Wildlife Product Trading in Online Social Networks: A Case Study on Ivory-Related Product Sales Promotion Posts

Abstract

Wildlife trafficking (WLT) has emerged as a global issue, with traffickers expanding their operations from offline to online platforms, utilizing e-commerce websites and social networks to enhance their illicit trade. This paper addresses the challenge of detecting and recognizing wildlife product sales promotion behaviors in online social networks, a crucial aspect in combating these environmentally harmful activities. To counter these environmentally damaging illegal operations, in this research, we focus on wildlife product sales promotion behaviors in online social networks. Specifically, 1) A scalable dataset related to wildlife product trading is collected using a network-based approach. This dataset is labeled through a human-in-the-loop machine learning process, distinguishing positive class samples containing wildlife product selling posts and hard-negatives representing normal posts misclassified as potential WLT posts, subsequently corrected by human annotators. 2) We benchmark the machine learning results on the proposed dataset and build a practical framework that automatically identifies suspicious wildlife selling posts and accounts, sufficiently leveraging the multi-modal nature of online social networks. 3) This research delves into an in-depth analysis of trading posts, shedding light on the systematic and organized selling behaviors prevalent in the current landscape. We provide detailed insights into the nature of these behaviors, contributing valuable information for understanding and countering illegal wildlife product trading.

Figures (15)

• Figure 1: An example for wildlife product trading related post in the Online social networks. The subcaption is the text content where links and user mentions are masked. We investigate whether a post is WLT-related through its post text, images, and the linked webpages whenever necessary.
• Figure 2: Illustration for Collecting Data. Nodes are the users, and edges represent their relationships. Given seed posts as nodes in black, we fetch their following/follower (blue and yellow edges) network for several hops. Eventually, we collect all these users' timelines as candidate data for further processing. Ideally, researchers can keep expanding the dataset scale by extracting more user hops, given their budget and computation limits.
• Figure 3: A human-in-the-loop process for labeling data.
• Figure 4: Visualizing the deep learning framework for human-in-the-loop labeling process.
• Figure 5: Distribution of text length for (a) positive and (b) negative posts, as well as average token length (c) w/ and (d) w/o stop words. It is worth noting that normal posts tend to have long tails on the post length and average token length, but we had cutoffs in the visualizations to focus on the left parts. For more information on the max length, please refer to Table \ref{['tab:text_stat']}.

Theorems & Definitions (2)

• Definition 1
• Definition 2