Un modèle d’apprentissage en profondeur pour détecter L’hameçonnage des SMS (smishing)

Abstract

Smishing, a form of social engineering attack involving fraudulent SMS messages, has become a major cybersecurity issue in mobile communications. In this study, we propose a new smishing detection method based on federated learning, a decentralized learning technique that preserves privacy. Using deep learning algorithms including LSTM, BiLSTM, CNN and MLP, we build a robust smishing detection model in a federated learning framework. Experiments show that the federated learning method using CNN achieves an accuracy of 92.38 %, demonstrating the efficacy of federated learning in solving the challenges of smishing detection while preserving data confidentiality. The proposed method offers a solution to smishing attacks, and paves the way for future research into privacy-preserving mobile security.