Sajedul Talukder is a Ph.D. candidate in Computer Science at Florida International University (FIU) and currently working as a research assistant in the Cyber Security and Privacy Research (CaSPR) Lab under the supervision of Dr. Bogdan Carbunar. His research interests include security and privacy with applications in online and geosocial networks, machine learning, wireless networks, distributed systems and mobile applications. Sajedul Talukder received his B.Sc. degree in Computer Science and Engineering from Bangladesh University of Engineering and Technology (BUET), in 2014. His current research focuses on building an automated system that aims to reduce the online social networking risks for the general users and seeks to detect and defend abuses that can arise from social networking friends. His research works have been published on top-tier social networking conferences like ICWSM and ACM WebSci and have been invited by Facebook in their headquarter. During his Ph.D., he received several student travel grant awards to present his papers. At FIU, Sajedul worked as a research mentor for several undergraduate students in their summer research programs (such as Science without Borders, NSF-RET, NSF-REU). During his undergrad, he worked as a research intern in Ministry of Foreign Affairs, Bangladesh and Samsung R&D Institute Bangladesh.
Online social networks like Facebook promote friend relationships that can lead to a number of abuses ranging from collection and misuse of sensitive private user information to cyberbullying and distribution of malware, fake news, and propaganda. However, social networks often fail to aware the users of the dangers that can arise from making their information public. Many users still allow their Facebook friends to access their information, including timeline and news feed. This, coupled with the fact that people often have significantly more than 150 Facebook friends, the maximum number of meaningful friend relationships that humans can manage, suggests that Facebook users are vulnerable to attacks. Adversaries may launch a number of attacks and abuses that have already brought intense criticism and scrutiny from the users, media and politicians alike. Facebook estimated at least 13% of their users are either bots or clones who might befriend the user and infer sensitive information by stalking the data, identify “deep-seated underlying fears, concerns” by companies such as Cambridge Analytica, perform profile cloning, initiate sextorsion, steal identity, and perform spear phishing attacks, or share it with an unintended audience.
Within the scope of this dissertation, an automated online social network assistant system is proposed for detecting abuse and effectively defending against them. We observe that abuse detection and prevention, as proposed in academic work and implemented in popular online social networks, is unable to ensure that the users are protected from the perceived friend abuse. In this thesis, we propose that potentially abusive pending friends in online social networks need to be proactively detected and filtered, along with reactively detecting and preventing the already existing friends who are perceived to be abusive.
We introduce two novel approaches to prevent friend abuse in online social networks.
First, we seek to detect abusive friends in Facebook like social networks, who are already in the friend list, and to prevent the abusive friend from doing abuse by taking some protective actions against them.
Second, we propose the problem of abuse prevention during the time of friend invitation: detect the potentially abusive friend from the list of pending friend invitations, and provide the user with compelling reasons to reject the friend invitation, thus protecting the user from abuse in advance.