研究成果

Currently, resource-constrained devices, which are known as one of the Internet of things (IoT) devices, have been widely used for healthcare systems. Most healthcare systems store users’ health data, which is encrypted by ordinary symmetric-key encryption and/or public-key encryption schemes, in a (cloud) server. However, the encrypted data needs to be decrypted for data analysis, and it means that sensitive information would be leaked to the server. One promising solution is to use fully homomorphic encryption (FHE), which enables ones to perform any computation among encrypted data while keeping it encrypted, though FHE generally requires high computational and communication costs in the theoretical sense.
In this paper, we investigate practical feasibility of FHE in resource-constrained devices for healthcare systems. First, we define a privacy-preserving protocol for healthcare systems, and implement it on PC and Raspberry Pi by using a network simulator to measure its communication overhead, computational cost, and energy consumption over wireless body area network (WBAN). For this implementation, we suppose PC and Raspberry Pi as a cloud server and a resource-constrained device such as a smartphone or a wearable device, respectively. We use two FHE libraries, HElib and SEAL, for the implementation. Our result shows that the protocol with SEAL is better than that with HElib in terms of the communication overhead and energy consumption in transmission. On the other hand, HElib is better than SEAL regarding the running time, while SEAL can perform more homomorphic operations than HElib for the almost same plaintext size. Furthermore, the energy to execute each algorithm in the libraries is very small compared to the energy required in transmission. SEAL produces smaller sizes of ciphertexts than HElib, and therefore consumes few energy consumptions. As a result, we observe that both HElib and SEAL would be used on restricted resource devices, and in particular, SEAL would be more suitable for practical use in resource-constrained devices from our analysis.