Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluating the Potential of In-Memory Processing to Accelerate Homomorphic Encryption

Mpoki Mwaisela, Joel Hari, Peterson Yuhala, Jämes Ménétrey, Pascal Felber, Valerio Schiavoni

TL;DR

The paper investigates in-memory processing using UPMEM DPUs to accelerate homomorphic encryption, focusing on polynomial operations that underpin FHE schemes. By integrating DPUs into two open-source libraries, OpenFHE and HElib, the authors conduct extensive experiments that reveal substantial speedups for large polynomials when data-copy overhead is ignored, but show that host–DPU data movements dominate total cost in practical scenarios. The work highlights critical takeaways: DPUs scale with workload size, yet data transfer costs often negate speedups; zero-copy data residency and language/tooling improvements are essential for practical PIM adoption in HE. Overall, PIM has clear potential for HE acceleration on large datasets, but achieving real-world benefits requires minimizing data movement and enhancing software ecosystem support. The study provides concrete architectural and software design insights that will guide future PIM-based HE research and deployment.

Abstract

The widespread adoption of cloud-based solutions introduces privacy and security concerns. Techniques such as homomorphic encryption (HE) mitigate this problem by allowing computation over encrypted data without the need for decryption. However, the high computational and memory overhead associated with the underlying cryptographic operations has hindered the practicality of HE-based solutions. While a significant amount of research has focused on reducing computational overhead by utilizing hardware accelerators like GPUs and FPGAs, there has been relatively little emphasis on addressing HE memory overhead. Processing in-memory (PIM) presents a promising solution to this problem by bringing computation closer to data, thereby reducing the overhead resulting from processor-memory data movements. In this work, we evaluate the potential of a PIM architecture from UPMEM for accelerating HE operations. Firstly, we focus on PIM-based acceleration for polynomial operations, which underpin HE algorithms. Subsequently, we conduct a case study analysis by integrating PIM into two popular and open-source HE libraries, OpenFHE and HElib. Our study concludes with key findings and takeaways gained from the practical application of HE operations using PIM, providing valuable insights for those interested in adopting this technology.

Evaluating the Potential of In-Memory Processing to Accelerate Homomorphic Encryption

TL;DR

The paper investigates in-memory processing using UPMEM DPUs to accelerate homomorphic encryption, focusing on polynomial operations that underpin FHE schemes. By integrating DPUs into two open-source libraries, OpenFHE and HElib, the authors conduct extensive experiments that reveal substantial speedups for large polynomials when data-copy overhead is ignored, but show that host–DPU data movements dominate total cost in practical scenarios. The work highlights critical takeaways: DPUs scale with workload size, yet data transfer costs often negate speedups; zero-copy data residency and language/tooling improvements are essential for practical PIM adoption in HE. Overall, PIM has clear potential for HE acceleration on large datasets, but achieving real-world benefits requires minimizing data movement and enhancing software ecosystem support. The study provides concrete architectural and software design insights that will guide future PIM-based HE research and deployment.

Abstract

The widespread adoption of cloud-based solutions introduces privacy and security concerns. Techniques such as homomorphic encryption (HE) mitigate this problem by allowing computation over encrypted data without the need for decryption. However, the high computational and memory overhead associated with the underlying cryptographic operations has hindered the practicality of HE-based solutions. While a significant amount of research has focused on reducing computational overhead by utilizing hardware accelerators like GPUs and FPGAs, there has been relatively little emphasis on addressing HE memory overhead. Processing in-memory (PIM) presents a promising solution to this problem by bringing computation closer to data, thereby reducing the overhead resulting from processor-memory data movements. In this work, we evaluate the potential of a PIM architecture from UPMEM for accelerating HE operations. Firstly, we focus on PIM-based acceleration for polynomial operations, which underpin HE algorithms. Subsequently, we conduct a case study analysis by integrating PIM into two popular and open-source HE libraries, OpenFHE and HElib. Our study concludes with key findings and takeaways gained from the practical application of HE operations using PIM, providing valuable insights for those interested in adopting this technology.

Paper Structure

This paper contains 13 sections, 9 equations, 8 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. UPMEM PIM design
  4. Homomorphic Encryption
  5. Programming with UPMEM PIM
  6. Distributing data on DPUs
  7. Adapting existing libraries for UPMEM PIM DPUs
  8. Evaluation
  9. Polynomial evaluations
  10. Case studies with existing HE libraries
  11. Discussion
  12. Related work
  13. Conclusion

Figures (8)

  • Figure 1: DPU architecture
  • Figure 2: Homomorphic encryption allows computation on encrypted data without decrypting it.
  • Figure 3: Cost of polynomial operations on CPU and DPUs.
  • Figure 4: Cost of polynomial addition on DPUs in OpenFHE with and without the host to DPU data copying vs. single and multi-threaded CPU variants.
  • Figure 5: Cost of polynomial multiplication on DPUs in OpenFHE with and without host to DPU data copying vs single and multi-threaded CPU variants.
  • ...and 3 more figures