1. DETECT FAKE
ACCOUNT USING
MACHINE LEARNING
TEAM MEMBERS
1.SHARAN S - 210420205050
2.VIGNESH K - 210420205031
3.MURALIDHARAN R - 210420205037
GUIDED BY : A MARIUMUTHU
ASSISTANT PROFESSOR
Project Presentation: Department of Information Technology
3. ABSTRACT
The proliferation of fake accounts on social media platforms
poses a significant challenge in maintaining online authenticity
and security. This project aims to address the issue of fake
account detection using machine learning algorithms. The
primary objective is to develop a robust model capable of
identifying fraudulent or deceptive accounts by analyzing
various user-related features and behavioral patterns.
The project involves collecting a diverse dataset comprising
both genuine and fake accounts, encompassing information
such as profile details, activity logs, network interactions, and
content shared.
Several machine learning algorithms, including but not limited to
logistic regression, decision trees, random forests, and neural
networks, will be explored and evaluated for their efficacy in
distinguishing between genuine and fake accounts. Model
performance will be assessed using metrics such as accuracy,
4. OBJECTIVES
The objective of this presentation is to discuss the importance of
fake account detection in today's digital landscape, and to
provide an overview of the various methods and tools available
for detecting and preventing fake accounts.
1. Identify and Mitigate Risks: Detect and mitigate the risks
posed by fake accounts, including misinformation,
cyberbullying, and financial fraud.
2. Preserve Platform Integrity: Enhance the integrity of online
platforms by removing fake accounts, ensuring a trustworthy
and credible online environment.
3. Protect User Privacy: Safeguard user privacy by identifying
and eliminating fake accounts engaged in malicious activities
such as identity theft and stalking.
5. LITERATURE SURVEY
S.NO NAME OF
THE AUTHOR
PAPER TITLE ISSUE
NO/YEAR
DESCRIPTION OF THE PAPER
1. Sybil Guard A Review on
Detecting Fake
Accounts in Social
Media
Vol. 154, no.
15, 2015
This paper reviews the state-of-the-art
research on detecting fake accounts in
social media. The paper discusses various
approaches, including content analysis,
network analysis, and machine learning.
2. Chakraborty,
P., Muzammel,
C.S., Khatun
Detection of Fake
Profiles in Social
Media- Literature
Review
Vol. 37, no.
23, 2012
This paper reviews the literature on fake
account detection in social media using
machine learning methods.
3. M., Islam, S.F.
and Rahman, S
Fake account
detection in social
media using
machine learning
methods
Vol. 271, no.
41, 2017
The paper discusses various machine
learning algorithms that have been used
for this task, as well as the challenges and
future research directions.
6. EXISTING WORK:
6
The existing systems use very fewer factors to decide whether an account is Fake or not.
The factors largely affect the way decision making occurs. When the number of factors is
low, the accuracy of the decision making is reduced significantly. There is an exceptional
improvement in fake account creation, which is unmatched by the software or application
used to detect the fake account. Due to the advancement in creation of fake account,
existing methods have turned obsolete. The most common algorithm used by fake
account detection Applications is the Random forest algorithm. The algorithm has few
downsides such as inefficiency to handle the categorical variables which has different
number of levels. Also, when there is an increase in the number of trees, the algorithm's
time efficiency takes a hit.
7. This system leverages the Random Forest algorithm, utilizing features
like spam commenting, interaction rate, and artificial behavior to build
multiple decision trees for identifying fake accounts. Its robustness to
missing data and ability to achieve high accuracy, even with default
hyperparameters, makes it a powerful tool for online security, significantly
outperforming existing approaches.
PROPOSED WORK:
8. To detect human fake accounts
▶ Different research steps were executed to discover deceptive
accounts. To discover such deceptive accounts, we followed the
research steps in the below figure.
Gather and Clean Data Create FictiousAccounts Validate Data
Inject FictitiousAccounts
Create New Feautures
Supervised Machine
Learning
Evaluate Results
9. CONCLUSION
▶ Human accounts and bot accounts have similar attributes and also share
same characteristics.
▶ Engineered features created to detect bot accounts can be applied to
existing corpus of human accounts but they are not as successful as they
are in detecting bot accounts
▶ The machine learning models were trained to use engineered features
without relying on behavioural data. This made it possible for these
machine learning models to be trained on very little data, compared to
when behavioural data is included.
▶ Findings indicate that engineered features used to detect fake bot
accounts at best predicted human accounts with F1 score 89.75%.
10. ▶ E. Van Der Walt and J. Eloff, "Using Machine Learning to
Detect Fake Identities: Bots vs Humans," in IEEE Access, vol.
6, pp. 6540-6549, 2018.
▶ S. Gurajala, J. S. White, B. Hudson, B. R. Voter, and J. N.
Matthews, “Profile characteristics of fake twitter accounts,”
Big Data & Society, vol. 3, no. 2, p. 2053951716674236, 2016.
▶ C. Xiao, D. M. Freeman, and T. Hwa, “Detecting clusters of
fake accounts in online social networks,” in Proceedings of
the 8th ACM Workshop on Artificial Intelligence and Security.
ACM, Conference Proceedings, pp. 91– 101.
▶ S. Mainwaring, We first: How brands and consumers use social
media to build a better world. Macmillan, 2011
REFERENCES