Secure Cryptographic Techniques for Protecting Machine Learning
PDF

Keywords

Machine Learning Security
Homomorphic Encryption
Secure Multiparty Computation
Federated Learning
Differential Privacy
Model Confidentiality
Adversarial Threats

Abstract

As machine learning (ML) models become increasingly central to decision in sensitive domains such as healthcare, finance, and defense, securing these models from adversarial threats is of paramount importance. Cryptographic techniques offer a   foundational approach to safeguarding the confidentiality, integrity, and privacy of ML models and their data. This paper reviews and analyzes various cryptographic mechanisms including homomorphic encryption, secure multiparty computation, differential privacy, and federated learning that have been proposed or implemented to protect ML workflows. We compare the techniques based on their computational overhead, security guarantees, and deployment feasibility, aiming to offer a comprehensive guide for practitioners and researchers in both the ML and security communities.

PDF

All articles published in the American Journal of Cryptography and Network Security (AJCNS) are licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). This license allows users to share (copy and redistribute) and adapt (remix, transform, and build upon) the material for any purpose, including commercial use, provided appropriate credit is given to the original author(s) and the source, a link to the license is provided, and any changes made are indicated.