Identifying Behavioral Factors Leading to Differential Polarization Effects of Adversarial Botnets

Abstract

In this paper, we utilize a Twitter dataset collected between December 8, 2021 and February 18, 2022, during the lead-up to the 2022 Russian invasion of Ukraine. Our aim is to design a data processing pipeline featuring a high-accuracy Graph Convolutional Network (GCN) based political camp classifier, a botnet detection algorithm, and a robust measure of botnet effects. Our experiments reveal that while the pro-Russian botnet contributes significantly to network polarization, the pro-Ukrainian botnet contributes with moderating effects. To understand the factors leading to these different effects, we analyze the interactions between the botnets and the users, distinguishing between barrier-crossing users, who navigate across different political camps, and barrier-bound users, who remain within their own camps. We observe that the pro-Russian botnet amplifies the barrier-bound partisan users within their own camp most of the time. In contrast, the pro-Ukrainian botnet amplifies the barrier-crossing users on their own camp alongside themselves for the majority of the time.