A Privacy-Preserving Centralized Hybrid Movie Recommendation System for Internet of Vehicles Mahremiyeti Koruyan, Merkezi, Hibrit Film neri Sistemi: Ara lar Arasi ?Internet i in Bir Yaklasim

Simsek M., TÜYSÜZ ERMAN A.

33rd IEEE Conference on Signal Processing and Communications Applications, SIU 2025, İstanbul, Türkiye, 25 - 28 Haziran 2025, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/siu66497.2025.11111781
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: centralized learning, hyperparameter optimization, internet-of-vehicles, privacy enhancing technologies, recommendation systems
  • İstanbul Üniversitesi Adresli: Evet

Özet

This study presents a hybrid recommendation model supported by differential privacy, aiming to enhance recommendation accuracy while preserving user data privacy. The proposed architecture integrates Matrix Factorization (MF), a Multilayer Perceptron (MLP), and a Long Short-Term Memory (LSTM) network. Differential privacy is ensured during training by injecting Laplace noise into model gradients, and hyperparameter optimization is also applied. The model is evaluated on the MovieLens 100K dataset, which contains user-item interaction data. Performance evaluation based on MSE, MAE, and NDCG metrics shows that hyperparameter optimization yields approximately a 4% improvement in MSE compared to the baseline model. In contrast, under high privacy settings, a significant degradation in accuracy is observed, with MSE increasing by approximately 39%.