RISKS OF DATA PRIVACY VIOLATION IN DEEP LEARNING

Abstract

Thanks to the superior predictability of deep learning methods, artificial intelligence (AI)- applied technologies solve a wide range of problems and are increasingly widely used in many fields and industries. However, deep learning-based machine learning models are good at many tasks, problems but not perfect, typically these models are very vulnerable to various attacks which violate information security criteria. In particular, the risk of data privacy breaches is an itchy issue because it not only affects the system, service providers, users but also the safety and trust of people in using these technologies, thereby seriously leading to social and legal issues. In this article, we summarize and analyze the related works of privacy violation issues in deep learning in recent years, thereby modeling and giving warnings when building deep learning models.