SEMMI: Multi-party security decision-making scheme for linear functions in the internet of medical things

Abstract

In the Internet of Medical Things(IoMT), developing models using machine learning algorithms can detect and assist users effectively in identifying health issues. Due to the risk of private user information being leaked during the machine learning application process, the widespread application and development of IoMT applications are hampered. Encrypting data is a good way to protect user privacy. However, given the participants’ limited resources, processing and analyzing ciphertext data presents a significant challenge.As a result, this paper proposes a secure and efficient assisted decision-making scheme (SEMMI) that is appropriate for the IoMT applications. SEMMI performs a thorough analysis of each participant’s resource constraints, divides the data circulation process into four stages, and constructs a data circulation and ciphertext calculation protocol. Data transmission security is ensured through the use of stream encryption and homomorphic encryption. Each participant sends the ciphertext to the cloud, and the cloud calculates the ciphertext data, effectively relieving each participant’s computational load. The security of the final result is guaranteed by matching the result with pre-decryption. The scheme’s security and efficiency are demonstrated experimentally. The results indicate that the accuracy loss of each data set under the ciphertext is no more than 3% at most and that the cloud performs most of the calculations for each participant. Finally, SEMMI is applied to the FedAvg algorithm, demonstrating the scheme’s universality.