Advances and Challenges in Privacy-Preserving Machine Learning

Traditional machine learning relies on collecting data in a centralized location for training algorithms, this raises privacy concerns, especially when using sensitive information like financial data or medical records. Therefore, there is a growing need to protect the privacy of training data in machine learning systems. This study adopts a systematic review approach to examine recent applications of privacy-preserving machine learning in safeguarding training data over the last decade. Our focus is on how privacy-preserving schemes have been implemented in the different phases of the machine learning process and the unique challenges of preserving privacy at these phases. Moving forward we proposed future directions highlighting concerns to be addressed to enhance efficiency and scalability in privacy-preserving machine learning. This paper provides a sterling guide to machine learning engineers and researchers as it can be used as a pointer to the unique constraints of preserving privacy at the different phases of the machine learning process and the appropriate steps to be taken to enhance the fight against privacy breaches in machine learning systems.