Privacy Preserving Support Vector Machines using Ring Learning with Errors

The growing concern of protecting privacy through machine learning has led to an emphasis on data sharing and analysis. Support Vector Machines (SVMs) are frequently used for classifying and predicting tasks and are renowned for their efficiency in numerous applications and their capacity to handle both linear and non-linear data. This paper presents a privacypreserving SVM training scheme, maintained through using Ring Learning with Errors (Ring- LWE). The Ring-LWE function, which is composed of lattice-based cryptography, offers a secure framework for realizing privacy-preserving machine learning. Encrypted data sets of various sizes are trained, and the difference in accuracy between encrypted and normal datasets is negligible.For that reason, during the training of models that use confidential information, BFV(Brakerski/Fan-Vercauteren) encryption method can be applied to maintain the confidentiality of the data