2020/01/28 by Luca Luceri, Silvia Giordano, Luceri, Luca +3
Computer Science · #Advanced Malware Detection Techniques #FOS: Computer and information sciences #Internet Traffic Analysis and Secure E-voting #Machine Learning (cs.LG) #Social and Information Networks (cs.SI) #Spam and Phishing Detection
paper · pdf · doi:10.48550/arxiv.2001.10570
openalex publication_date 2020/01/28 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Since the 2016 US Presidential election, social media abuse has been\neliciting massive concern in the academic community and beyond. Preventing and\nlimiting the malicious activity of users, such as trolls and bots, in their\nmanipulation campaigns is of paramount importance for the integrity of\ndemocracy, public health, and more. However, the automated detection of troll\naccounts is an open challenge. In this work, we propose an approach based on\nInverse Reinforcement Learning (IRL) to capture troll behavior and identify\ntroll accounts. We employ IRL to infer a set of online incentives that may\nsteer user behavior, which in turn highlights behavioral differences between\ntroll and non-troll accounts, enabling their accurate classification. As a\nstudy case, we consider the troll accounts identified by the US Congress during\nthe investigation of Russian meddling in the 2016 US Presidential election. We\nreport promising results: the IRL-based approach is able to accurately detect\ntroll accounts (AUC=89.1%). The differences in the predictive features between\nthe two classes of accounts enables a principled understanding of the\ndistinctive behaviors reflecting the incentives trolls and non-trolls respond\nto.\n