How dataflow influences FHE and ZKP acceleration

FHE and ZKP systems perform computations involving large data. Their acceleration on hardware platforms is not just about compute acceleration. This talk discusses data handling challenges and how they influence accelerator design.

The large amount of data involved in these computations presents challenges beyond improving computational speed. Efficient hardware acceleration therefore requires attention not only to how computations are performed, but also to how the associated data is handled throughout the process. These considerations are closely connected, as the requirements of data handling can affect the design and overall efficiency of an accelerator. Understanding this relationship is an important part of developing hardware platforms for FHE and ZKP systems.

Dr. Sujoy Sinha Roy is a tenured Associate Professor at Graz University of Technology (TU Graz), where he leads the CryptEng group. His research lies in cryptographic engineering, with a focus on secure and efficient implementation of post-quantum cryptography, fully homomorphic encryption, side-channel security, and hardware/software acceleration of advanced cryptographic protocols. He received his PhD degree from COSIC group of KU Leuven in 2017. His thesis received the IBM Innovation Award. He has obtained his habilitation in January 2024 from TU Graz in Applied Computer Science.