An Efficient and Privacy-Preserving Sensitive

Medical Data Sharing Protocol

Abstract: The aim of digital healthcare systems and learning methodology. By eliminating or doing away with the
the Internet of Medical Things (IOMT) is the ease of requirement of direct transmission of data, this framework
accessing high-quality healthcare from the comfort of one’s offers increased confidentiality, autonomous compliance to
own residence.Therefore, the widespread implementation data safeguarding laws, and enhanced scaling capability.(Li
of home-based healthcare systems is the goal of IoMT. et.al.) [7]Nonetheless, privacy problems still continue to
Millions of lives can be saved by developing intelligent and arise in FL-based medical recommender systems (FMRS),
effective systems for the early detection of serious diseases, primarily owing to the potential of leakage of confidential
which also reduces the load on established healthcare information during gradient transmission, demonstrating
institutions like hospitals.Patients and physicians now have how susceptible the model is to alterations in information
access to real-time data because to IOT advancements. provided.
Because of this breakthrough’s use of effective sensors and This project prioritises on establishing a privacy-preserving
communication technologies, digital healthcare systems are system leveraging homomorphic encryption with the objective
now less expensive and energy-intensive. This document to safeguard user security and fixate the shortcomings in
introduces a method, for handling security and privacy FMRS. Homomorphic encryption safeguards the privacy
issues related to the gathering, storage and use of data and confidentiality of sensitive genetic medical data through
in a healthcare setting. Sensitive genetic data is securely facilitating computation over encrypted data without the need
transmitted and stored using a novel protocol through for decryption. To permit authorised entities to gain access
use of homomorphic encryption (FHE). Users can upload to and utilise the data for medical research and development
their data to cloud servers where it is encrypted and purposes, our protocol intends to deal with the privacy
stored.Furthermore, the use of homomorphic encryption concerns related to the storage, distribution, and utilisation of
eliminates the need for encryption decryption cycles, guar- genetic medical data.
anteeing data security, and enables AI models to be trained This research focuses on three specific objectives:
directly upon encrypted data. Additionally, The encrypted (1) to assess the privacy and security related inefficiencies in
anonymous data may be shared for training purposes present-day FMRS frameworks, especially as they pertain to
with outside organisations. Finally, the article examines gradient transmission;
various digital innovations in the healthcare context, their (2) to propose alternative approaches to overcoming the
opportunities, limitations, and current challenges facing FMRS privacy flaws which have been found.
the e-health industry. (3) to offer a secure and reliable system which utilises
homomorphic encryption to safeguarding genetic medical
I. I NTRODUCTION data at every stage from processing to distribution and storage.
Healthcare has been through an enormous change, turning
entirely digital owing to the rapid development in medical The objective of this research is to put down a foundation for
research as well as the emergence of Internet of Medical privacy protection in FMRS by addressing the shortcomings
Things (IOMT) appliances (Joyia et.al.) [6]. This digital in existing frameworks and offering innovative remedies. Our
revolution has produced a multitude of opportunities via the research simplifies the process to securely manage, transmit,
use of machine learning (ML) and artificial intelligence (AI) and analyse genetic medical data, which in turn improves data
technology specially (Wiens et.al.) [13]. These sophisticated privacy and security in the healthcare sector. In addition, our
algorithms’ capacity of analysing enormous volumes of strategy gives patients authority over their data, facilitating
information and to provide personalised suggestions for safer and more effective machine learning model training,
medication, diagnosis, and therapy possesses shows great and eventually enhancing patient and community health out-
potential to completely transform the healthcare sector. On comes.We provide a thorough examination of each component
the other hand, there are additionally significant privacy issues as we go deeper into the historical background, current issues,
associated with collecting and archiving of massive quantities and suggested protocol design in the following sections. By
of clinical information in centralised data centres, Federated our work, we provide the foundation for a future in which
learning (FL) has been presented as a potential fix for such genetic medical data exchange may be conducted in a private
problems(Yang et.al.) [15]. Machine learning (ML) models and secure manner..
are cooperatively trained on user data using FL’s decentralized Safeguarding privacy in medical data constitute the focus of
many latest developments, an important feature of these is B. Risks and Shortcomings Analysis
homomorphic encryption, which renders it possible to compute
upon encrypted data and does away with the need for repeated Although homomorphic encryption is vital for safeguarding
encryption and decryption(Salavi et.al.) [10]. By enabling genomic data, it could also result in a computational burden
authorised entities to perform computations on encrypted, that might lower operational speed and efficiency (Vizitiu
homomorphic encryption guarantees the protection of sensitive et.al.)(Cheng et.al.) [12] [5]. Processes and approaches to
patient information throughout the analysis process (Song encryption have to be upgraded so as to mitigate these negative
et.al.)(Zhang et.al.) [11] [17]. Access control methods are impacts. For genomic information systems to prove effective
another essential piece for giving patients control over their and efficient, the data must be standardised and made interop-
medical data and for supporting homomorphic encryption. erable. Reliable and high-quality data is essential for both data
Patients can manage who has access to their encrypted medical analysis and training models; therefore, it is essential to verify
records by putting in place procedures with strong access and evaluate the quality of uploaded data so as to prevent
restrictions. This proactive participation guarantees patient disruption, but having to do this while preserving privacy is
data privacy until explicitly disclosed to authorised parties. challenging.Other risks to privacy involve the unauthorised
exposure of encrypted information and breaches of the adopted
encryption strategy. Hackers might try to intercept encrypted
data in transit or may use advanced computing techniques and
fresh innovations to try to decode the encryption; therefore,
it is imperative to secure sensitive genetic information using
robust security protocols.

III. SECURITY AND PRIVACY REQUIREMENTS

In this section, we provide the security and privacy require-

ments for the data confidentiality and patients privacy. The
main aim of this project is to protect the privacy and security
of genomic data obtained from doctors’ reports and sensors in
a manner that ensures confidentiality and guarantees individual
privacy of contributors. The most crucial requirement here is
Fig. 1. Internet of medical things for the implementation of a strong encryption method using
fully homomorphic encryption (FHE) (Bourse et.al.) [4] which
will secure all the genomic data before it gets transmitted or
II. D ETAILS OF S YSTEM AND R ISK A NALYSIS
stored on cloud servers. It is worth noting that, despite progress
A. About System Model in computing technology, this FHE system has been designed
This system aims to provide a private and secure environ- meticulously to withstand attacks from quantum computers
ment to upload, access, and share genomic data for individ- such that data being protected will remain impenetrable. An
uals. It incorporates the following components and features: authenticated channel for easy sharing of information between
Uploading user data to the system and controlling who may healthcare server (H’s server), cloud servers and medical prac-
access it are two features that users are able to make use of. titioners’ computers, which are used by medical practitioners
Only those with permission may access the data thanks to themselves, is established in order to minimize risks associated
homomorphic encryption, which protects data privacy. Until with data transmission and storage. The fortified communica-
users explicitly authorise access, the system preserves their tion channel will protect the encrypted genetic data from being
confidentiality as they upload genetic data through anonymity intercepted or accessed by unauthorized persons. Specifically,
and selective access. Users possess the liberty to designate before storing or transmitting them, genomic data are designed
access privileges, bestow certain reports, or enable read-only to be anonymous so that individuals contributing to these sets
or upload access to chosen entities. Because homomorphic can feel safe from any possible identification. Additionally, ho-
encryption allows for the training of AI models over encrypted momorphic image processing techniques (Yang et.al.) [16]are
data without requiring decryption, it eliminates repetitive en- used to manipulate encrypted genomic data without risking
cryption and decryption operations, which leads to efficient any identifiable elements such as pixel values or feature point
operations and lower computational overhead (Lin.et.al) [8]. positions. Consequently, even as it is applied for diagnostic
The trained machine learning models can ultimately be purposes or machine learning algorithms, personal privacy is
leveraged for predictive purposes, which assists individuals maintained intact while using these details. From the beginning
with limited access to medical facilities and will also reduce to the end of the diagnostic journey, which involves interaction
the workload of doctors while ensuring quality healthcare with data in a hospital and medical practitioners making final
services (Alan et.al.) [2]. diagnosis, privacy requirements are strict.
 matrix is performed. To do this box filter is convoluted with
discrete image I, exploiting the integral image ˜I. Thereafter,
both first and second order partial derivatives are computed
with much precision at different scales using box filter.
2) Feature Point Description: Surf Scheme puts much
stress on meticulous extraction of descriptors for feature
points: Each feature point is put into precisely squared region
Fig. 2. Smart healthcare around it, with its orientation at the centre. For this, the
region is split into 4×4 subregions, with computation of values
performed diligently for each one. At last, every feature point
IV. PRELIMINARIES AND RELATED WORK
has a detailed descriptor of an exact vector measuring 64
This section covers machine learning schemes, Speed up dimensions each.
Robust Features (Surf) (feature extraction from a medical
picture), and homomorphic encryption. These plans serve as C. Machine Learning Scheme
the cornerstone for building our effective MIoT system.
Next let’s consider the intricacies involved in the Machine
A. Encryption Scheme Learning Scheme (Norgeot et.al.) [9] which is using multi-
Homomorphic encryption encompasses partially homomor- image matching via density-based clustering termed multi-
phic encryption (PHE), somewhat homomorphic encryption image matching. The Quick-Match scheme is fueled by this
(SHE), and fully homomorphic encryption (FHE) schemes efficient data clustering mechanism having linear time com-
(Acar et.al.) [1]. While PHE offers limited support for either plexity as its root element.
multiplicative or additive homomorphic evaluation (e.g., RSA
or Paillier), constructing an FHE scheme requires first estab- D. Quick-Match Scheme
lishing an SHE scheme capable of supporting a defined num- : A meticulously collection and organization of feature
ber of additive and multiplicative homomorphic evaluations. points from retinal images into a tree structure Through
Comparative studies between SHE and FHE schemes reveal density-based clustering approaches, trees are dissected into
that an SHE scheme based on Rlwe offers greater efficiency clusters meticulously. Through the definition of a careful
than an FHE scheme based on Lwe. density function to estimate density correctly in order to
For instance, researchers have compared the SHE FV build the feature point tree. By using a meticulous parent
scheme with the Yet Another Somewhat Homomorphic En- function, parents for feature points from other images are
cryption (YASHE) scheme. Additionally, the SHE BGV located accurately.
scheme proposed by Brakerski, Gentry, and Vaikuntanathan
has been contrasted with the YASHE scheme, with the SHE V. S YSTEM A RCHITECTURE
FV scheme demonstrating smaller ciphertexts than the BGV
scheme. Given these considerations, we opt to utilize the SHE A. System Overview
FV scheme as our foundational encryption scheme. Under Our solution is designed to safeguard user privacy using
this scheme, denoted as E˜, where mi (i = 1, 2) represents encryption techniques and cloud services. It operates in this
plaintexts and c1 and c2 denote ciphertexts (c1 = E˜(m1 ) and manner:
c2 = E˜(m2 )), the SHE FV scheme facilitates operations such To begin the process, patients supply their data via sensors and
as addition (cAdd =[c1 +c2 ]q = E˜(m1 +m2 )) and multiplication medical reports. This concludes our period of data collecting.
(cM ult = [c1 + c2 ]q = E˜(m1 m2 )). Further details regarding The data is encrypted using FHE to safeguard its confidential-
the scheme are provided in Supplement A. ity before storing it in cloud servers. This ensures that data is
securely stored on servers to preserve privacy and anonymity.
B. Surf Scheme Patients/Users have supreme control over their data and they
The used method of Surf Scheme as invented by (Bay can control who may access, see, and add information. Users
et.al.) [3] is a broad approach to enhance the efficiency in are able to handle their data as they want because of this strict
image feature detection unlike SIFT (Scale Invariant Feature control.
Transform) algorithm (Wu et.al.) [14] How about digging To avoid repeating the encryption and decryption steps, homo-
deeper into the intricacies of Surf Scheme? morphic encryption uses the encrypted genetic data as a basis
1) Conversion from Grey Image to Integral Image: One of for training AI models this eventually improves privacy and
it’s most basic steps is converting grey image (˜I) of image security.
I into an integral image. Each pixel (x, y) belonging to pixel AI technology can provide useful information about health
grid Ω , gets carefully processed, where corresponding integral hazards to those who have access to medical services. This
image pixel gets carefully calculated by simply adding up gives people the power to assess their own wellbeing and
pixels across this region up to and including (x, y). By using decide whether to seek guidance, which eventually relieves
box filters (BF) an approximate computation of the Hessian demand on medical personnel.
B. System Design directly support division operations. We guarantee that ge-
1) Key. Management (KGM): A group of servers centrally nomic data can be processed securely while preserving privacy
oversees the creation, distribution and validation of keys. and confidentiality by converting division into additive and
Public and private keys are generated for encryption and multiplicative operations.
decryption purposes with digital certificates being provided
for keys. This process guarantees the implementation of D. Our Innovative Homomorphic SURF Approach
encryption methods. Establishes authenticated channels, for
transmitting data. 1) Integral Image Calculation: The encrypted grayscale
2) Data Acquisition and Encryption (DAE): Genomic data image is further transformed into an image using homomor-
is acquired and encrypted using the Fully Homomorphic phic operations that involve both addition and multiplication.
Encryption (FHE) scheme through sensors and interfaces used This process entails calculating the sum of intensities within
by doctors. The encrypted data is then. Securely stored on specific rectangular regions, which enhances the efficiency of
servers. feature extraction.
3) Data. Analysis (DPA): Cloud servers carry out feature 2) Homomorphic Calculation of Hessian Matrix Approx-
extraction and machine learning tasks on the encrypted data. imation: Determinant Estimation; We estimate the determi-
This involves analyzing features categorizing them while nant of the matrix through operations involving addition and
ensuring that sensitive information remains protected during multiplication. This estimation method involves approximating
processing. the determinant as a composite of order derivatives enabling
4) Diagnostic Decision Making (DDM): Diagnosticians reliable feature identification.
retrieve encrypted data, from servers decrypt it and make 3) Feature Point Detection:
diagnoses based on the decrypted information. This process
• Feature Point Filtering: We calculate the scale normalized
guarantees information confidentiality throughout the phase.
determinant of the Hessian matrix to filter out relevant
5) Result Distribution (RD): Following diagnoses comple-
feature points. To achieve this the determinant is com-
tion encrypted diagnostic outcomes are uploaded to centralized
pared with a threshold using SIMD HCS method to filter
servers which then securely distribute them to individuals.
out noise disturbances and pinpoint characteristics.
VI. F RAMEWORK , FOR P RIVACY P RESERVING S ERVICES • Selection of Feature Points: Feature points are chosen
through a 3x3 filtering approach, where every pixel point
In this section we outline the system design aimed at
is assessed against its points using SIMD HCS method.
safeguarding privacy in the collection of data from reports
Feature points that surpass a threshold are kept for
and sensors. We kick off by introducing our encoding and
analysis.
SIMD SHE techniques, which play a role in upholding data
• Refinement of Feature Points; The enhanced feature
integrity and confidentiality. Subsequently we delve into the
points are calculated through the use of operations in-
development of SIMD comparison, division, surf, clustering
tegrating the discrete gradient and Hessian matrix. This
and multi-image matching strategies to support data analysis
refinement process boosts the precision and dependability
while ensuring privacy.
of identified feature points.
A. Encoding Strategy for Genomic Data 4) Description of Feature Points:
Given the nature of information our IoT system utilizes
• Calculation of Weighted Gradient; The encrypted
an encoding strategy tailored for fixed point real numbers
weighted gradient, for each feature point is determined
to ensure precision and accuracy. This encoding process is
using SIMD SHE FV and HCS techniques. This in-
crucial for converting data into a format, with homomorphic
volves applying Gaussian filters to the encrypted image
encryption thereby protecting the privacy and integrity of
computing the gradients magnitude and direction thereby
genetic details.
facilitating a depiction of features.
B. Homomorphic Comparison Scheme for Secure Analysis • Determining Orientation; The orientation of each feature
point is established based on directions through evalu-
In the realm of genetic data examination conducting com-
ation. This stage allows for calculating the direction in
parisons on encrypted data is crucial, for detecting trends and
encrypted form for examination.
irregularities without jeopardizing privacy. Our homomorphic
comparison approach utilizes circuits to perform encrypted
comparison tasks making it possible to analyze data securely E. Machine Learning Approach
while upholding confidentiality.
1) Overview of Genomic Data Analysis: Within healthcare
C. Homomorphic Division Approach, for Safeguarded Data genomic data analysis plays a role in identifying illnesses by
Handling analyzing genetic markers, gene expression patterns and DNA
We use a homomorphic division scheme to allow safe sequences. Analyzing data enables medical treatments tailored
data manipulation because homomorphic encryption does not to individual genetic profiles.
 security of our primitives. Theorems validate the security
assertions of our schemes showcasing their resilience against
COA and KPA attacks. By employing a blend of methods
and mathematical evidence we enhance the integrity and
confidentiality of our genomic data system.

B. Homomorphic. Multi Image Clustering

The scrutiny, on security extends to our surf and multi image
clustering schemes. These schemes ensure the confidential-
ity and integrity of data during feature point detection and
multi image clustering by leveraging the security assurances
Fig. 3. Comparison on transaction latency of [3] with Proposed scheme provided by our underlying primitives. By rearranging and
mixing up the images before encrypting them our innovative
surf scheme boosts security measures ensuring that patient data
2) Utilizing Encryption: Our method utilizes encryption remains private and guarded against breaches.
methods to ensure data privacy and security during genomic
data analysis. Homomorphic encryption enables computations, TABLE I
on encrypted data without decryption maintaining confiden- AUTHENTICATION COST TEST
tiality. First we extract characteristics from genetic data sets
Authentication Fuction Computation Cost primitives
such, as gene expression levels, SNP information and DNA Hash function 5.6 ms
sequences. These characteristics are then transformed into Request encryption 12.3 ms
a format that works with encryption to allow for analysis, Transection decoding 3.2 ms
without compromising privacy. Request decryption 4.1 ms
User verification 22.5 ms
3) Advanced Clustering Methods, for Genomic Data: PBFT consensus commitment 30 ms
Complex clustering techniques like k means and hierarchi-
cal clustering are utilized to group profiles together. These
methods work directly with encrypted data allowing for the
grouping of information. VIII. O UTCOMES
4) Disease Identification Using Encrypted Data: After the
The utilization of our driven genomic data system produced
stage disease detection models such as Support Vector Ma-
results demonstrating its efficiency in upholding privacy stan-
chines (SVM) or Random Forests are trained using encrypted
dards strengthening security protocols and facilitating smooth
data. These models learn from the encrypted data attributes
data analysis processes. After testing and evaluation significant
and labels enabling disease identification while maintaining
achievements were observed in key areas;
data privacy.
5) Upholding Privacy and Security: Throughout our anal-
ysis approach we prioritize the privacy and security of infor- A. Confidentiality and Privacy Protection
mation by conducting computations on encrypted data. This Our system showcased confidentiality features that shielded
ensures the protection of details and upholds the confidentiality genomic information from unauthorized access prioritizing
of individual genetic profiles. patient privacy. By employing encryption methods such, as
SIMD SHE FV and homomorphic evaluation techniques we
VII. E VALUATION OF S YSTEM S ECURITY AND P RIVACY
effectively secured data during its transmission, storage and
In safeguarding the security and privacy of our IoT based processing stages. This encryption strategy combined with
system for data we thoroughly assess strategies, within our stringent access controls and anonymization procedures en-
threat model focusing on SIMD SHE FV, HCS, HDS, homo- sured that authorized individuals could access and analyze the
morphic surf algorithms and clustering methods. Our threat data responsibly to safeguard confidentiality.
model emphasizes the importance of maintaining confidential-
ity during transmission, storage and processing of information.
B. Security Resilience
By employing a quantum encryption scheme based on lattice
structures our SIMD SHE FV approach guarantees confiden- A comprehensive security assessment of our system high-
tiality during data transmission. lighted its defense mechanisms against threats, like curious
server scenarios (CS) known plaintext attacks (KPA) and
A. Theoretical Frameworks chosen plaintext attacks (CPA). The integration of HCS, HDS
The security claims of our schemes are supported by foun- and surf schemes bolstered protection levels significantly by
dations. For example the Ring Learning with Error Problem thwarting adversaries from jeopardizing the integrity or con-
and the BDD Problem serve as the basis for assessing the fidentiality of information.
C. Efficiency and Scalability [2] Abdullah Alanazi. Using machine learning for healthcare challenges
and opportunities. Informatics in Medicine Unlocked, 30:100924, 2022.
Our system displayed efficiency and scalability smoothly [3] Herbert Bay, Andreas Ess, Tinne Tuytelaars, and Luc Van Gool.
handling amounts of genomic data while maintaining good Speeded-up robust features (surf). Computer vision and image under-
performance levels. By using SIMD technology and homomor- standing, 110(3):346–359, 2008.
[4] Florian Bourse, Michele Minelli, Matthias Minihold, and Pascal Paillier.
phic evaluation techniques we were able to process encrypted Fast homomorphic evaluation of deep discretized neural networks. In
data in parallel improving efficiency without compromising Advances in Cryptology–CRYPTO 2018: 38th Annual International
security or privacy. Moreover the system proved its scalability Cryptology Conference, Santa Barbara, CA, USA, August 19–23, 2018,
Proceedings, Part III 38, pages 483–512. Springer, 2018.
by adapting to the increasing volume and complexity of [5] Wenzhi Cheng, Wei Ou, Xiangdong Yin, Wanqin Yan, Dingwan Liu,
datasets with impact on performance. and Chunyan Liu. A privacy-protection model for patients. Security
and Communication Networks, 2020:1–12, 2020.
D. Interoperability and Integration [6] Gulraiz J Joyia, Rao M Liaqat, Aftab Farooq, and Saad Rehman. Internet
of medical things (iomt): Applications, benefits and future challenges in
Extensive testing showed that our system is compatible healthcare domain. J. Commun., 12(4):240–247, 2017.
[7] Li Li, Yuxi Fan, Mike Tse, and Kuo-Yi Lin. A review of applications in
with existing healthcare infrastructure and various genomic federated learning. Computers & Industrial Engineering, 149:106854,
data formats and standards. It seamlessly integrates with third 2020.
party tools and platforms supporting research efforts while [8] Hui Lin, Sahil Garg, Jia Hu, Xiaoding Wang, Md Jalil Piran, and
M Shamim Hossain. Privacy-enhanced data fusion for covid-19 ap-
adhering to data privacy regulations. The systems modular plications in intelligent internet of medical things. IEEE Internet of
design and flexible APIs make it easy to incorporate into Things Journal, 8(21):15683–15693, 2020.
existing workflows empowering healthcare professionals and [9] Beau Norgeot, Benjamin S Glicksberg, and Atul J Butte. A call for
deep-learning healthcare. Nature medicine, 25(1):14–15, 2019.
researchers to utilize data. [10] Rashmi R Salavi, Mallikarjun M Math, and UP Kulkarni. A survey
of various cryptographic techniques: From traditional cryptography to
E. Overall Performance and User Satisfaction fully homomorphic encryption. In Innovations in Computer Science and
Engineering: Proceedings of the Sixth ICICSE 2018, pages 295–305.
Healthcare experts and Users commended the user design of Springer, 2019.
the system its security features and extensive privacy settings [11] Wei-Tao Song, Bin Hu, and Xiu-Feng Zhao. Optimizing lwe-based
that allowed them to securely retrieve and examine genetic fhe for better security and privacy protection of smart city. Journal of
Information Science & Engineering, 33(4), 2017.
data, for both research and medical reasons. Patients also [12] Anamaria Vizitiu, Cosmin Ioan Niă, Andrei Puiu, Constantin Suciu, and
appreciated how transparent the system was, about how it uses Lucian Mihai Itu. Applying deep neural networks over homomorphic
data and protects privacy building a sense of trust in handling encrypted medical data. Computational and mathematical methods in
medicine, 2020, 2020.
information. [13] Jenna Wiens and Erica S Shenoy. Machine learning for healthcare: on
the verge of a major shift in healthcare epidemiology. Clinical infectious
IX. CONCLUSION diseases, 66(1):149–153, 2018.
[14] Jian Wu, Zhiming Cui, Victor S Sheng, Pengpeng Zhao, Dongliang Su,
The rise of IoT has brought about changes, in fields, and Shengrong Gong. A comparative study of sift and its variants.
including healthcare by introducing new ways to manage Measurement science review, 13(3):122–131, 2013.
and analyze data. Using advanced encryption methods like [15] Qiang Yang, Yang Liu, Tianjian Chen, and Yongxin Tong. Federated
machine learning: Concept and applications. ACM Transactions on
encryption we’ve created a strong system for protecting the Intelligent Systems and Technology (TIST), 10(2):1–19, 2019.
privacy of genomic data. Our model has made strides in [16] Yang Yang, Xingxing Xiao, Xue Cai, and Weiming Zhang. A secure
keeping information secure ensuring data remains confidential and high visual-quality framework for medical images by contrast-
enhancement reversible data hiding and homomorphic encryption. IEEE
and improving the analysis of genomic data. Access, 7:96900–96911, 2019.
Our approach includes cutting edge tools such as SIMD [17] Yinghui Zhang, Xiaofeng Chen, Jin Li, Duncan S Wong, Hui Li, and
SHE FV scheme, homomorphic comparison and division Ilsun You. Ensuring attribute privacy protection and fast decryption
for outsourced data security in mobile cloud computing. Information
schemes to process data securely and efficiently. By utilizing Sciences, 379:42–61, 2017.
these methods we’ve developed privacy preserving techniques
for surfacing images and clustering images. This allows for
feature detection and image matching in datasets.
Our hard work has led to the development of a diagnosis
system for diseases that is both efficient and privacy conscious.
Through the use of image matching techniques we’ve estab-
lished secure diagnostic procedures while maintaining patient
confidentiality, with homomorphic evaluation. Additionally
our encryption method built on lattice cryptography provides
defense against quantum attacks to ensure data confidentiality
in the face of evolving security risks.

R EFERENCES
[1] Abbas Acar, Hidayet Aksu, A Selcuk Uluagac, and Mauro Conti. A sur-
vey on homomorphic encryption schemes: Theory and implementation.
ACM Computing Surveys (Csur), 51(4):1–35, 2018.