A Survey on Datamining in Cyber Bullying

International Journal on Recent and Innovation Trends in Computing and Communication
Volume: 2 Issue: 7

ISSN: 2321-8169
1865 – 1869

_______________________________________________________________________________________________

A survey on Datamining in Cyber Bullying
K. Nalini

Dr. L. Jaba Sheela

Research Scholar
Bharathiyar University
Coimbatore
[email protected]

Professor, MCA Department
Panimalar Engineering College
Chennai
[email protected]

Abstract—CRIME has always the very first thing that people want to avoid and police officer want to stop. A major challenge facing all lawenforcement and intelligence –gathering organizations is accurately and efficiently analyzing the growing volumes of crime data. Crime activity
reports available from victims, governmental organizations, news press, and social networks play a significant role in public safety, including
crime prevention, suppression and investigation, uniformed patrol and response. Cyber Crime is an area which covers crimes committed in the
internet. Opportunities for connecting with classmates, friends and people with shared interests abound. Email, online chat, and social
networking sites allow us to interact with people in the same town and people on the other side of the world. Unfortunately the opportunity for
misuse comes with any new technology. There were sexual predators and bullies long before the advent of the internet and chat rooms. Cyber
bullying and internet predation threaten minors, particular teens and teens who do not have adequate supervision when they use the computer.
As the amount of criminal records growing every day, it is impossible to perform manual analysis on the dataset and extract useful information.
Data mining has been studying for decades trying to get useful information out of large amount of data. Many efforts have used automated
techniques to analyze different types of crimes, but without a unifying framework describing how to apply them. In particular, understanding the
relationship between analysis capability and crime type characteristics can help investigators more effectively use those techniques to identify
trends and patterns, address problem areas and even predict crimes.
Keywords—Data Mining, Cyber Bullying, Sexual Predation, Machine Learning, Datasets

__________________________________________________*****_________________________________________________
I.

Introduction

“Cyber bullying is defined as an aggressive,
intentional act carried out by a group or individual using
electronic forms of contact (eg. Email and Chat rooms),
repeatedly or overtime, against a victim who cannot easily
defined herself”.
Cyber bullying consist in sending
messages containing slanderous expressions, harmful for
other people or verbally bullying other people in front of the
rest of online community. According to recent studies
almost 43%
of teens in the United States alone reported being victims of
cyber bullying. In 2011, 70% of teens use social media sites
on daily basis and nearly one in four teens hit their favorite
Social-media sites 10 (or) more times a day. Scan safe’s
monthly “Global Threat Report” found that up to 80% of
blogs contained offensive contents and 74% included porn
in the format of image, video (or) offensive languages.
In particular the story of 13-year-old Megan Meir
brought notoriety to the subject of Cyber bullying when she
committed suicide after being harassed though a popular
social networking site. (ABC NEWS, whereas 10% of the
cases took place for less than 10 years. A 14-year-old
Estonian teenager S.K[18] , who committed suicide in 2009,
after being recurrently harassed by a Paedophile
through the internet. The Paedophile pretended to be a
teenager girl in order to gain access to dozens victims. He
could therefore internet with many children in a seemingly
natural way. Sadly S.K could not bear the constant coercion
from the Paedophile and this soon led to his suicide. The
“MySpace Mom” case[19] is another tragic example:- L.D
and Cyber bullies, pretended to be a teenager boy on
MySpace and befriended a teenager girl (M.M). After
several weeks exchanging messages abruptly ended their

friendship, telling M.M that she was cruel. Some days later
M.M committed suicide. Many other that occur on a yearly
basis across the world, are highly indicative of how severe
Cyber threats are.
Online security has been an important and urgent
problem ever since the creation of the Internet. One of the
burning online security problems are online slandering ,
bullying and particularly when the target subjects are underage victims. Sexual predators have adapted their predatory
strategies to these platforms and usually the target victims
are kids. The number of children who are approached (or)
solicited for sexual purpose through the Internet is
staggering [1] and unfortunately online sexual predators
always outnumber the law enforcement officers available in
Police Cyber Crime Units [2].
Most adverse effects of cyberbullying are seen in
adolescents, though this menace exists among all age
groups. According to a report by Microsoft, India, ranks
third in the world when it comes to cyberbullying of
children and first in cyberbullying cases of adults. It
includes all those activities that are meant to humiliate,
disturb, defame, threaten or insult an individual.
Photoshopping target’s face over obscene pictures, posting
and spreading defaming rumors and blackmailing the victim
over his/her objectionable videos are common examples of
cyber crime. Like other cases of bullying, children may not
inform their parents of the cyber humiliation or threat they
suffer out of fear of social stigma. Results are depression,
anxiety, loss of self esteem, fear and isolation.
Cases of cyberbullying have been reported all over
the world and India is no exception. Twenty one percent of
the suicides among Indian adolescents are due to the trauma
of cyber bullying suffered by the victims, as shown by
studies. This is both due to easy availability of internet as
1865

IJRITCC | July 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication
Volume: 2 Issue: 7

ISSN: 2321-8169
1865 – 1869

_______________________________________________________________________________________________
well as lack of rules against cyber defamation. A recent case
that horrified me to the core was the one in which an Indian
boy committed suicide because one of his friends posted his
video online which had him making out with his cousin
sister. Amanda’s case gained much concern, as that tenth
grade girl suicide after posting her own video recounting the
cyber blackmail she was suffering.
Akash Ambani, the elder son of Mukesh and Nita
Ambani, was one of the top trending tags on the microblogging site Twitter post his brother Anant’s appearance at
an IPL match. He became the subject of snide remarks
targeting his weight. The most re-tweeted post was by
@SirJadeja16h which read: “Red alert. = Expected
earthquake in Kolkata later tonight coz Akash Ambani will
be doing jhamping jhapang after MI win #IPLFinal”.
What many people do not know about the younger
Ambani scion is that the reason behind his obesity is the
usage of steroids to treat his asthma. These tweets not only
openly slander the victim they also breach the fine line
between freedom of expression and potential cyber crime.
What makes cyber bullying so widespread in India is
the fact that unlike many developed countries, India does
not presently have laws to curb it. There are no serious
punishments for cyber offenders. The recent most cyber
crime laws which were made by the parliament in Feb 2013
include only financial matters like cases of fraud and
phishing scams. Indian government must realize that India
being a prominent IT hub, with a large section of society
having access to internet, cyber laws are need of the hour.
In Mumbai, Two cases of cyber bullying, where
profiles of two young women on social networking sites
were replaced with obscene material, ended in acquittal after
the prosecution failed to produce evidence. In both the
cases, the court said the prosecution failed to submit
electronic evidence. In one of the cases, the investigating
officer was not available for examination in the court.
In the first case, in September 2006, a Thane resident
prepared a fake profile of a college student, posted obscene
comments about her and also provided her telephone
number. Following this, the student started receiving vulgar
messages and calls. On her complaint, the police registered a
case, identified the cyber cafe from where the accused had
posted the obscene material and arrested him. The accused
had used his personal computer for the activity.
II.
Literature Survey
Patchin and Hinduja[3] define Cyber bullying as
“Willful and repeated harm inflicted through the medium of
electronic text”. In their recent studies found that students
who experienced Cyber bullying (Both those who were
victims and those who admitted to Cyber bullying others)
perceived a poorer climate at their school than those who
had not experienced Cyber bullying.
In a recent study on Cyber bullying detection Yin, et
al [4] used a supervised learning approach for detecting
harassment. They determined that the base line text mining
system (using a bag of words approach) was significantly
used content, sentiment and contextual features of
documents to train a Support Vector Machine Classifier for

Corpus of Online posts and only the contents of the posts
were used to determine either a post is harassing (or) not and
the characteristics of the author of the posts were not
considered. They have used the combination of 3 features
such as N-gram, TFIDF weighting and foul words frequency
were used as the baselines.
The results shows
improvements over the baselines. Both C4.5 decision tree
learner and an Instance-based learner were used to identify
the true positive with 78% accuracy, by recording the
percentage of curse and insult words within a post.
Dinakar et al.,[5] applied a range of binary and
multiclass classifiers on a manually labeled Corpus of You
tube Comments.
Their findings showed that binary
individual topic-sensitive classifiers can outperform the
detection of textual Cyber Bullying compared to merge data
sets or multiclass classifiers. They have illustrated the
application of commonsense knowledge in the design of
social network software for detection Cyber Bullying. The
authors treated each comment on its own and did not
consider other aspects to the problem as such the pragmatics
of dialogue, conversation and the social networking graph.
They concluded that taking into account such features will
be more useful on social networking websites and can lead
to a better modeling of the problem.
April Kontostathis et al.,[6] used a Language-Based
method of detecting Cyber Bullying. They collected the
data from the website Formspring.me, a question and
answer formatted website that contains a high percentage of
bullying content. The data was labeled using a web service,
Amazon’s mechanical turk. They used the labeled data in
conjunction with machine learning techniques provided by
the Weka Tool Kit, to train the computer to recognize
bullying content. Both C4.5 decision tree learner and an
Instance-based Learner were used to identify the true
positive with 78.5% accuracy, by recording the percentage
of curse and insult words within a post.
Maral Dadvar et al., [7] have investigated the GenderBased Approach for Cyber Bullying detection in Myspace.
They have used the content of the text written by the users
but not the user’s information. They approached Support
Vector Machine model to train a gender-specific text
classifier using WEKA. They have utilized the Myspace
posts as dataset which was provided by Fundacion
Barcelona Media. The dataset consists of more than
3,81,000 posts in about 16,000 threats. Overall 34%of posts
are written by female and 64% by male authors. The
Gender Specific Approach improved the Baseline by 39% in
precision, 6% in recall, 15% in F-measure.
Ying Chen et al.,[8] investigated existing text mining
methods in detecting offensive contents for protecting
adolescent online safety. Specifically, they proposed the
Lexical Syntactical Feature(LSF) approach to identify
offensive contents in social media and further predict a
user’s potentiality to send out offensive contents. Their
research has several contributions. First they practically
conceptualize the notion of online offensive contents and
further distinguish the contribution of pejoratives/profanities
and obscenities in determining offensive contents, and
introduce hand authoring syntactic rules in identifying
name-calling harassment. Second, they improved traditional
1866

IJRITCC | July 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication
Volume: 2 Issue: 7

ISSN: 2321-8169
1865 – 1869

_______________________________________________________________________________________________
Machine-Learning methods by not only using lexical
features to detect offensive language, but also incorporating
style feature, structure features and Content-specific features
to better predict a user’s potentiality to send out offensive
content in social media. Experiment result shows that the
LSF Sentence offensiveness prediction and user
offensiveness estimate algorithm outperform, traditional
learning-based approaches in turns of precision, recall and
F-score.
The LSF tolerates informal and misspelling
contents and it can easily adapt to any formats of English
writing styles.
Chou et al.,[9] applied two term weighting method to
detect internet abuse in the workplace of software
programmers. They have used six classification methods
such as Naïve Bayes, Multinomial Naïve Bayes, Back
propagation neural network, K-nearest neighbor, C4.5
decision tree, Support Vector Machine in Online news
websites such as New York Times online with several
sections such as general news, sports, entertainment,
business and technology.
They approached text
categorization to detect internet abuse in the workplace.
Androutsopouloes et al.,[10] used Naïve Bayesian
Classification, memory-based classification and total cost
ratio that allows that performance of a filter to be compared
easily to that of the Baseline inorder to filter unsolicited
bulk email. They applied both the methods to achieve very
high classification accuracy and clearly outperformed the
anti-spam keyword patterns of a widely used e-mail reader.
Their findings suggest that it is entirely feasible to construct
learning-based anti-spam filters when spam messages are
simply to be flagged or when addition mechanism are
available to inform the senders of block messages.
Javier Paraper et al.,[11] have presented automatic
methods for detecting sexual predation in Chat rooms. They
have successfully shown that a learning based method is a
feasible way to approach this problem and have proposed
innovative sets of features to derive the classification of chat
participants as predators or non-predators.
They
demonstrated that the set of features utilized and the relative
weighting of the misclassification costs in the SVMs are two
main factors that should be taken into account to optimize
performance.
They carefully analyzed the relation
importance of the classifier’s features as a preliminary effort
to understand the psycho-linguistic, contextual and
behavioural characteristics of several predators in the
internet. Their approach is promising for intelligence
gathering and prioritizations of investigative resources to
assist
police
Cyber
Crime units in their hunt for sexual predators in the Internet.
In a recent study on cyberbullying detection [12],
Electronic aggression, or cyber bullying, is a relatively new
phenomenon. In a series of two studies, exploratory and
confirmatory factor analyses (EFAs and CFAs respectively)
were used to examine
whether electronic aggression can be measured using items
similar to that used for measuring traditional bullying, and
whether adolescents respond to questions about electronic
aggression in the same way they do for tradition al bullying.
EFA and CFA results revealed that adolescents did not

differentiate between bullies, victims, and witnesses; rather,
they made distinctions among the methods used for the
aggressive In general, it appears that adolescents
differentiated themselves as individuals who participated in
specific mode of online aggression, rather than as
individuals who played a particular role in online
aggression.
A wide range of learning strategies have been adopted
for sexual predation classification in the literature.
Villatoro-Tello et al.[25] applied two-stage approach with
an initial conversation-level classification that tries to filter
out conversation with no sexual predation, and a subsequent
predator-victim classification.
The two-stage method
designed in was highly effective but the main reason behind
such high performance was a pre-processing step that
removed 90% of the conversations : a) conversations that
had only on participant were removed, b) conversations that
had less than six interventions per-user were removed, and
c) conversations that had long sequences of unrecognized
characters (apparently images) were removed.
Such
heuristic pruning was favourable for a particular
experimental setting but can most likely not be used with
other datasets.
There are also some software products available for
fighting against cyberbullying e.g., [13], [14],[15], [16],
[17]. However, filters generally work with a simple key
word search and are unable to understand the semantic
meaning of the text. While some filters block the webpage
containing the keyword, some shred the actual offensive
words themselves. Other software products exhibit a blank
page on detection of the keywords. However, removal of the
offensive word from the sentence can totally distort the
meaning and sense of the sentence. Moreover, Internet
programmers can easily dodge filters. It can be argued that
filters are not an effective anti-cyberbullying solution as
there are many ways to express inappropriate, illegal and
offensive information. Another limitation is that filtering
methods have to be set up and manually.
III.

Proposed work

Challenges in detecting Cyber bullying
Bullying is a social problem that is too large and old in
nature, there are various peoples are involved directly or
indirectly. This problem is unsolved in real life and it is just
like a hard problem. In this age of technology there are too
few places to spend time thus youngsters are involve in
internet net based social networking web sites or different
kinds of web based applications where they are able to
search new friends, persons and may be able to share
personal data over internet. Secondly the social networking
websites provide the provision for their privacy and content
management scheme, but most of the users are now much
aware about these privacy policies. Required to improve the
user interaction with these privacy policies by which the surf
internet in secure environment. The problem exist in the real
world system of bullying and their effects, we find that it is
a problematic situation of internet surfing.
1867

IJRITCC | July 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication
Volume: 2 Issue: 7

ISSN: 2321-8169
1865 – 1869

_______________________________________________________________________________________________
For a number of issues related to cyber bullying
detection, research has been done based on the text
mining paradigm such as online sexual predator recognition
[20] and spam detection [21]. Nevertheless, very little study
has been done on technical solutions, for which is why there
is insufficient proper training datasets. Moreover, privacy
issues and ambiguities can be the reasons in describing
cyber bullying.
Although bullying messages are posted everyday
comparing to hundreds of thousands of messages posted
every second, they are very sparse. Collecting enough
training data is a big challenge, since random sampling will
lead to few bully messages. One possibility is using hash
tags(#bully etc) [22] or using a set of commonly used terms
of abuse[23], however it leads to a very biased training
dataset.

In order to face these types of challenges we need to
design an effective framework that incorporates word-level
features and user based features to detect and prevent
offensive content IRC logs. We should also design the
effective strategy to detect and evaluate the level of
offensiveness of a user and word level offensiveness in a
message and we need to check whether this proposed
framework is efficient and effective enough to be deployed
on real time. In this proposed solution we provide the
primary way by which we identify the bullying, additionally
using the text and data mining technique we analyse text
content in the posts and provide the conclusion is there any
kind of bullying exist or not.
System Architecture

Word offensiveness calculation
Contextual and
word level
features

IRC logs

INPUT:User’s
conversations

Extracted
features

Words

Posts

Extracted
features

1

Word level
offensiveness test

Classification using
supervised learning
approach

Output

2

User offensiveness estimation
User
specific
Estimation
1. Match rules
language features
2. User offensiveness level

Figure 1. User offensiveness classification
We propose a contextual and word level features based
framework to detect offensive content and identify offensive
users in IRC logs. We would like to include 2 phases of
offensiveness detection. Phase 1 aims to detect the
offensiveness on the word level and phase 2 derives
offensiveness of user level. In phase 1 we need to apply the
natural language processing techniques such as word level
features and contextual level features. In phase 2 we
incorporate user-level features by using style ,structure and
cyber bullying features. The framework is illustrated in
fig.1.

IV.

Datasets

Data collection is the first step in any research project in
text mining. Data collection for the study of cybercrime
needs to focus primarily on capturing data from and social
networking sites; however, there are both legal and technical
issues that must be overcome. There is very little reliable
labeled data about predator communication; much of the
work that has appeared in both computer science and
communication studies forums is focused on anecdotal
evidence and chat log transcripts from Perverted Justice
[26]. Perverted-Justice .com began as a grass-roots effort to
1868

IJRITCC | July 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication
Volume: 2 Issue: 7

ISSN: 2321-8169
1865 – 1869

_______________________________________________________________________________________________
identify cyber predators. PJ volunteers pose as youth in chat
rooms and respond when approached by an adult seeking to
begin a sexual relationship with minor. When these
activities result in an arrest and conviction, the char log
transcripts are posted online. New chat logs continue to be
added to the web site. There were 325 transcripts,
representing arrests and convictions, on the site as of July
2009. Using these datasets, we would like to compare the
performance of different classification algorithms included
in WEKA. 1. Random Forest 2. J48 (WEKA’S C4.5
implementation) 3 Sequential Minimal Optimization.

Conclusion
Cyber Bullying is a growing problem in the social web
and it is becoming major threat to teenagers and adolescents.
In this paper we represented a survey on the current scenario
of cyberbullying and various methods available for the
detection and prevention of cyber harassment. Our concept
depends upon the text analysis, the data which is uploaded
or text written by any user is first analyzed and after that, we
estimate the roles of user, is it a bully? or a victim? As more
researchers enter this field of research should attempt

to be more proactive in addressing the role that newer
technologies, particularly cell phones are peer-to
peer devices, play in new incarnations of Cyber Crime.
References
[1]

[2]
[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

Mcghee, Jennifer Bayzick, April Kontostathis, Lynne
Edwards, Alexandra Mcbride, and Emma Jakubowski.
Learning to identify internet sexual predation.
Int. J. Electron. Commerce, 15(3):103–122, April 2011.
Nick Pendar. Toward spotting the pedophile: Telling victim
from predator in text chats. In Proc. First IEEE
International Conference on Semantic Computing, pages
235–241, 2007.
Patchin, J., & Hinduja, S. “Bullies move beyond the
schoolyard; a preliminary look at cyberbullying.” Youth
violence and juvenile justice. 4:2 (2006). 148-169.
D. Yin, B. D. Davison, Z. Xue, L. Hong, A. Kontostathis,
and L. Edwards,"Detection of Harassment on Web 2.0," In
Proceedings of the Content Analysis In The Web 2.0
(CAW2.0) Workshop at WWW2009, 2009.
K. Dinakar, R. Reichart, and H. Lieberman, "Modeling the
Detection of Textual Cyberbullying," International
Conference on Weblog and Social Media - Social Mobile
Web Workshop, Barcelona, Spain 2011, 2011.
A. Kontostathis, L. Edwards, and A. Leatherman,
"ChatCoder: Toward the Tracking and Categorization of
Internet Predators," In Proceedings of Text Mining
Workshop 2009 held in conjunction with the Ninth SIAM
International Conference on Data Mining (SDM 2009).
M. Dadvar, F. d. Jong, R. Ordelman, and D. Trieschnigg,
"Improved cyberbullying detection using gender
information," In Proceedings of the Twelfth Dutch-Belgian
Information Retrieval Workshop (DIR 2012), pp. 23-25,
February 2012.
Ying Chen, Sencunzhu, Yiluzhou, Heng Xu, “Detecting
Offensive Language in Social Media to protect Adolescent
online safety”, International conference on social
computing, p71-80,sep-2012.
C.H. Chou, Atish P. Sinha, Huimin Zhaor, “A Text mining
Approach to Internet Abuse Detection” , Proc. Of the 5th
Workshop on e-business (WeB), Milwarkee, WI 2006.
Ion Androutsopoulos, John Koutsias, Konstantinos V.
Chandrinos, and Constantine D. Spyropoulos. An
experimental comparison of naive bayesian and keywordbased anti-spam filtering with personal e-mail messages. In
Proceedings of the 23rd annual international
ACM SIGIR conference on Research and development in
information retrieval, SIGIR ’00, pages 160–167, New
York, NY, USA, 2000. ACM..
Javier Parapar, David E. Losada, A´ lvaro Barreiro,
Combining Psycho-linguistic, Content-based and Chatbased Features to Detect Predation in Chatrooms, Journal

[14]

[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]

[25]

[26]
[27]
[28]

[29]

of Universal Computer Science, vol. 20, no. 2 (2014), 213239.
Daniell M. Law A , Jennifer D. Shapkaa, Shelley Hymel A,
Brent F. Olson A, Terry Waterhouse B. The changing face
of bullying: An empirical comparison between traditional
and internet bullying and victimization, Computers In
Human Behavior, Vol 28,Issue-1,Jan 2012, Pages 226-232.
Bsecure.
Available:
http://www.safesearchkids.com/BSecure.html
Cyber
Patrol.
Available:
http://www.cyberpatrol.com/cpparentalcontrols.asp
eBlaster. Available: http://www.eblaster.com/
IamBigBrother. Available: http://www.iambigbrother.com/
Kidswatch. Available: http://www.kidswatch.com/
http://www.publico.es/espana/263683/retrato-de-unacibervictim.
http://www.foxnews.com/story/2007/11/16/mom-myspacehoax-led-todaughter-suicide/
A. Kontostathis, "ChatCoder: Toward the tracking and
categorization ofinternet predators." In:
Proceedings of SDM 2009, Sparks, NV, May 22009.[8]
P.N. Tan,F. Chen, A. Jain, "Information assurance:
Detection
of
webspam attacks in social media." Proceedings
of Army
Science Conference, Orland, Florida. 2010.
J-M.Xu, K-S.Jun, X.Zhu and A.Bellmore, Cyber bullying
detection in twitter, Vol-7, Issue-16, Pages 10-14, Reporter
2014.
H. Sanchez and S. Kumar, Twitter bullying detection, Data
mining Course Report, 2011.
www.noswearing .com/dictionary.
Esau villatoro-Tello, Antonio Juarez-Gonzalez, Hugo Jair
Escalante, Manuel Montes-y Gomez, and Luis Villasenor
Pineda. A Two-step Approach for Effective Detection of
Misbehaving Users in Chats – Notebook for PAN at CLEF
2012.
www.Perverted-justice.com

1869
IJRITCC | July 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________