Nowadays the adoption rate of Facebook like social networking website is increasing with
significant ratio. These sites are widely used for posting posts, sharing images and group
communications. Possibilities of illegal activities are unavoidable as the post and activities are
public to everyone directly or indirectly. These illegal activities include creating fake accounts,
posting malicious posts, adult images etc. Anything posted on OSN gets viral within a short
span of time. If the post is malicious in nature it may cause a riot which would disturb the
normal working of society. Our proposed system is addressing this issue by automatically
removing malicious posts in zero hour by creating a portal which would classify the user's
post into different categories and to further analyze and recognize the malicious post using
NLP (Natural Language Processing). Sentiment analysis is done on user posts and comments
to detect user sentiments. Adult images are blocked using adult detection based on image
processing.
iii
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Chapter 1
INTRODUCTION
Web applications, especially social networks (such as Facebook, Twitter etc.) are enjoying ever
growing popularity. One of the most famous and popular social networks is Facebook with 1.17
billion monthly active users in 2016and has recently surpassed Google as the most visited site
on the Internet.[1] A multitude of examples exist which demonstrate how Facebook influences
our daily life. Even in areas of life which have always considered being private and/or intimate
is shared publically now, the usage of Facebook has become more popular .So eventually
the rise in Facebook activities is rapidly increasing day by day. For example Facebook saw
350 million users generating over 3 billion posts, comments and likes during the 32 days of
the FIFA world cup 2014. Every 60 seconds on Facebook 510 comments are posted,293000
statuses are updated and 136000 photos are uploaded. This enormous magnitude of activities
makes Facebook a lucrative venue for malicious entities to seek monetary gains and compromise
system reputation. Today Facebook, being the most preferred OSN for users to interact with
each other, group communications, to post their opinions and get news, is potentially the
most attractive platform for malicious entities to launch cyber-attacks. These cyber-attacks
include misinformation on Facebook, luring victims into scams, phishing attacks, malware
infections, malicious post etc. It has been claimed that Facebook spammers make 200 dollar
million just by posting links. Such activity not only degrades user experience but also violates
Facebook's terms of service. Lately one post can create a havoc or cause riots if group of
people find it offending. Thus the environment of the society is disturbed as well as properties
are damaged because of a single malicious post. There have been numerous real time examples
of this over the world where facebook posts caused riots. For eg In Mumbai on 21 June 2014,
riots took place in Dhule district because objectionable content about minority community
had been posted on Facebook. Another famous riot took place in Pune in June 2014 where
a controversial facebook post that contained defamatory pictures with allegedly derogatory
references to warrior King Shivaji Maharaj were posted on Facebook which caused a violent
protest and affected the place for two days.
1.1 INTRODUCTION OF THE SYSTEM
In this project, we address the problem of automatic real-time detection of malicious content
posted by user. We intend to develop a portal that classifies user posts with the help of NLP
into different categories such as Politics, Education, Entertainment and Sports etc. Classifying
the post gives probable effect of the original post, so it's easier to understand social effect of
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 1
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
any post on Society or any such social media. So our system focuses on user statuses, which
can be viewed as opinions of users or their reaction on concern we want to analyze. Texts
are extracted from posts, images to know how people feel about different posts thus sentiment
analysis is done. Sentiment analysis is applied on classified post to identify good and bad words;
the post containing maximum bad words are further automatically removed by implementing
NLP. Therefore, any user posting any malicious post which would cause disturbance in society
is automatically removed from this Social Networking Portal. Adult images are blocked using
adult image detection based on Image Processing.
1.2 BRIEF DESCRIPTION
1.2.1 Aim
Our aim is to provide a portal which would automatically remove malicious post at zero hour
from the portal and to classify them into different categories like history, sports , education ,
entertainment and politics with the help of NLP. And also to identity and remove adult images
using AID algorithm .
1.2.2 Objectives
1. To block Adult images using AID Algorithm based on Image Processing.
2. To do Sentiment analysis on user posts and comments using NLP algorithm .
3. To classify post into different categories like history, sports , education , entertainment
and politics
4. To track bad count of user and ban users who exceed bad count.
1.2.3 Motivation
In 20 minutes on an average day in Facebook:
1. 1.3 millions photos are tagged
2. 1.9 million statuses are updated
3. 2.7 million photos are uploaded
4. 2.7 million messages are sent
5. 10 million comments will be made.
This shows a huge amount of activity taking place in short span of time. This enormous
magnitude of activities makes Facebook the most attractive platform for malicious entities to
launch cyber-attacks. These cyber-attacks include misinformation, scams, phishing attacks,
cyber stalking, cyber bullying, malicious posts and etc. Often people express their views on
Facebook with statuses or comments, which can be malicious in nature; word malicious means
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 2
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
to 'cause harm'. At times it happens malicious posts/comment can hurt the sentiments of individuals,
religious groups or organization.In return it can trigger uproar in society which can
disturb the normal functionality of society.Some examples which caused riots because of malicious
posts are UK riots 2012, Iranian election protests of 2009-2010, Egyptian protests 2011
etc[14].Therefore we intend to develop a portal which would automatically remove malicious
content on zero-hour using NLP.
1.2.4 Summary of the System Functionality
The portal/system will be able to detect and remove malicious post from the portal at zero
hour posted by the user. The post can be in text or an image format. For removing malicious
post in text format NLP is used. If the post is not malicious in nature it is further classified into
various categories like politics, sports, entertainment and education using sentiment analysis.
If the post is found malicious, it will be stopped from posting. In addition, adult image
detection algorithm is used for detecting adult images; if the image crosses the set threshold
value then it is declared as adult image and will not be used for further processing.
1.3 PROJECT SCOPE
1.3.1 Overview of the Target for the Final System:
The system is mainly targeted to the users using OSN like Facebook for securing users from
malicious post and adult images .
1.3.2 Overview of the Technical Area
1. JDK:
The Java Development Kit (JDK) is an implementation of either one of the Java SE,
Java EE or Java ME platforms released by Oracle Corporation in the form of a binary
product aimed at Java developers on Solaris, Linux, Mac OS X or Windows.
The JDK includes a private JVM and a few other resources to finish the recipe to a Java
Application. Since the introduction of the Java platform, it has been by far the most
widely used Software Development Kit (SDK).
2. MySQL:
MySQL is an open-source relational database management system (RDBMS) .The SQL
acronym stands for Structured Query Language. The MySQL development project has
made its source code available under the terms of the GNU General Public License, as
well as under a variety of proprietary agreements.
1.4 APPLYING SOFTWARE ENGINEERING
APPROACH
The development model that we will be following for implementation of the software is the
waterfall model.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 3
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
1. Requirement Gathering and Planning:
This is the first stage of our system where we gathered information related to our project
by visiting social sites and go through different papers.
2. Implementation:
In implementation stage we programmed different modules .In addition, different modules
are combined together and executed as single program.
3. Verification:
In verification stage we did various testing like unit testing and system testing.In unit
testing we tested modules individually. And in system testing , we tested our system.
4. Deployment and maintenance:
This is the final stage of our system where our system is ready for use.
1.5 ORGANIZATION OF THE PROJECT REPORT
Chapter 1: Introduction
This section consists of basic information about proposed system. This chapter includes
various goals and objectives which are to be achieved by the proposed system. It helps
to focus on desired aim of the system.
Chapter 2: Background and Literature Survey
This section consists of literature survey of proposed system.
Chapter 3: Requirement and Analysis
This section consists of the problem statement which need to achieved .System requirements
specifications are defined in this section. A system requirement specification is
a structured collection of information that embodies the requirements of a system i.e
software and hardware requirement for running the proposed system. In addition, it also
contains use case diagram for the system. This section also contains the various method
used for achieving the desired aim.
Chapter 4: Design
This section includes the E-R Diagram: An entity relationship diagram shows the relationships
of entity sets stored in a database.ER diagrams illustrate the logical structure of
databases. It also includes the UML diagrams such as State diagram, Activity diagram,
Component diagram, Deployment diagram etc.
Chapter 5: Implementation
This section includes actual implementation of the system. System Architecture for
the proposed system is defined in this section. This part mainly focuses on coding for
different modules. It includes some important screen-shots of the project.
Chapter 6: Result Analysis and Evaluation
This section includes the final result and related discussion. It gives tabular representation
of test cases.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 4
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Chapter 7: Conclusion and Future scope
This section concludes the project from evaluation of the result and defines scope of the
project in future.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 5
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Chapter 2
LITERATURE SURVEY
OSN like Facebook are popular collaboration and interpersonal communication tool for millions
of users and their friends, with 500 active million users. Facebook has gained popularity
among all age group of people; it encourages users to create profiles that contain information
about themselves, like their photo, name, occupation, interests, address etc. Because of this
huge amount of information, Facebook is prone to various malicious attacks like frauds, phishing
attacks, malware infections, malicious post etc. Researchers have used various supervised
learning models to detect spam and other types of malicious content on OSNs like Facebook
and achieved positive results [10]. One of the studies on detecting malicious content is experimented
on 4.4 million public posts generated during 17 news-making events on Facebook[1]
which found substantial presence of malicious content . This study observed characteristic
differences between malicious and legitimate posts and used them to train machine learning
models for automatic detection of malicious posts. The extensive feature set was completely
derived from public information available at post creation time, and was able to detect more
number of malicious posts as compared to existing clustering based spam campaign detection
techniques. This study also deployed a real world solution in the form of a REST API and a
browser plug-in to identify malicious Facebook posts in real time. Some of the study done to
detect malicious content on Facebook and other OSN's are stated in following paragraphs.
2.1 Detection of malicious content on Facebook:
Gao et al.[2] used facebook accounts of different users to do an initial study to quantify and
characterize spam campaigns with the help of a set of automated techniques to detect and characterize
the coordinated spam campaigns .The authors observed a huge anonymized dataset
of 187 million asynchronous wall messages between various Facebook users. In return, authors
detected approximately 200,000 malicious wall posts with embedded URLs, which were originating
from more than 57,000 user accounts. Following this, Gao et al. [3] then proposed an
online spam filtering system to inspect messages generated by users in real time as a component
of the OSN platform. Rather than analyzing each post individually, this approach mainly
focused on redeveloping spam messages into campaigns for classification. Resultant was that,
using 187 million facebook wall posts as their dataset they got true positive rate of roughly
over 80 percent. In addition, authors achieved 1,580 messages/sec as average throughput and
21.5m as an average processing latency rate. However, this approach was not successful in
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 6
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
detecting any new malicious post if the system has not noted it previously as the clustering
approach used always marked a new cluster as legitimate. To protect Facebook users from
real time malicious posts, Rahman et al.[4] took advantage of the social context of posts to
deploy a social malware detection method. Using a SVM based classifier trained on 6 features;
a maximum true positive rate of 97 percent was achieved by the authors. The classifier took
46ms to classify a post. MyPageKeeper, a facebook app was developed using this model to
protect its users from malicious posts. This model also targeted at detecting spam campaigns,
and depended on message similarity features. Such techniques are efficient in detecting content,
which they have seen in the past, for example, campaigns. However, if the system is
not familiar with the post in past, then these techniques are incapable of detecting malicious
posts in real time. Nevertheless, in our propose system we overcome this flaw by using NLP to
detect malicious post at zeroth hour. Below is a summary table i.e table 2.1 of the give study
above.
Table 2.1: Summary of Literature Survey
Sr.
No.
Paper Name Objective Data/Method
Used
Result and Limitation
1 Detecting and
characterizing
social spam
campaigns
To quantify
and characterize
spam campaign
Observed a huge
anonymized dataset
of 187 million
asynchronous wall
messages between
various Facebook
users
Detected approximately
200,00
malicious wall post
with embedded
URL
2 Towards online
spam filtering in
social networks
Proposed an
online spam
filtering software
to observe messages
send by
user
Used 187 million
facebook wall posts
as their data-set
Achieved true positive
rate of 80
percentage. However
this approach
was not successful
in detecting any
new malicious post
which the system
has not recorded
previously.
3 Efficient and
scalable socware
detection
in online social
networks
To protect users
from real-time
malicious post
Used a SVM based
classifier trained on
6 features
Achieved a true
positive rate of
97 percent. Took
46ms to classify
a post. But this
approach also
had the similar
problem as stated
in the above
approach.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 7
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
2.2 Detection of malicious content on other OSNs:
Many machine-learning models have been studied in the past to detect malicious content on
other OSN such as Twitter and YouTube, [5], [6]. The parameters that affect the efficiency of
such models are age of the account, number of social connections, past messages of the user, etc.
[10], which are not available on Facebook publicly. Other techniques make use of OSN specific
features like user mentions, user replies, retweets (Twitter), views and ratings (YouTube)[9],
which are not available in Facebook. Blacklists have been shown to be ineffective, capturing
less than 20 percent URLs at zero-hour [7].
2.3 Facebook's Current Techniques:
1. For detecting malicious URLs in real time and preventing them from entering the social
graph, Facebook's immune system uses multiple URL blacklists [8]. The limitation of
the blacklist is that it is incapable in detecting URLs at zero-hour which limits the
effectiveness of this technique [7].After taking an analysis using Graph API to check if
Facebook removed any of the 11,217 malicious posts identified by blacklists after being
posted. The result was disappointing as only 3,921 out of the 11,217 (34.95 percent)
malicious posts had been deleted the remaining got past Facebook's real time filters i.e.
almost two thirds of all malicious posts (65.05 percent) and it remained undetected even
after 4 months (July – November, 2014) from the date of post.
Figure 2.1: An example of a malicious post from Facebook, this URL in the post ask users to
like a post on Facebook to earn money as indirectly its pointing to a scam website.
2. To protect its users from malicious URLs, in 2011 Facebook collaborated with Web
of Trust. This partnership states that whenever a user clicks on a link which has been
reported on WOT as malware, phishing, spam or any other kind of abuse, then Facebook
shows a warning page to the user (Figure 2). To verify this claim and to cross check the
existence and effectiveness of the warning pages , we visited some random 1000 posts
on Facebook containing a URL marked as malicious by WOT, and clicked on the URL.
Surprisingly, the warning page did not appear even once.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 8
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Figure 2.2: Example of WOT warning page which Facebook claims to show whenever a user
clicks on a link which is noted as abusive on Web of Trust.
As stated above one of the problem with Facebook is malicious post.Although a single post is
considered as user's view, it can also be malicious in nature. If a post is found offensive it can
cause a riot. Table 2.2 consist of such incidents which are caused by Facebook post.
Table 2.2: Riots caused by Facebook status across world
Sr.
No.
Description Place Consequences
1 Communal violence erupted
over'objectionable video' posted on
Facebook over Hindu God and Goddess
Chhapra, Bihar
Aug 6, 2016
1.Mosques were
damaged by petrol
and shops of Muslim
were looted
and shops were set
to fire.
2 Communal violence erupted from an
alleged objectionable Facebook post
against Prophet Muhammad
Birbhum, West
Bengal March 3,
2016
1 killed in police firing
and1police station
was ransacked
3 Defamatory post morphing photos of
Chhatrapati Shivaji, Bal Thackeray and
others on Facebook sparks violence
across the city
Pune , Maharashtra
Jun 2,
2014
24 out of 33 police
stations were
affected stones were
pelted at vehicles
and damaged 130
PMPML buses and
21 private vehicles,
as also set fire to
one bus, tempo
4 Woman and two children killed by mob
in riots over 'blasphemous' Facebook
post
Pakistan July
2014
Houses of religious
minority group
Ahmadiyyas, were
torched by a mob
5 A boy posted a morphed image of a
Hindu goddess on Facebook
Gujarat's Vadodara
The place was disturbed
for a week
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 9
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Chapter 3
REQUIREMENTS AND ANALYSIS
Requirements analysis encompasses those tasks that go into determining the needs or conditions
to meet for a new product. We determined whether the stated requirements are clear,
complete, consistent and unambiguous, and resolving any apparent conflicts.
3.1 PROBLEM STATEMENT
Automatic removal of malicious posts from Facebook using NLP.
3.2 SPECIFICATIONS OF THE SYSTEM
3.2.1 Software requirements specification
' Hardware Requirements:
1. Personal Computers/Laptops
2. RAM : 8GB
3. System : I3 processor 2.4 GHz
4. Monitor : 15 VGA Colour
' Software requirements:
1. Operating System 7/8
2. Apache Tomcat Server 7
3. JAVA(1.8)
4. MySQL
5. Eclipse Mars
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 10
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
3.2.2 System Interfaces
' Admin –
1. Admin can access the Portal posts from the developers account.
2. Admin can keep track of banned users.
3. Admin can manage all the functionalities of user account
4. Admin can unbanned users
' User –
1. User will be able to login into the system.
2. User can send friend request and also accept friend request.
3. User can start a chat with their friends
4. User can upload images and post status
5. User can like or comment post/image.
6. User can upload profile picture
' Hardware Interface:
The server is directly connected to the client systems. Also the client has the access
to the database for accessing the account details and storing the login time. The client
access to the database in the server is only read only.
' Software Interface:
Integrated system for monitoring and management is a multi-user, multi-tasking environment.
It enables the user to interact with the server and leaves a record in the
database.
' Communication Interface:
This system uses java servlets and hence requires HTTP for transmission of data.
3.3 METHODS USED
3.3.1 Natural Language Processing (NLP)
NLP is used for detecting and removing malicious post on zero hour. Following are the steps:
1. Sentiment Analysis
Sentiment analysis is a technique that determines the attitude of text. Sentiment analysis
is a type of classification. It is a concerned with determining what text is trying to convey
to a reader, usually in the form of a positive and negative attitude. Over here, we use
'sentiment analysis' to refer to the task of automatically determining feelings whether
text, is malicious in nature or not.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 11
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
2. Tokenisation
Tokenization is the process of breaking a stream of text up into words, phrases, symbols,
or other meaningful elements called tokens. The list of tokens becomes input for further
processing such as parsing or text mining. We use tokenization for breaking the user's
post or comment for further processing.Then each token is inserted into stack.
3. RSW
Sometimes, some extremely common words which would appear to be of little value in
helping select documents matching a user need are excluded from the vocabulary entirely.
These words are called stop words. Stop words are filtered out before or after processing
of data. Stop words are usually referred to the most common words in a language, there is
no single universal list of stop words used by all natural language processing tools. After
tokenization, we remove stop words from the stack which are not useful in detecting the
malicious word eg. When , the , a etc.
4. Dataset Matching
The words which are left after the removal of stop words are compared with the dataset
which contains a list of abusive/ malicious words. If the word founds a match in the
dataset then it is malicious in nature and is blocked from further processing and the
words which are not malicious in nature are categorised into various category in the
dataset
3.3.2 Adult Image Detection
1. Skin tones
The goal of skin tone detection is to build a decision rule that will differentiate between
skin and non-skin pixels.
2. Threshold
The next step for skin detection in an image is by assigning a threshold value(skin tone)
i.e. if the probability of skin tone in image is more or equal to the assigned threshold
value then that image is considered as adult and will be blocked from uploading .
3. Skin tone(Hex value)
To do the comparing of the skin tone with the assigned threshold value, it is converted
into hex value.
4. Color Model (HSV)
HSV is named for three values – Hue Saturation Value. Hue-saturation based color
spaces describes hue or tint , saturation or amount of gray in term of colors and shade
and brightness value .Hue defines color (such as red, green, yellow and purple) of an
area, saturation measures the colorfulness of an area in proportion to its brightness .The
'intensity', 'lightness' or 'value' is related to the color luminance. To do all these
processing e we are using HSV color model
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 12
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
3.4 EXPECTED OUTCOMES
If post is malicious in nature or matches with the dataset , it will be blocked at zero hour. If
image uploaded is adult image i.e. containing more skintone than threshold value then it will
be blocked.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 13
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Chapter 4
DESIGN
During the design phase we designed the system architecture of our system and also other
diagrams like ER Diagram and UML diagram.
4.1 SYSTEM ARCHITECTURE DIAGRAM
Figure 4.1: System Architecture
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 14
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Figure 4.1 gives an overview of the system. Firstly, a user has to be registered with the
portal, after that only user is able to access the system functionalities. After user is logged
on to the system successfully , he/she has various options like upload profile picture, upload
image, like post/image, add friend, accept friend request, send message etc. On the uploaded
image, Adult Image Detection Algorithm is applied. In addition, on the upload post NLP and
sentiment analysis is applied for detecting and removing malicious post at zero hour. All the
data get stored in the database. For the communication between client and server, HTTP
protocol is used.
4.2 DIAGRAMS
4.2.1 ER-Diagram
Figure 4.2: ER Diagram
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 15
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
4.2.2 UML diagrams
1. Use Case Diagram –
Figure 4.3: Use Case Diagram
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 16
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
2. Class Diagram –
Figure 4.4: Class Diagram
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 17
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
3. Collaboration Diagram –
Figure 4.5: Collaboration Diagram
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 18
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
4. Sequence Diagram –
Figure 4.6: Sequence Diagram
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 19
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
5. State Diagram –
Figure 4.7: State Diagram
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 20
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
6. Component Diagram –
Figure 4.8: Component Diagram
7. Deployment Diagram –
Figure 4.9: Deployment Diagram
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 21
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Chapter 5
IMPLEMENTATION
This chapter gives an overview of the proposed system as well as the working of given system.
This chapter also contains the screen-shots of the working models.
5.1 Overview of proposed system
The purpose of proposed system is accessing the Facebook posts from the registered user
account.User can upload Images, Messages, Comments to posts as well as make friendship and
upload profile pictures etc. User can perform two main tasks i.e uploading textual post and
image.Sentiment analysis will be done on textual post. This analysis consists of tokenization
and removal of stop words.If it encounters any illegal post it wont upload the post.Further
verified user Posts will be classified into different categories such as Politics, History, Education,
Entertainment and Sports by using Text Mining from Data-set and Online API's.Recognition
of adult image is done by applying AID (Adult Image Detection). This technique uses skintone
pixels of an image to detect if it is explicit in nature or not. If skin-tone pixel is more
than threshold value then it will be considered as adult image and wont be posted. Figure 5.1
gives an overview of the above stated working of system.
Figure 5.1: Overview of proposed system
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 22
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Below figure 5.2 show the logic of detecting malicious post i.e text posted by the user.
Figure 5.2: Logic of proposed system
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 23
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
5.2 SCREENSHOTS
1. Screen-shot for Login Page
Figure 5.3: Login Page
Figure 5.3 shows the log in page of the portal where user can login using email id or
phone number and password.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 24
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
2. Screen-shot for User Registration Page
Figure 5.4: Registration Page
Figure 5.4 shows User registration. It is the first step of the system. User registers on
the server by filling up a form which includes personal information of the user.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 25
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
3. Screen-shot for chat window
Figure 5.5: View users
Figure 5.5 shows Chat window that shows the messages exchanged between two users
and also user can send message to their respective friends.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 26
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
4. Screen-shot for Post blocking
Figure 5.6: Post blocking
Figure 5.6 shows Post blocking of the system. If the post is malicious in nature or
matches with the data set it will be blocked.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 27
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
5. Screen-shot for Adult Image Detection
Figure 5.7: Adult Image Detection
Figure 5.7 shows the Adult Image Detection. If the image posted by the user exceeds
the set threshold value that means it is explicit in nature i.e. adult image, which will be
blocked from uploading.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 28
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
6. Screen-shot for Profile Page
Figure 5.8: Profile Page
Figure 5.8 shows Profile page of system. This image shows the profile page of user where
user can update his/her information as well as view their respective time-line .
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 29
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Chapter 6
RESULT AND EVALUATION
For checking the performance of our system we used WAPT tool(Web Application Performance
Testing).WAPT is a load and stress testing solution applicable for mobile applications,
web services and all types of web sites from online stores to corporate ERP and CRM
systems.Descriptive graphs and reports will let you analyze the performance of your system
components under various load conditions, isolate and fix any bottlenecks and optimize your
software and hardware configuration.
6.1 RESULT ANALYSIS
Following is a table which shows how the system will work with different number of user.
Different parameter like average response time, sending speed etc is taken into consideration.
Table 6.1: Result Analysis table
Sr
no
Active
no. of
user
Avg response
time, sec
(with
page
elements)
Sending
speed(kbit/s)
Receiving
speed,
kbit/s
CPU utilization
Memory
utilization
(Mb)
1 1 4.23(4.38) 1.70 2742 60
444(21)
2 5 12.8(12.9) 2.22 3430 91
539(25)
3 15 17.3(17.84) 8.20 13190 92
572(27)
Avg Response time, sec (with page elements): Shows values of average response time.
The first value is the response time without page elements, and the second value (in brackets)
is the response time with page elements
Sending speed, kbit/s :Shows how many kbits per second were sent to the server.
Receiving per user speed, kbit/s :shows the receiving speed per user.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 30
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Memory utilization: Memory utilization is represented by 2 values. The first value is the
amount of used memory in megabytes (Mb). The second value (in brackets) is the percentage
of memory utilization.
Thus the conclusion is, as number of user increases response time also increase.
6.2 TESTING
Software testing is the process which allows us to test each and every module of the system by
executing various tests on the system. Software testing is very helpful in order to check that
the system is working as per the requirements.
6.2.1 Unit Testing
Table 6.2: Test Cases for unit testing
Test
ID
Test Objective
Pre-Condition Steps Test Data Expected
Result
Actual
Result
1.
Successful
User Registration
A Registration
form to be
available
Enter all
required
information
All details
of the user
User
should be
registered
successfully
and
if any field
is empty/
incorrect
it gives an
alert
The user
is registered
successfully
2. Successful
User Login
A valid User account
should be
available
Enter
the username
and
password
in login
field and
click login
button.
A valid
username
and password
User
should be
logged in
successfully
and
if any field
is empty /
incorrect it
should give
an alert.
User is
logged
in successfully
.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 31
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Test
ID
Test Objective
Pre-Condition Steps Test Data Expected
Result
Actual
Result
3.
Post
uploading:image
User should be
logged in.
Upload
an image
and click
on post
button.
A valid
image
format.
If image
is not
adult it
should be
uploaded
successfully.
and
if image is
adult then
it should
not be
uploaded
Post is
uploaded
successfully.
4.
Post
uploading:Text
User should be
logged in.
Enter
the text
and click
on post
button.
Input
should
contain
only a
text.
If text is
malicious
in nature
it should
be blocked
from
further
processing
else
it should
be successfully
posted
Post is
uploaded
successfully.
5. ClassificationUser should be
logged in.
Enter
the text
and click
on post
button.
Input
should
contain
only a
text.
If the post
is not
malicious
in nature
the post
should get
classified
in different
categories
Post is
classifying
in
different
categories
successfully.
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 32
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
6.2.2 System Testing
Table 6.3: Test Cases for unit testing
Test
ID
Functionality
to tested
Test Procedure
Expected Result
Actual
Result
Pass/Fail
1
To verify
whether JDK is
installed
Java environment
checked
Java should be
installed
JDK installed
pass
2
To verify the
existence of
Eclipse IDE
Eclipse IDE
environment
checked
Eclipse should
be installed
Eclipse is
installed
Pass
3
To verify existence
of Apache
Tomcat server
To check the
installation of
Tomcat
Tomcat should
be installed
Tomcat is
installed
Pass
4
To verify installation
of MySQL
Workbench
Check the
MySQL environment
MySQL should
be installed
MySQL is
installed
Pass
AISSMS IOIT-INFORMATION TECHNOLOGY 2016-17 33
Facebook Watchdog : Automated Removal of Malicious Post from Facebook
Chapter 7
CONCLUSION AND FUTURE
WORK
7.1 CONCLUSION
Over the years many shortcomings on OSN like Facebook has be identified; one of them is to
identify and remove malicious content from Facebook especially which is posted by the user.
Many approaches have been used to identify and remove malicious content posted by user on
zero hour. None of them is successful in real time detection. Facebook is a social networking
site used for giving voice to ones thoughts however; a certain post can cause a riot or can hurt
the sentiment of the others. Our system overcomes this disadvantage by using NLP where it
protects it's user from malicious content using NLP. Our system has achieved the objective of
real time detection of malicious content and to further divide the post into various categories
like politics, sports, education, entertainment etc. The proposed system will also detect adult
images using adult image detection algorithm and will maintain bad count for every user. If
the bad count exceeds the set threshold value, the user will be blocked from the portal.
7.2 FUTURE WORK
1. We can enhance the system by adding facial recognition to precisely detect an adult
image as it classifies face as an adult image because of skin-tone
2. We can apply adult image detection technique on user's profile picture to ensure that
even profile picture is not an adult image
3. We can enhance our algorithm to detect a sarcastic post as now a days sarcasm is too
common.
4. We can add regional language to our system to attract user who don't know English.
...(download the rest of the essay above)