click here to download project abstract
– ABSTRACT
Social bots are considered to be the most popular type of spamming like spamming malware
links, produce fake news, spread rumors to manipulate public opinion. Recently large-scale
social bots have been created and are wide spread on social which have a bad impact on public
and internet users’ safety in all social media platforms. Bot detection aims to distinguish bots
from humans to aid understanding the news or opinions. In recent times, classification of bots in
social media have become more as they are populated everywhere. In this paper, we propose a
decision tree classifier and a deep learning method to classify bots and humans in Twitter using
Twitter API. This proposed model uses what an account has tweeted and cross reference against
a bag of words model. These methods are unique that applies deep learning concepts to
classification. Using real world data from twitter shows the validity of the model we proposed.
Introduction
Online social networks represent a global platform through which people share and
promote products, links, opinions and news. By 2007, Twitter had almost 330 million active
users and by 2015, the number of users had grown to 1.3 billion. The data sharing feature of
social networks allows users to share links, images, videos, however this feature is commonly
used by spammers and fraudsters. Social bots are programs that automatically generate content,
share it via a particular social network, and interact with the users. A study found that social bots
in Twitter are 15% of all accounts which is equivalent to 48 million accounts, generating 35% of
content that is posted on Twitter.
Studies aimed to address the problem associated the use of automated accounts on social
networks, which can spread spam, warms, and phishing links or manipulate legitimate accounts
by hijacking and deceiving users. Malicious accounts under bot-master, who controls a group of
social bots or distribute spam or manipulate behaviors on a given social network. Social bots
have also played a significant role in the uprisings that occur in the aftermath of major events
such as elections or conflicts. The malicious activities of bots during events can be used to
spread spam. In addition, they can also cause financial harm.
The activities of social bots also impacted the social graph of Online Social Networks
(OSN’s) because the large number of non-genuine social relationships. If social bots successfully
infiltrate users accounts, they can harvest social bot private data and subsequently use It for
phishing and spamming activities. In addition, they can aggregate information from the web to
impersonate others, replicate human behaviors, and influence people by ranking and retweeting.
In addition to essential misleading users, social bots can damage the ecosystem of the social
network by establishing fake relationships and poisoning the network content.
In attempt to limit the threats posed by social bots, researches have proposed different
methods by which social bots can be detected and blocked. The majority of the studies in this
domain to date have focused on studying behavior patterns. For example, a recent study was
performed, proposed a dynamic metric to measure the change in user’s activities as a means of
identifying the strategies employed by spammers. It is important to note that not all social botcan
be classified as malicious accounts. Some even explicitly state their nature in the profile of the
account. Social bots that operate without malicious intent may serve positive purposes, such as
managing news feeds or acting as customer care responders. The problem we are concerned with
in this paper is undisclosed social bots that have malicious intentions. As outlined above, these
social bots can pose fundamental financial, social, political, and security risks. They have
become increasingly sophisticated in their designs and capabilities to avoid social bot detection
techniques. There is a requirement to gain in-depth insights into the capabilities and limitations
