Detecting suspicious members in an online emotional support service

Abstract

Online emotional support systems provide free support to individuals who experience stress, anxiety, and depression by bridging individuals (i.e., users) with a crowd of voluntary paraprofessionals. While most users tend to legitimately seek mental support, others may engage maliciously by attacking volunteers with trolling, flaming, bullying, spamming, and phishing behaviors. Besides attacking the mental health of trained paraprofessionals, these suspicious activities also introduce threats against the long-term viability of the platform by discouraging new volunteers and encouraging current volunteers to leave. Towards curtailing suspicious users, we propose a novel system, namely TeaFilter, that effectively detects suspicious behaviors by integrating a collection of light-weight behavioral features together. We have performed extensive experiments based on real user data from 7 Cups, a leading online emotional support system in the world. Experimental results have demonstrated that our system can accomplish a high detection rate of 77.8% at a low false positive rate of 1%.