Robust and Secure Multiparty Collaborative QoS Prediction Scheme for IIoT Services

The rapid growth of intelligent manufacturing within the Industrial Internet of Things (IIoT) has highlighted the importance of quality of service (QoS) as a key metric for system performance and reliability. Due to its dynamic and heterogeneous nature, accurate QoS prediction is essential but remains difficult under data silos, privacy constraints, and limited trust among industrial entities. To overcome these limitations, this article proposes a robust and secure multiparty collaborative (RSMC) QoS prediction scheme for IIoT service recommendation. RSMC leverages a federated learning architecture with a secure aggregation strategy, facilitating joint model training across factories without exposing raw data. Specifically, a hybrid local time-aware prediction model integrating a temporal convolutional network with a bidirectional simple recurrent unit is employed to capture long-term dependencies and nonlinear dynamics in QoS time series. To ensure the security of collaborative training, a robust functional encryption mechanism is introduced to achieve secure parameter aggregation and effectively resist inference and collusion attacks. Experiments on real QoS datasets show that the proposed scheme achieves superior prediction accuracy compared with mainstream approaches, while preserving privacy and enabling large-scale collaboration.