"Hybrid Encryption Schemes for Secure Machine Learning"

Abstract

In recent years, the intersection of machine learning (ML) and cybersecurity has become increasingly critical as ML models are deployed in sensitive applications. One significant challenge in this domain is ensuring the confidentiality and integrity of ML models and data during training, inference, and deployment phases. Hybrid encryption schemes, leveraging both symmetric and asymmetric cryptographic techniques, offer a promising solution to address these security concerns effectively. This paper explores various hybrid encryption schemes tailored for secure machine learning applications. It discusses the theoretical foundations of hybrid encryption, including the roles of symmetric and asymmetric encryption algorithms in achieving confidentiality and authenticity. Furthermore, the paper examines practical implementations and optimizations of hybrid encryption schemes in the context of ML workflows.