International Journal of Data Science and Artificial Intelligence (IJDSAI)

Volume 02, Issue 03, May – June (2024)

RESEARCH ARTICLE

IOT-CENTRIC DATA PROTECTION USING

DEEP LEARNING TECHNIQUE FOR
PRESERVING SECURITY AND PRIVACY IN
CLOUD
C. Senthil Singh 1, *, Sameena Naaz 2 and G. Saranya 3
1
Department of Electronics and Communication Engineering, Shadan Women College of Engineering and Technology,
Hyderabad, India.
2
Department of Computer Science, Roehampton University, London, SW15 5PH, United Kingdom.
3
Department of computer science and Engineering, SA Engineering college, Poonamallee, Tamil Nadu 600077 India.
*Corresponding e-mail: senthilsingh@gmail.com

Abstract – The Internet of Things (IoT) describes a system data to the cloud results in the gadget losing ownership of it
where interconnected physical objects are connected online. As [5].
the collection and sharing of vast amounts of personal data
grow, so do concerns over user privacy within IoT An average data loss expense worldwide will be $4.35
environments. While IoT devices offer significant advantages in million in 2022, up to 2.6% and 12.7% from 2021 and 2020,
terms of productivity, accuracy, and financial benefits by respectively, according to research by IBM and Ponemon
minimizing human intervention and providing exceptional Institute [6]. These factors have made data protection a major
flexibility and convenience, they also face challenges related to problem, which has inspired researchers to provide strategies
communication overhead, security, and privacy. To address
for maintaining data privacy [7]. Most of the models that are
these issues, a novel Internet of Things-based Cloud
Information Security Preservation (IoT-CISP) has been
now in use are based on encryption techniques [8], and
proposed. This approach enhances the model’s effectiveness differential privacy, but they have a limit According to
and ensures security by first separating sensitive data from non- research by IBM and the Ponemon Institute, the average cost
sensitive data using an SVM classifier, and then employing this of a data breach globally will be $4.35 million in 2022, an
data for partial decryption and analysis. Sensitive data is increase of 2.6% and 12.7% from 2021 and 2020,
protected through Okamoto-Uchiyama encryption, ensuring respectively.
that data storage, analysis, and sharing are conducted securely
to maintain the system’s safety and privacy. The effectiveness These reasons have made data protection a significant
of this novel method was assessed against existing issue, prompting researchers to offer techniques for
methodologies using parameters like precision, accuracy, F1 preserving data privacy. The majority of models now in use
score, and recall, revealing its superior security and efficiency are based on encryption methods as well as differential
compared to other schemes. Results demonstrate that the IoT- privacy, although their accuracy, utility, and efficiency might
CISP approach offers encryption times that are 31.24%, be improved. As far as the authors are aware, there are
23.12%, and 33.03% shorter than those of the CP-ABE, GDBR,
currently no models that adequately balance the accuracy and
and HP-CPABE algorithms, respectively.
privacy of the data that is being outsourced [9,10]. To
Keywords – cloud computing, Internet of things, Support vector overcome these challenges Internet of Things based Cloud
machine, Okamoto Uchiyama. Information Security preservation (IoT-CISP) has been
1. INTRODUCTION proposed. The following are the paper's main contributions:

IoT devices, generate enormous amounts of data but lack • The proposed IoT-CISP approach enhances the
the storage and the processing capability to process it, which model’s effectiveness and ensures security by first
are the foundation of the cyber-physical system [1, 2]. The separating sensitive data from non-sensitive data
cloud platform has established itself as a top-tier method of using an SVM classifier, and then employing this
storing, analyzing, and exchanging data with numerous data for partial decryption and analysis.
stakeholders for the best results [3, 4]. However, it is not • Sensitive data is protected through Okamoto-
advisable to put your faith in a third-party cloud platform, Uchiyama encryption, ensuring that data storage,
particularly for preserving private information, as renting analysis, and sharing are conducted securely to
maintain the system’s safety and privacy.

ISSN: 2584-1041 ©KITS PRESS Publications

 C. Senthil Singh et al. / IJDSAI, 02(03), 81-87, 2024

• The effectiveness of this novel method was assessed in the context of GDPR compliance. Additionally, discussed
against existing methodologies using parameters the challenges and identified gaps, offering suggestions for a
like precision, accuracy, F1 score, and recall, new modeling technique. This technique not only models’
revealing its superior security and efficiency conventional risks to data privacy but also efficiently does
compared to other schemes. GDPR compliance checks.
The remaining sections are arranged as follows: Section In 2023 Huang, B. et al [16] Proposed Ciphertext-Policy
3 presents the proposed framework. The experimental setting Attribute-Based Encryption (CP-ABE) as a lattice-based
is demonstrated in Section 4. The collected results are then encryption technique. Compared to techniques built on the
detailed in Section 5. With a conclusion and future work, the Learning with Error problem, this particular approach is
paper is completed. more efficient because it is based on the Ring Learning By
Error issue. Due to unreliable cloud service providers, the
2. LITERATURE SURVEY approach addresses security and access control concerns for
In 2021, Zhang et al. [11] proposed the HP-CP-ABE critical data. Evaluations and experimental simulations
hidden access policy to safeguard data security and demonstrate the scheme's excellent applicability and
authenticate authorized users. One potential issue is the efficiency.
possibility of attackers launching attribute values to reveal. 3. PROPOSED METHODOLOGY
The parameter data is included in several HP-CP-ABE
methods' access controls, employing guessing attacks A malicious utility provider might take data that has
(AVGA). To address the computational PBDHE assumption been outsourced from the cloud, store it, analyze it, and share
as a means to establish the selective IND-AVGA security of it with the parties involved to gather private information that
the proposed scheme, this is the first time this assumption has could be abused. As a result, protecting data has grown to be
been applied. a difficult undertaking that must be handled carefully. This
paper provides a safe data protection technique for
In 2023 Vaidya S [12] Suggested a hidden approach preserving confidentiality in a cloud context to address this
framework paired with a based encrypted approach to crucial and difficult problem. It does so by effectively
improve security associated with healthcare IoT information. separating the data into sensitive and non-sensitive
The system ensures precise control over access to encrypted categories using an SVM classifier, partially decrypting the
data and safeguards the privacy of clinical clients. The data, and performing data analysis that increases the model's
proposed system far surpasses existing systems in terms of, effectiveness while maintaining security. Okamoto
security, storage load, and computing efficiency, especially Uchiyama encryption has been used to protect sensitive data.
when the approach structure is concealed. It presents a new By carrying out safe data storage, analysis, and exchange, the
approach for strongly transmitting data in the context of IoT. model guarantees the security and privacy of the system.
In 2023 Wang. C et al. [13] Proposed a robust data Specific criteria, including precision, accuracy, F1-score, and
encryption scheme called Attribute Hiding and Multiple recall, have been used to compare the suggested method to
Authorization Centers-based Data Hierarchical Encryption the existing methodologies. Figure 1 shows the overall
Scheme (AH-MAC-DHE). This strategy Maintains private framework for the suggested work.
data by disguised access controls and user features. to handle 3.1 Support Vector Machine
the problem of private data being leaked. Assuming
judgmental q-parallel the coefficient of the Bilinear Diffie- Support vector machines (SVMs) are used to collect data
Hellman, they have shown that AH-MAC-DHE is reliable from IoT devices, which is divided into two categories:
and offers security for privacy as well as anti-collusion. sensitive and nonsensitive data. SVMs are a type of
According to experimental findings, AH-MAC-DHE supervised machine learning technology that converts
performs better than current methods. complex, highly nonlinear circumstances into binary
classification models [5]. The SVM must construct the
In 2022 Li, M., Xiao, D., et al. [14] proposed based on decision surface, which is a hyperplane, utilizing data
compressive sensing, utilizing private cloud for three samples to maximize the margin around it. During the
different levels of cloud service users. This enables the training stage of the algorithm, each data sample is assigned
sensor-cloud system to offer a variety of multimedia service a class designation and the projected value. The data sample
levels and security assurances. From the standpoint of contains what are referred to called characteristics, these are
consumers of cloud-based services, ensuring the privacy of the variables in the data that specify the data sampling
essential data is a challenge. The suggested approach vector's activity. The weights applied to each input feature
successfully balances the relationships between cloud service and a collection of support vectors that construct the ideal
providers, sensor network suppliers, and cloud service hyperplane are used to forecast the results of the SVM
customers, according to theoretical studies and experimental training phase. In contrast to other neural networks, the SVM
simulations. maximizes the number of nonzero weights while lowering
In 2022 Wei et al. [15], proposed Scourge modeling the overall number of nonzero weights by maximizing the
techniques for addressing privacy of data threats in margin. These only match the important traits that provide
independent systems, and we have analyzed these techniques information that is useful for selecting the hyperplane.

82
 C. Senthil Singh et al. / IJDSAI, 02(03), 81-87, 2024

Figure 1. Block diagram of the proposed method

The kernel function modifies the data dimensions to
define the hyperplane's form, which is a crucial step in the
SVM. Simply put, the kernel function increases the
hyperplane's size to help distinguish between the classes. The
employment of many kernel types is possible, including the
polynomial, linear, both sigmoid and the Gaussian radial
basis function. The type of data sample affects how each
kernel performs. The simplest kernel, the linear kernel,
performs better when applied to linear problems. The
supplied characteristics are combined by the polynomial,
RBF, and sigmoid kernels to produce support vectors. They
work best with non-linear data, but their complexity depends
on how many additional features they find. In Figure 2, the
SVM flowchart is displayed.
3.2 Okamoto Uchiyama Encryption
Numerous concepts from number theory, discrete
mathematics, and abstract algebra are used in the Okamoto-
Uchiyama cryptosystem. Numerous of these ideas are
fundamental and are applied in various areas of
cryptography. However, even though beyond integer
computations, there is no need to provide detailed or rigorous
treatments, those essential notions are still important. They
are not sufficiently covered in mathematics curricula in
underdeveloped and developing countries, by examining the
fundamental concepts and mathematics used in the
Okamoto-Uchiyama algorithm. Figure 2. Flow chart of Support Vector Machine (SVM)

83
 C. Senthil Singh et al. / IJDSAI, 02(03), 81-87, 2024

3.2.1 Key Generation Algorithm 1: PHE Cryptosystem

The following procedure generates a public/private key: Function Keygen ()
1. Create the two big primes, A and B. Output: Public key 𝐾𝑃 :(N, G), private key:(⋌, 𝜇)
2. Compute N=𝐵2 A
3. Select an integer with a random value G2 n-1 such Choose two large prime numbers P and Q of equal bit-
that G(B-1) length and compute N=P.Q;
4. Computes H=GN mod N. G←N+1;
Next, we have (N, G, H) as the public key and (B, A) as the ⋌=∅(𝑁) = (𝑃 − 1). (𝑄 − 1), where ∅(∙) denotes the
private key. Euler’s totient function;
3.2.2 Encryption 𝜇 = ∅−1 (N)mod N, which is the modular
Using the public key (B, A), the following can be done multiplicative inverse of ∅(𝑁);
to encrypt a message M < B.
Function ℇ(𝐿)
1. A random number R1n-1 should be chosen so that
Output: Ciphertext c
G(B-1)
2. Compute C= GN HR mod N. the random 𝑟∈
The value C is the encryption of M ∅∗ −1 ={𝑧|𝑧 ∈ 𝑍, 0 ≤ 𝑧 < 𝑁, 𝑔𝑐𝑑(𝑧, 𝑛) = 1};

3.2.3 Decryption element the ciphertext by c=𝐺 𝐿 .𝑟 𝑁 mod 𝑁 2 ,where m∈

𝑍𝑁 = {z|z ∈ 𝑍, 0 ≤ 𝑧 < 𝑁}, 𝑐 ∈ 𝑍𝑁 ∗ 2;
An encrypted message C can be decrypted with the
private key (B, A) as follows: Function D(C)

1. Compute P= l (𝐶 𝐵−1 mod 𝐵2 Output: Message L

2. Compute Q= l (𝐺 𝐵−1 mod 𝐵2 p and Q will be 𝜇−1
Define the integer division function M (𝜇)= ;
integers. 𝑁
3. Using the extended Euclidean technique, compute Generate the plaintext as L=m (𝑐 ⋌ 𝑚𝑜𝑑 𝑁 2 ). 𝑚𝑜𝑑N;
the opposite value of Q modulo B.
𝑄′ = 𝑄−1 𝑚𝑜𝑑 𝐵
4. Compute M=𝑃𝑄 ′ 𝑚𝑜𝑑 𝐵 Following data entry into the BiGRU network, the
The value M is the decryption of C. processed data undergoes full learning, which enhances the
accuracy of temporal feature extraction even further. The
3.3 Partially Homomorphic Encryption Using Sensitive completely connected layer ultimately produces the final
Data forecasting results.
Schemes for partial homomorphic encryption (PHE) 3.4.1 Convolutional Neural networks
provide the performance of certain accurate processes on
scrambled data. The PHE cryptosystem uses public keys. It The deep design and integration of convolution
comprises three stages: encryption, decryption, and key processing define convolutional neural networks (CNNs), a
creation. Specifically, since it possesses the public key and subclass of feed-forward neural networks. They usually deal
secret key throughout the phase of generating keys, every with spatial data loss, inefficiencies, and overfitting.
device receives the public key, which is used for encrypted Combining both convolutional and pooling layers in the
data. The system operator or just one agent has access to the CNN model framework enables efficient feature-learning
private key needed to decrypt messages. Specifically, there and classification activities by automatically obtaining
is a time variation in both the public and private keys. features at different scales and levels.
𝐸𝑛𝑐(𝐿1 )°𝐸𝑛𝑐(𝐿2 ) = 𝐸𝑛𝑐(𝐿1 + 𝐿2 ) (1) 3.4.2 Gated Recruitment Unit (GRU)
In particular the ciphertext of 𝐿1 + 𝐿2 ) can be obtained from One model that processes sequence data using the deep
the ciphertexts of 𝐿1 and 𝐿2 As seen in Algorithm 1, the learning technique is called BIGRU. It is predicated on
three roles execute all three PHE encryption system features. enhancing GRU with the addition of a bidirectional loop
structure, which improves the collection of contextual
3.4 BiGRU-CNN information in sequence data. Two gated loop units are
Figure 3 shows the process of entering current load data into included in every BiGRU unit; one is used to process
the CNN network's computational layer from the data source sequence data in the direction that is forward, and the other
layer. The convolutional layer is used to record data is used to process data in the opposite direction.
correlation. By applying pooling operations to reduce data
dimensionality, the pooling layer improves the efficiency of
network learning.

84
 C. Senthil Singh et al. / IJDSAI, 02(03), 81-87, 2024

Figure 3. Bi-GRU CNN Neural networks

It lowers the model's training complexity by combining
an LSTM's input and remembering gates into just one
updating gate complexity and convergence time, as well as
its number of parameters, and achieving faster training
convergence.
𝑟𝑡 = 𝜎(𝑊𝑟 𝑥𝑡 + 𝑈𝑟 ℎ𝑡−1 ) (1)
𝑧𝑡 = 𝜎(𝑊𝑟 𝑥𝑡 + 𝑈𝑟 ℎ𝑡−1 ) (2)
ℎ̌𝑡 = tanh (𝑊𝑥𝑡 + 𝑈(𝒓𝒕 ⊙ ℎ𝑡−1 )) (3)
ℎ𝑡 = (1 − 𝑧𝑡 )ℎ𝑡−1 + 𝑧𝑡 ℎ̌𝑡 (4)
Here, 𝑊𝑡, 𝑊𝑧, 𝑊, 𝑈𝑘, 𝑈 are the GRU weight matrix
values. 𝜎 indicates the reasonable sigmoidal function; 𝑡𝑎𝑛ℎ
indicates the tanh function; ⊙ indicates the element Figure 4. Performance Evaluation
multiplication operation; 𝑧𝑡 indicates the update gate, which
Figure 4 illustrates the findings, which demonstrate that
determines the degree of informing of the GRU unit's
all deep learning methods offer high evaluation metrics in
activation value based on the state of the input and the state
contrast, the averages for benign are 99.92%, 98.85%, and
of the earlier hidden layer in tandem; 𝑟𝑡 indicates the
99.90% for precision, recall, and F1-score.
rearrange gate, whose informing procedure is comparable to
that of 𝑧𝑡; The candidate hidden layer I indicated by ℎ𝑡̃ ,
while the hidden layer indicated by ℎ𝑡 .
4. RESULT AND DISCUSSION
As shown in the next sections, we tried several
experiments in this work to address the privacy issue using
deep learning algorithms. By carrying out safe data storage,
analysis, and exchange, the model guarantees the security
and privacy of the system. Specific criteria, including
precision, accuracy, F1-score, and recall, have been used to
compare the suggested method to the existing
methodologies.

Figure 5. Encryption Time

Figure 5 illustrates the encryption time for the IoT-CISP
technique, defined as the duration an encryption algorithm
takes to generate ciphertext from plaintext. The encryption
time depends on the number of images used in the process.

85
 C. Senthil Singh et al. / IJDSAI, 02(03), 81-87, 2024

The proposed method demonstrates faster encryption CONFLICTS OF INTEREST

compared to existing techniques like CP-ABE, GDBR, and
The authors declare that they have no known competing
HP-CPABE, which require more time. The proposed IoT-
financial interests or personal relationships that could have
CISP technique not only achieves quicker encryption but also
appeared to influence the work reported in this paper.
performs better with larger datasets. According to the results,
IoT-CISP achieves encryption times that are 31.24%, FUNDING STATEMENT
23.12%, and 33.03% faster than CP-ABE, GDBR, and HP-
Not applicable.
CPABE, respectively.
ACKNOWLEDGEMENTS
The author would like to express his heartfelt gratitude
to the supervisor for his guidance and unwavering support
during this research for his guidance and support.
REFERENCES
[1] K. Singh and I. Gupta, “Online information leaker
identification scheme for secure data sharing”, Multimedia
Tools Appl., vol. 79, no. 41, pp. 31165–31182, 2020.
[CrossRef] [Google Scholar] [Publisher Link]
[2] I. Gupta, A. K. Singh, C. N. Lee, and R. Buyya, “Secure data
storage and sharing techniques for data protection in cloud
environments: A. systematic review, analysis, and future
directions”, IEEE Access, vol. 10, pp. 71247–71277, 2022.
Figure 6. Decryption time [CrossRef] [Google Scholar] [Publisher Link]
[3] D. Saxena, I. Gupta, A. K. Singh, and C. N. Lee, “A fault-
Figure 6 shows the decryption time for the IoT-CISP tolerant elastic resource management framework toward high
technique, which refers to the time it takes for the decryption availability of cloud services”, IEEE Trans. Netw. Service
algorithm to convert ciphertext back into plaintext. This Manage., vol. 19, no. 3, pp. 3048–3061, Sep. 2022. [CrossRef]
decryption time varies based on the number of data involved [Google Scholar] [Publisher Link]
in the process. The proposed IoT-CISP method achieves [4] R. Gupta, D. Saxena, I. Gupta, A. Makkar, and A. K. Singh,
faster decryption compared to existing techniques like CP- “Quantum machine learning driven malicious user prediction
ABE, GDBR, and HP-CPABE, which require more time. for cloud network communications”, IEEE Netw. Lett.,
Furthermore, the IoT-CISP technique performs better with [CrossRef] [Google Scholar] [Publisher Link]
[5] I. Gupta and A. K. Singh, “SELI: Statistical evaluation-based
larger datasets. The results indicate that IoT-CISP is 21.24%,
leaker identification stochastic scheme for secure data
13.12%, and 23.03% faster in decryption times compared to sharing”, IET Commun., vol. 14, pp. 3607–3618, 2020.
CP-ABE, GDBR, and HP-CPABE, respectively. [CrossRef] [Google Scholar] [Publisher Link]
5. CONCLUSION [6] IBM Ponemon, “Cost of a data breach study”, 2022. [Online]
Available: https://www.ibm.com/security/data-
In this research a novel Internet of Things based Cloud breach[CrossRef] [Google Scholar] [Publisher Link]
Information Security Preservation (IoT-CISP) has been [7] I. Gupta and A. K. Singh, “Dynamic threshold-based
proposed. This approach enhances the model’s effectiveness information leaker identification scheme”, Inf. Process. Lett.,
and ensures security by first separating sensitive data from vol. 147, pp. 69–73, 2019. [CrossRef] [Google Scholar]
[Publisher Link]
non-sensitive data using an SVM classifier, and then [8] X.Ma, J.Ma, H. Li, Q. Jiang, and S. Gao, “PDLM: Privacy-
employing this data for partial decryption and analysis. preserving deep learning model on cloud with multiple keys”,
Sensitive data is protected through Okamoto-Uchiyama IEEE Trans. Serv. Comput., vol. 14, no. 4, pp. 1251–1263,
encryption, ensuring that data storage, analysis, and sharing Jul./Aug. 2021. [CrossRef] [Google Scholar] [Publisher Link]
are conducted securely to maintain the system’s safety and [9] W. Zhao, W. Yang, H. Wang, T. Zhang, D. Man, T. Liu, J. Lv,
privacy. The effectiveness of this novel method was assessed and M. Guizani, “Privacy-Preserving Outsourcing of K-Means
against existing methodologies using parameters like Clustering for Cloud-Device Collaborative Computing in
precision, accuracy, F1 score, and recall, revealing its Space-Air-Ground Integrated IoT”, IEEE Internet of Things
Journal, vol. 10, no. 23, pp.20396-20407, 2023. [CrossRef]
superior security and efficiency compared to other schemes.
[Google Scholar] [Publisher Link]
Results demonstrate that the IoT-CISP approach offers
[10] S. Fugkeaw, L. Wirz, and L. Hak, “Secure and Lightweight
encryption times that are 31.24%, 23.12%, and 33.03% Blockchain-Enabled Access Control for Fog-Assisted IoT
shorter than those of the CP-ABE, GDBR, and HP-CPABE Cloud Based Electronic Medical Records Sharing”, IEEE
algorithms, respectively. In the future, the current model will Access, vol. 11, pp.62998-63012, 2023. [CrossRef] [Google
be tested on additional benchmark datasets as part of our Scholar] [Publisher Link]
research plans. [11] Z. Zhang, W. Zhang, and Z. Qin, “A partially hidden policy
CP-ABE scheme against attribute values guessing attacks with
online privacy-protective decryption testing in IoT-assisted
cloud computing”, Future Generation Computer Systems, vol.
123, pp. 181-195, 2021. [CrossRef] [Google Scholar]
[Publisher Link]
[12] S., Vaidya, A., Suri, V., Batla, I. Keshta, S.S.M. Ajibade, and
G. Safarov, “A computer-aided feature-based encryption

86
 C. Senthil Singh et al. / IJDSAI, 02(03), 81-87, 2024

model with concealed access structure for medical Internet of

Things”, Decision Analytics Journal, vol. 7, pp. 100257, 2023. Sameena Naaz is currently working as a Senior
[CrossRef] [Google Scholar] [Publisher Link] Lecturer at the University of Roehampton, London,
United Kingdom. She holds the post of Professor at
[13] C. Wang, J. Pan, J. Lu, and Z. Wu, “A Data Hierarchical
the Department of Computer Science and
Encryption Scheme Based on Attribute Hiding under Multiple Engineering, Jamia Hamdard, New Delhi, India (On
Authorization Centers”, Electronics, vol. 13, no. 1, pp. 125, leave). She has a total experience of more than 23
2023. [CrossRef] [Google Scholar] [Publisher Link] years. She received her B.Tech (Computer Engg.)
[14] M. Li, D. Xiao, H. Huang, and B. Zhang, “Multi-level video from Aligarh Muslim University, in 1998 and the
quality services and security guarantees based on compressive M.Tech. Degree in Electronics with Specialization in Communication and
sensing in sensor-cloud systems”, Journal of Network and Information Systems from Aligarh Muslim University, in 2000. She
Computer Applications, vol. 205, pp.103456, 2022. [CrossRef] completed her Ph. D from Jamia Hamdard in the field of distributed systems
in year 2014. Sameena Naaz has published several research articles in
[Google Scholar] [Publisher Link]
reputed International Journals and Proceedings of reputed international
[15] N. Azam, L. Michala, S. Ansari, and N.B. Truong, “Data conferences published by IEEE and Springer. Her research interests include
privacy threat modeling for autonomous systems: A survey Distributed Systems, Cloud Computing, Big Data, Machine Learning, Data
from the GDPR's perspective”, IEEE Transactions on Big Mining and Image Processing.
Data, vol. 9, no. 2, pp.388-414, 2022. [CrossRef] [Google
Scholar] [Publisher Link]
[16] B. Huang, J. Gao, and X. Li, “Efficient lattice-based revocable G Saranya received her B.E degree in Computer
attribute-based encryption against decryption key exposure for Science and Engineering from Anna University,
Chennai and M.E degree in Computer Science and
cloud file sharing”, Journal of Cloud Computing, vol. 12, no
Engineering from Hindustan University, Chennai. She
1, pp.37, 2023. [CrossRef] [Google Scholar] [Publisher Link] started her career as an Assistant Professor and has 9
years and 6 months of experience. Currently she is
AUTHORS working as an Assistant Professor in S.A. Engineering
C. Senthil Singh received his B. E, MTech, PhD College, Chennai. Her research interests include Deep
degree in Information and Communication from Learning and Cloud Computing. She is a lifetime member of ISTE.
Anna University, Chennai, India. He has a great
flair for teaching and research and has a total
experience of 22 years in teaching in Engineering.
His professional interests include VLSI, Wireless
Communication, Embedded Systems and Arrived: 29.04.2024
Telemedicine.
Accepted: 12.06.2024

87