COSIC Seminar "Exploring Practical Privacy-Preserving Data Processing with..." (Yarne Thijs, COSIC)

COSIC Seminar – Exploring Practical Privacy-Preserving Data Processing with Transciphering and Fully Homomorphic Encryption – Yarne Thijs (COSIC) As the demand for secure, privacy-preserving data processing continues to grow, Fully Homomorphic Encryption (FHE) has emerged as a promising approach—allowing computation directly on encrypted data. However, its high computational cost limits real-world scalability. This talk explores two complementary strategies: pure FHE pipelines and transciphering, which combines symmetric cryptography with FHE to enable more efficient encrypted computation and transfer of data. We present a privacy preserving system, comparing both direct FHE and transciphering approaches. In particular, we introduce two algorithmic variations: One where all data is transferred and processed without being able to see the content of the data. And a second version where the use of a choice vector (CV) to control data access is used. We demonstrate how these mechanisms enhance anonymity, access control, and input/output privacy while preserving correctness. Benchmarking results across schemes (BFV/BGV/CKKS) reveal substantial trade-offs between ciphertext size, encryption latency, and computational overhead. We further simulate a FHE-friendly block cipher (e.g., Pasta) for transciphering, paving the way for real-world hybrid encryption deployments. This presentation will walk through the core algorithms, discuss their security properties, and showcase benchmarking results.