Transciphering: implementing Chacha20 with homomorphic encryption

• Grant number: 24/11006-4
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): September 01, 2024
• Effective date (End): August 31, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Theory of Computation
• Principal Investigator: Hilder Vitor Lima Pereira
• Grantee: Gabriela Martins Jacob
• Host Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

Fully homomorphic encryption (FHE) allows a server to compute functions on encrypted data without ever decrypting it, guaranteeing the security and privacy of the client. This technology is valuable for many practical applications, such as private artificial neural networks, and also theoretical applications, such as building new cryptographic primitives. To use FHE, the client generates the keys and encrypts their data before sending it to the server, which then computes the necessary functions and returns the result, also encrypted. The client then decrypts the result with their private key to get the final outcome, without the server ever accessing the client's data.However, despite being extremely useful, FHE has practical challenges related to computational costs, especially communication overhead. The reason is that the ciphertext tends to be over a thousand times larger than the original message, leading to a significant increase in the amount of data sent to the server. A solution to this problem is the technique called transciphering, which combines traditional encryption methods with homomorphic encryption. With this technique, the client uses a conventional cipher to encrypt their data and sends to the server the secret key of that cipher encrypted homomorphically, i.e., under FHE's public key. The server, then, homomorphically decrypts the data to compute functions on it, maintaining privacy. AES is the most widely used cipher, but despite of being standardized, it is slow when used to construct transciphering. For this reason, other ciphers have been considered and studied. This project proposes the study, implementation, and evaluation of the Chacha20 cipher with FHE to construct transciphering. Chacha20 is also widely used and well-established and, thus, can be an alternative to AES.