A Study of Friend Abuse Perception in Facebook SAJEDUL TALUKDER, Edinboro University, USA BOGDAN CARBUNAR, Florida Int’l University, USA Social networks like Facebook provide functionality that can expose users to abuse perpetrated by their contacts. For instance, Facebook users can often access sensitive profle information and timeline posts of their friends, and also post abuse on the timeline and news feed of their friends. In this article we introduce AbuSnif, a system to identify Facebook friends perceived to be abusive or strangers, and protect the user by restricting the access to information for such friends. We develop a questionnaire to detect perceived strangers and friend abuse. We train supervised learning algorithms to predict questionnaire responses using features extracted from the mutual activities with Facebook friends. In our experiments, participants recruited from a crowdsourcing site agreed with 78% of the defense actions suggested by AbuSnif, without having to answer any questions about their friends. When compared to a control app, AbuSnif signifcantly increased the willingness of participants to take a defensive action against friends. AbuSnif also increased the participant self-reported willingness to reject friend invitations from strangers and abusers, their awareness of friend abuse implications and their perceived protection from friend abuse. CCS Concepts: · Security and privacy → Privacy protections; Social aspects of security and privacy; Spoofng attacks; Additional Key Words and Phrases: Social network friend abuse, friend spam, supervised detection ACM Reference Format: Sajedul Talukder and Bogdan Carbunar. 2020. A Study of Friend Abuse Perception in Facebook. ACM Trans. Soc. Comput. 1, 1, Article 1 (January 2020), 33 pages. https://doi.org/10.1145/3408040 1 INTRODUCTION Infuential social networks like Facebook encourage casual friendship relations. Social network users often have signifcantly more than 150 friends 1 , which is the number of meaningful friend relations that a person can manage [26]). Past work has shown that adversaries, including bot- operated user accounts [71] 2 , can establish friend relations with unsuspecting social network users, then expose them to vulnerabilities and abuse that include the collection and misuse of private information [24, 35, 59, 83, 84], identity theft [49] and spear phishing [28] attacks, the distribution of ofensive, misleading, false or malicious information [2, 4, 19, 74], and cyber abuse that includes cyberstalking [25], doxing [24, 59], sextorsion [84] and cyberbullying [36, 37, 56]. High-profle cases of abuse perpetrated through Facebook include Cambridge Analytica’s use of data collected from 87 million Facebook users [40] to identify łdeep-seated underlying fears, concernsž [39] and to inject content to change user perception [50] and infuence the outcome of elections [9, 10]. 1 The average number of friends per Facebook user is 338, while the median is 200 [58]. 2 Facebook estimated that 13% (i.e., 270 million) of their user accounts are either bots or clones [32]. Authors’ addresses: Sajedul Talukder, Edinboro University, USA, stalukder@edinboro.edu; Bogdan Carbunar, Florida Int’l University, USA, carbunar@gmail.com. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specifc permission and/or a fee. Request permissions from permissions@acm.org. © 2020 Association for Computing Machinery. 2469-7818/2020/1-ART1 $15.00 https://doi.org/10.1145/3408040 ACM Trans. Soc. Comput., Vol. 1, No. 1, Article 1. Publication date: January 2020.