PrivFT: Private and Fast Text Classification With Homomorphic Encryption

The need for privacy-preserving analytics is higher than ever due to the of privacy risks and to comply with new privacy regulations leading to amplified interest in privacy-preserving techniques that try to balance privacy and utility. In this work, we present an efficient method for Classification while preserving the privacy of the content using Fully Encryption (FHE). Our system (named (PrivFT)) performs two tasks: 1) making inference of encrypted inputs using a plaintext model and 2) training an effective model using an dataset. For inference, we train a supervised model and outline a for homomorphic inference on encrypted user inputs with zero loss to accuracy. In the second part, we show how to train a model using encrypted data to generate an encrypted model. We provide a GPU of the Cheon-Kim-Kim-Song (CKKS) FHE scheme and compare it with CPU implementations to achieve 1 to 2 orders of magnitude speedup at parameter settings. We implement PrivFT in GPUs to achieve a run time inference of less than 0.66 seconds. Training on a relatively large dataset is more computationally intensive requiring 5.04 days.