Abstract

Tinder is a popular dating app that allows users to discover potential dating partners with close geographical proximity. Tinder is the first dating app in several countries and has more than 50 million users. However, many of these users are bots with malicious intent. The first step in dealing with this issue is understanding the characteristics of Tinder bots. Toward this aim, we have proposed a ground truth collection method to acquire bots to study. Our method combines honeypot methods and manual annotation. We find that probing messages is a reliable method to distinguish bots from humans as bots promote malicious URLs and direct users to phishing sites. Our observations on the collected bots show that they are more complex than bots that are studied in other social media sites. Tinder bots have profiles that are very hard to differentiate from normal users. We explore activity and profiles of these bots and report the characteristics that can be used in building a supervised learning approach for bot detection.

Date

2017

Authors

Tahora H Nazer, Fred Morstatter, Gareth Tyson, Huan Liu

Conference

SDM