GPU-Based Offloading of Curve25519 in Hyperledger Indy for IoT Environments

초록

Cloud and fog computing are increasingly used to deliver low-latency services and to process large-scale data in distributed environments. However, their decentralized and heterogeneous nature poses challenges for secure communication and trust establishment. Identity verification is crucial in addressing these issues, but traditional identity systems that rely on centralized authorities face scalability issues, privacy risks, and single points of failure. Decentralized identity frameworks such as Hyperledger Indy have emerged, offering privacypreserving authentication through decentralized identifiers and verifiable credentials. Despite its advantages, Hyperledger Indy suffers from performance bottlenecks under increasing workload. Profiling identified Curve25519 scalar multiplication as a major bottleneck in Hyperledger Indy, due to its frequent use in cryptographic operations for DID-based identity interactions. To improve scalability, we propose a graphics processing unit (GPU) offloading approach to accelerate scalar multiplication. Our implementation leverages parallel computation, shared memory, and pre-allocated GPU buffers to improve execution efficiency. Experiments show up to a 3.04 times speedup over the central processing unit (CPU) implementation, demonstrating reduced processing delay and improved scalability for decentralized identity in fog-enabled IoT environments. © 2025 IEEE.

키워드