Social-Guard: Detecting Scammers in Online Dating Suhaib Al-Rousan 1 , Abdullah Abuhussein 2 , Faisal Alsubaei 3 , Ozkan Kahveci 2 , Hazem Farra 2 , and Sajjan Shiva 1 1 Department of Computer Science, University of Memphis, Memphis, TN 38152, USA {slrousan, sshiva}@memphis.edu 2 Information Systems Department, St. Cloud State University, St. Cloud, MN 56301, USA {aabuhussein, okahveci, hfarra}@stcloudstate.edu 3 College of Computer Science and Engineering, University of Jeddah, Jeddah 23890, Saudi Arabia fsalsubaei@uj.edu.sa Abstract— Scamming users of online dating sites has rapidly increased in recent years. According to the Federal Trade Commission, 21,000 online dating scams were reported in 2018, with total reported losses of $143 million. A common tactic used by scammers to trick dating site users is using celebrity profile photos to increase attractiveness. In this paper, we present a prototype tool to assist online dating users in identifying scammers and interacting safely on online dating platforms. The tool recognizes celebrity photos in online dating profiles and identifies the associated profiles as possible fakes, thereby equipping online dating site users with a method to discover scammers. We created fake profiles using celebrity and non-celebrity photos to evaluate the tool’s effectiveness. The tool evaluation results were satisfactory. This work aims to increase public awareness of and alert users to signs of scams on online dating websites. Keywords—scam detection, deep learning, usable security, image recognition. I. INTRODUCTION Online social networking has not only changed how people look for, send, and receive information, but has also shaped how people interact with each other in the rapidly evolving online world [1]. Many websites offer online dating services to match users based on user profile information. An increasing number of Americans use dating sites and mobile dating applications to find love and friendship [2]. A Pew Research Center study [3] found that roughly 60 percent of U.S. adults considered Internet dating an acceptable way to start a relationship. Moreover, one in five relationships in the United States started on a dating website [4]. Users of these websites read other users’ profile biographies and decide whether to interact with them or not on the basis of this information. However, looking for romantic connections online can have noteworthy drawbacks. Unlike social platforms such as Twitter, Facebook, and LinkedIn, which allow users to search personal identifiers like name, email, or phone number, users on dating websites cannot search for other users with their personal identifiers [5]. Instead, specialized algorithms recommend matching profiles based on users’ predefined preferences. This business model thus attracts scammers who are anxious to exploit lonely hearts. According to the Federal Trade Commission (FTC) [6], 21,000 online dating scams were reported in 2018, with a total reported losses of $143 million. In one of those reported scams, a male victim reported an online dating scam, in which the scammer sent him intimate photographs and then revealed that the photographs were sent by a minor [7]. Soon after the photos were sent, the victim received a message from a man pretending to be the parent of a minor and threatening to report the victim to the police unless he sent the scammer $300 in gift cards. The existence of reciprocity on these dating websites creates a new challenge for the public in determining whether a given user profile is genuine [8]. Scammers on dating sites often: • send or use an image that looks like a model from a style magazine; • need to leave the dating site and speak with the victim via email or text; • lavish the victim with attention (e.g., texts, emails, and phone calls); • promise to meet the victim face to face, yet consistently invent a reason to avoid doing so; and • request money and promise to pay the victim back. Unlike other digital crimes, dating scams have a high success rate because they exploit basic human desires and needs. Most victims blame themselves for what happened, and rather than receiving support, they incur disapproval from relatives and others [8]. However, hiding the details about scams in dating sites does not help address the problem. While online dating services have focused on finding ways to better match users, they have not paid enough attention to protecting them. A better recommender system does not prevent scammers from creating multiple fake profiles with different information to trick recommendation algorithms. Therefore, this paper aims to help the public identify fake profiles on dating websites and improve awareness of potential scams to help users interact with others safely. The rest of this paper is organized as follows. In section 2, we present background information about this problem and discuss the motivations for our research. Section 3 discusses previous work in this field. A tool we developed to address the problem of scams on online dating sites is presented in section 4. Section 5 presents our evaluation method and its results. A discussion and future research directions are presented in section 6. Finally, we conclude the paper in section 7.