A Novel Method for Spam Call Detection Using Graph Convolutional Networks

Trung-Kien Nguyen, Viet-Trung Tran, Huy-Anh Nguyen, Khac-Hoai Nam Bui

Abstract: Unwanted calls from advertisers or phishers are a major nuisance for millions of mobile phone users. Manual blacklists are ineffective against this growing problem. A sophisticated approach is needed on the telephone operator side to detect spam calls based on malicious patterns. In this paper, we present a framework that uses network-relationship between mobile users for spam detection. Our framework has two main contributions: i) We propose an efficient and cost-effective method to construct a large, labeled spam detection graph by leveraging end-user collaboration; ii) We propose a graph neural network (GNN)-based model for the spam detection task. Our experiments show that our model outperforms strong baseline models in this research field.