gabor Filtering Simulation for Biometric Protection Systems. CPITS
2020 pp. 14-22
[16] Medykovskyy M., Pasyeka M., Pasyeka N. & Turchyn O. (2017). Scientific research of life cycle
perfomance of information technology. 12th International Scientific and Technical Conference on
Computer Sciences and Information Technologies, CSIT 2017, , 1 425-428. doi:10.1109/STC-
CSIT.2017.8098821
[17] Mishchuk, O., R. Tkachenko, & I. Izonin Missing Data Imputation through SGTM Neural-Like
Structure for Environmental Monitoring Tasks. Advances in Intelligent Systems and Computing.
Vol. 938. 2020, pp. 142-151, doi:10.1007/978-3-030-16621-2_13
[18] Mykhailyshyn H., Pasyeka N., Sheketa V., Pasyeka M., Kondur O. & Varvaruk M. (2021).
Designing network computing systems for intensive processing of information flows of data
doi:10.1007/978-3-030-43070-2_18
[19] N. Pasieka, V. Sheketa, Y. Romanyshyn, M. Pasieka, U. Domska & A. Struk «Models, Methods
and Algorithms of Web System Architecture Optimization» IEEE International Scientific-
Practical Conference Problems of Infocommunications, Science and Technology (PIC S&T),
Kyiv, Ukraine, 2019, pp. 147-153, doi: 10.1109/PICST47496.2019.9061539.
[20] Nazarkevych M., Logoyda M., Dmytruk S., Voznyi, Y. & Smotr O. (2019). Identification of
biometric images using latent elements. Paper presented at the CEUR Workshop Proceedings,
2488 pp. 99-108.
[21] Nazarkevych M., Logoyda M., Troyan O., Vozniy Y. & Shpak Z. (2019, September). The Ateb-
Gabor Filter for Fingerprinting. In International Conference on Computer Science and Information
Technology pp. 247-255. Springer, Cham.
[22] Nazarkevych M., Logoyda, M., Troyan, O., Vozniy, Y., & Shpak, Z. (2019, September). The Ateb-
Gabor Filter for Fingerprinting. In Conference on Computer Science and Information
Technologies (pp. 247-255). Springer, Cham.
[23] Nazarkevych M., Lotoshynska N., Klyujnyk I., Voznyi Y., Forostyna S. & Maslanych I. (2019,
July). Complexity Evaluation of the Ateb-Gabor Filtration Algorithm in Biometric Security
Systems. In 2019 IEEE 2nd Ukraine Conference on Electrical and Computer Engineering
(UKRCON) pp. 961-964
[24] Nazarkevych M., Lotoshynska, N., Brytkovskyi, V., Dmytruk, S., Dordiak, V., & Pikh, I. (2019).
Biometric identification system with ateb-gabor filtering. Paper presented at the 2019 11th
International Scientific and Practical Conference on Electronics and Information Technologies,
ELIT 2019 - Proceedings, 15-18. doi:10.1109/ELIT.2019.8892282
[25] Nazarkevych M., Oliiarnyk R., Nazarkevych H., Kramarenko O., & Onyshschenko I. (2016,
August). The method of encryption based on Ateb-functions. In 2016 IEEE First International
Conference on Data Stream Mining & Processing (DSMP) pp. 129-133.
[26] Nazarkevych, M., Oliarnyk, R., & Dmytruk, S. (2017, September). An images filtration using the
Ateb-Gabor method. In 2017 12th International Scientific and Technical Conference on Computer
Sciences and Information Technologies (CSIT) (Vol. 1, pp. 208-211
[27] Neal, T., Sundararajan, K., Fatima, A., Yan, Y., Xiang, Y., & Woodard, D. (2018). Surveying
stylometry techniques and applications. ACM Computing Surveys, 50(6), 1-36.
doi: 10.1145/3132039
67
[28] Nirkhi, S., Dharaskar, R., & Thakare, V. (2016). Authorship verification of online messages for
forensic investigation. Procedia Computer Science, 78, 640-645. Retrieved from
https://www.sciencedirect.com/science/article/pii/S1877050916001137
[29] Pascual, F. (2020, August 19). Twitter sentiment analysis with machine learning. MonkeyLearn
Blog. Retrieved from https://monkeylearn.com/blog/sentiment- analysis-of-twitter/
[30] Pasieka, N., Sheketa, V., Romanyshyn, Y., Pasieka, M., Domska, U., & Struk, A. (2019). Models,
methods and algorithms of web system architecture optimization. Paper presented at the 2019
IEEE International Scientific-Practical Conference: Problems of Infocommunications Science and
Technology, PIC S and T 2019 – pp. 147-152. doi:10.1109/PICST47496.2019.9061539
[31] Pasyeka M., Sheketa V., Pasieka N., Chupakhina S. & Dronyuk, I. (2019). System analysis of
caching requests on network computing nodes. 3rd International Conference on Advanced
Information and Communications Technologies, AICT2019 - Proceedings, pp. 216-222,
doi:10.1109/AIACT.2019.8847909
[32] Pasyeka M., Sheketa V., Pasieka N., Chupakhina S. & Dronyuk, I. (2019). System analysis of
caching requests on network computing nodes. Paper presented at the 2019 3rd International
Conference on Advanced Information and Communications Technologies, AICT 2019 -
Proceedings, 216-222. doi:10.1109/AIACT.2019.8847909
[33] Riznyk O., Povshuk O., Kynash Y., Nazarkevich M., & Yurchak I. (2018). Synthesis of non-
equidistant location of sensors in sensor network. 14th International Conference on Perspective
Technologies and Methods in MEMS Design, MEMSTECH 2018 - Proceedings, 204-208.
doi:10.1109/MEMSTECH.2018.8365734
[34] S. Babichev, A. Sharko, B. Durnyak, V. Zhydetskyy and I. Izonin, "Application of Huang
Transform and Wavelet Analysis for Acoustic Emission Signal Filtering," 2019 IEEE 2nd Ukraine
Conference on Electrical and Computer Engineering (UKRCON), 2019, pp. 859-863, doi:
10.1109/UKRCON.2019.8879839.
[35] Sharon Belvisi, N. M., Muhammad, N., & Alonso-Fernandez, F. (2020). Forensic authorship
analysis of Microblogging texts using N-grams and Stylometric features. 2020 8th International
Workshop on Biometrics and Forensics (IWBF), 1-6. doi: 10.1109/iwbf49977.2020.9107953
[36] Sikora, L., Lysa, N., Fedyna, B., Durnyak, B., Martsyshyn, R., & Miyushkovych, Y. (2018).
Technologies of development laser based system for measuring the concentration of contaminants
for ecological monitoring. Paper presented at the 2018 IEEE 13th International Scientific and
Technical Conference on Computer Sciences and Information Technologies, CSIT 2018 -
Proceedings, 1 93-96. doi:10.1109/STC-CSIT.2018.8526602
[37] Singh, T., & Kumari, M. (2016). Role of text pre-processing in Twitter sentiment analysis.
Procedia Computer Science, 89, 549-554. https://doi.org/10.1016/j.procs.2016.06.095
[38] Tkachenko, R., Izonin, I., Kryvinska, N., Dronyuk, I., & Zub, K. (2020). An approach towards
increasing prediction accuracy for the recovery of missing iot data based on the grnn-sgtm
ensemble. Sensors (Switzerland), 20(9) doi:10.3390/s20092625
[39] Tkachenko, R., Izonin, I., Vitynskyi, P., Lotoshynska, N., & Pavlyuk, O. (2018). Development of
the non-iterative supervised learning predictor based on the ito decomposition and sgtm neural-
like structure for managing medical insurance costs. Data, 3(4) doi:10.3390/data3040046
[40] V. Buriachok, et al., Invasion Detection Model using Two-Stage Criterion of Detection of Network
Anomalies, Cybersecurity Providing in Information and Telecommunication Systems (CPITS),
pp. 23–32, Jul. 2020.
[41] Y. Romanyshyn, V. Sheketa, L. Poteriailo, V. Pikh, N. Pasieka and Y. Kalambet Social-
communication web technologies in the higher education as means of knowledge transfer. IEEE
2019 14th International Scientific and Technical Conference on Computer Sciences and
Information Technologies (CSIT). – Vol.3. – 2019. – Lviv, Ukraine. – pp. 35–39.
[42] Zharikova M. & Sherstjuk, V. (2017). “Academic integrity support system for educational
institution,” 2017 IEEE 1st Ukraine Conference on Electrical and Computer Engineering,
UKRCON 2017 - Proceedings, 1212-1215. doi:10.1109/UKRCON.2017.8100445
68