Kuantum sonrası kriptografi: homomorfik şifreleme
Uluışık, Emrullah
This thesis presents a comprehensive analysis of Fully Homomorphic Encryption (FHE), a transformative advancement in contemporary cryptography. The exponential growth in digital data, coupled with the escalating demand for remote data processing, has elevated the importance of robust information security and privacy mechanisms. Homomorphic Encryption (HE) schemes facilitate the outsourcing of computations to untrusted third parties without exposing the underlying sensitive information, significantly enhancing data privacy assurances. Traditional encryption methods effectively protect data at rest and in transit; however, they inherently lack the capability to perform operations directly on encrypted data. FHE addresses this fundamental shortcoming by enabling intricate computations to be conducted on encrypted data without the necessity of decryption. The thesis begins by outlining the historical development of encryption systems and fundamental cryptographic concepts. It details the progression from the "privacy homomorphism" concept introduced by Rivest, Adleman, and Dertouzos in 1978 to Craig Gentry's seminal work in 2009, which presented the first practical FHE scheme. The key milestones in the evolution of FHE and major contributions in the literature are presented from both theoretical and practical perspectives. Subsequently, the mathematical structures and concepts underlying FHE are discussed in depth. Topics such as lattices, rings and fields, polynomials, number fields, canonical embedding, and ideal structures are examined as the foundational building blocks ensuring the security of FHE. In particular, challenging problems provided by Lattice-Based Cryptography, such as Learning with Errors (LWE), serve as the mathematical basis for FHE's resistance to attacks. The thesis also analyzes the structure, functioning, and differences between modern FHE schemes (such as Brakerski, Gentry and Vaikuntanathan (BGV), Fan and Vercauteren (BFV), and Cheon-Kim-Kim-Song (CKKS)). The practical applications, advantages, limitations, and recent optimization techniques of these schemes are thoroughly examined. Special attention is given to the opportunities FHE offers in sectors where data privacy is paramount, including healthcare, finance, and cloud computing. Key practical challenges such as high computational costs, large key and ciphertext sizes, and noise management are highlighted. Innovations to address these challenges, including hardware accelerators, software optimizations, and open-source libraries (like HELib and SEAL), are explored in detail. The thesis also covers contemporary research areas such as multi-key FHE, privacy-preserving computation in artificial intelligence, and post-quantum cryptography. In conclusion, this thesis presents a broad perspective on FHE, covering its mathematical foundations and practical application potential, while evaluating current developments, limitations, and future research directions. The wider and more efficient xxv adoption of FHE will open new possibilities for information security and privacy. In this context, the thesis aims to serve as a comprehensive resource for both academic and industrial communities. Keywords: Bootstrapping, Relinearization, Ring Learning with Errors, BGV, BFV, CKKS.
•Jan 1, 2025•0 citations