Privacy-preserving neural networks with Homomorphic encryption: Challenges and opportunities

Abstract Classical machine learning modeling demands considerable computing power for internal calculations and training with big data in a reasonable amount of time. In recent years, clouds provide services to facilitate this process, but it introduces new security threats breaches. Modern encryption techniques ensure are considered as the best option protect stored transit from an unauthorized third-party. However, decryption process is necessary when must be processed or analyzed, falling into initial problem vulnerability. Fully Homomorphic Encryption (FHE) holy grail cryptography. It allows non-trustworthy third-party resource encrypted information without disclosing confidential data. paper, we analyze fundamental concepts FHE, practical implementations, state-of-the-art approaches, limitations, advantages, disadvantages, potential applications, development tools focusing on neural networks. FHE demonstrates remarkable progress. current literature homomorphic networks almost exclusively addressed by practitioners looking suitable implementations. still lacks comprehensive more thorough reviews. We focus privacy-preserving cryptosystems targeted at identifying solutions, open issues, challenges, opportunities, research directions.