Cloud computing has strong computing power and huge storage space. Machine learning algorithm, combining with cloud computing, makes the processing of large-scale data practical. Logistic regression algorithm is a widely popular machine learning-based classification algorithm that can be implemented in cloud. However, data privacy cannot be guaranteed in big data processing as privacy leakage of the training data may occur. In order to prevent the privacy leakage of logistic regression algorithm in the cloud and promote the processing efficiency of training data, this paper offers a Privacy Preserving Logistic Regression Algorithm (PPLRA). The homomorphic encryption is used to encrypt the private data when they are uploaded for training. Moreover, the approximation of the Sigmoid function in logistic regression using Taylor's theorem can support the safe calculation using homomorphic encryption. The Experimental results show that PPLRA has significant effects in data privacy preserving, and is more effective in data processing. Comparison with Non-Privacy Preserving Logistic Regression Algorithm (NPPLRA) shows that the computational efficiency is improved by about 1.2 times.

云计算具有强大的计算能力和巨大的存储空间。机器学习算法与云计算相结合，使处理大规模数据变得切实可行。Logistic回归算法是一种广泛流行的基于机器学习的分类算法，可以在云中实现。但是，在大数据处理中不能保证数据隐私，因为可能会丢失训练数据的隐私。为了防止云计算中逻辑回归算法的隐私泄露，提高训练数据的处理效率，提出了一种隐私保护逻辑回归算法（PPLRA）。同态加密用于在上传私有数据进行训练时对私有数据进行加密。此外，使用泰勒（Taylor）对逻辑回归中的Sigmoid函数进行逼近 定理可以支持使用同态加密的安全计算。实验结果表明，PPLRA在保护数据隐私方面具有显着效果，并且在数据处理方面更有效。与非隐私保留逻辑回归算法（NPPLRA）的比较表明，计算效率提高了约1.2倍。