“…To protect the data owned by multiple data providers, scheme [37] presents a privacy-preserving neural network prediction model, which can implement the prediction task safely in multiple-client model. Meanwhile, a machine learning method for training word vectors with security and privacy protection is proposed by Zhang et al [38]. It is worth mentioning that both schemes [37] and [38] use the innerproduct functional encryption algorithm to train the datasets provided by multiple participants, while ensuring the security of the datasets.…”