Work

SMS spam detection using BERT and multi-graph convolutional networks

Year: 2025

Type: article

Abstract: The surge in smartphone usage has significantly increased SMS traffic and, consequently, SMS spam, posing risks such as phishing, financial losses, and privacy breaches. Traditional rule-based and blacklist methods fail against evolving spamming techniques, prompting the adoption of machine learning and deep learning approaches. However, models like CNNs and RNNs struggle to capture global co-occurrence patterns and complex semantics, while transformer-based models like BERT lack explicit syntactic and co-occurrence modeling. To address these limitations, we propose the BERT with Triple-Graph Convolutional Networks (BERT-G3CN) model, the first framework to integrate BERT word embeddings with graph embeddings from Cooccurrence, Heterogeneous, and Integrated Syntactic Graphs. This multi-graph approach captures diverse features and models both global and local structures using tailored Graph Convolutional Networks (GCNs). Experiments on two benchmark datasets demonstrate that BERT-G3CN achieves superior accuracy of 99.28% and 93.78%.

Source: International Journal of Intelligent Networks

DOI: 10.1016/j.ijin.2025.06.002

Authors: Linjie Shen, Yanbin Wang, Zhao Li, Wenrui Ma

Institutions: Zhejiang Gongshang University, Shenzhen MSU-BIT University, Zhejiang Lab

Language: English

Topic: Spam and Phishing Detection

Subfield: Information Systems

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $700

Funder: Key Research and Development Program of Zhejiang Province