Scheme-Based Fully Homomorphic Encryption for Secure Logistic Regression in Privacy-Sensitive Data

Abstract

Introduction: Hospitals and research facilities are increasingly using technological methods to share patient data while protecting privacy. Distributed ledger technology and homomorphic encryption are two of these technological fixes. Data may be computed on using homomorphic encryption without ever needing to be decrypted. Even now, attackers are attacking cloud data because to the technology's intrinsic instability and fast progress. Consequently, homomorphic encryption offers a viable way to carry out viability tests on sets of private patient data kept in different places. As a result, a homomorphic encryption method relied on matrix transformations that included shifting, rotating, and transposing each letter in the raw text transformed binary ASCII value. Symmetric cryptography uses the same secret key for both encryption and decoding. A desired aspect of symmetric encryption is the "avalanche effect" which occurs when two different keys produce different cipher texts for the same communication. This effect is achieved in this manner because the key has different circumstances. A cryptanalysis of the proposed algorithm shows that it is more robust against different types of attacks than the current encryption techniques.

Objectives: Cloud computing in healthcare lowers costs while also boosting efficiency. It is quick, but the most important thing is to preserve patient-sensitive data since doing so will boost patient trust and support economic growth. The goal of digitizing patient medical records is to save expenses while increasing the effectiveness and quality of service. However, patient records include a significant amount of private information. Patients must thus be able to quickly provide a variety of medical affiliations access to their private data in a straightforward, reliable, effective, and safe manner.

Methods: homomorphic encryption is compared and examined, beginning with partially and moderately homomorphic encryption techniques and moving on to completely homomorphic techniques. Four categories were created out of completely homomorphic encryption techniques, and important methods within each category were contrasted. The usage of homomorphic encryption in the medical field is covered in the following section. In order to ensure security, homomorphic encryption algorithms were evaluated based on computation and transmission cost.

Results: We construct a computational protocol. The protocol is a series of actions that include parties sending messages to other parties and doing local calculations. Our goal is to achieve the concept of security in cryptography using a "semi-honest" approach. This security model makes the assumption that all parties will adhere to the rules and provide their real input values, but they will also be interested in learning what the other parties hidden inputs are. As long as the messages that the parties exchange while the protocol is being executed don't reveal information about the confidential inputs that each party has, the protocol is considered safe. In this context, security requires that the output of the protocol and the input known to a party be all that is needed to "simulate" the transcript of messages received by that party. According to formal specifications, this calls for a probabilistic polynomial time algorithm (the simulator), which receives input from a party, output, and a random seed that the party uses. It then produces a transcript of messages that is computationally identical to a transcript produced during a protocol run.

Conclusions: FHE has been used for statistical and machine learning applications, such as the private assessment of logistic regression. There will be more applications for FHE to find as its speed and usefulness increase.