An Efficacy Analysis of Data Encryption Architecture - 2023 - Procedia Computer
An Efficacy Analysis of Data Encryption Architecture - 2023 - Procedia Computer
com
ScienceDirect
Available
Availableonline
onlineatatwww.sciencedirect.com
www.sciencedirect.com
Procedia Computer Science 00 (2022) 000–000
www.elsevier.com/locate/procedia
ScienceDirect
ScienceDirect
Procedia Computer Science 218 (2023) 989–1002
Procedia Computer Science 00 (2022) 000–000
www.elsevier.com/locate/procedia
All hardware resources are viewed as services and delivered via the Internet in order to provide reliable, quick, and
efficient data management and net computing services. Benefits of the cloud include integration, adaptability,
flexibility, availability, the ability to adjust to changes in demand, the power to speed up development activity, and
the potential for cost savings through simplified and effective processing.
In order to meet users' computing needs, CC combines a number of computational ideas and technologies, including
Service Oriented Architecture (SOA), Web 2.0, selection and placement of virtual machine, and others with a
dependence on the Internet. Customers' operating system and data are kept on web server, and common business
applications are made available online through web browsers. In some aspects, the term "cloud computing" refers to
the development of these technologies and serves as a marketing term for both that development and the services they
provide [3].
Despite the many advantages of cloud computing, the most crucial area of concern is privacy and security. Cloud
computing's key security aspects are data security, access management, data utilization management, and trust. We
give a comparative assessment of the existing Encryption Architectures for Cloud platforms in this study, which allows
us to determine the best data security and retrieval model to use.
The cloud-based information must be protected, especially data protection in the cloud storage, in order to be
secure. By prohibiting un-authorized access, this enhances data security [4]. Although it is the cloud based supplier's
duty to offer users a stable and reliable storage service, a number of circumstances could compromise the safety and
integrity of user data. In such a case, the service provider might be forced to forgo the loss of the user's banking system
and reputation in an effort to cover the full extent of the damage.
In order to prevent data loss due to a single storage node's original reliability of the data being subverted, data in
cloud services is often separated into slices and encrypted before being stored in multiple storage nodes [2].
Because of these traits, certain standard network and storage security technology are no longer fully relevant in
cloud storage environments. For instance, message digital signatures are employed in traditional storage technology
to confirm file integrity, but with cloud storage, the data is stored on a remote server, making it impossible to
periodically retrieve the data and validate the signature to ensure data integrity.
There is a requirement for a trustworthy and efficient fault - tolerant system that can guarantee that even if a number
of slices are lost, the security of files may be restored utilising the remainder slices in addition to cloud storage, which
separates data into fixed-size parts.
Agencies are now attempting to avoid focusing on the IT infrastructure. They must focus on their business
operations to boost productivity. In comparison to the traditional IT model, cloud computing has several advantages.
Questions regarding cloud computing security, on the other hand, are a key barrier to cloud computing adoption from
the user's perspective. The availability of computer network services, primary data storage, and processing power,
without explicit user active control, is referred to as cloud computing. Using cloud service suppliers' services, cloud
data is processed and retrieved on a web server.
As a result, the value of cloud computing is increasing, making it a rising market that is attracting a lot of interest
from the educational and business sectors. The cloud storage solution, on the other hand, has several drawbacks,
including a lack of access and security concerns. Because the cloud storage service is centered on two-way data sharing
between the service provider and user, cloud computing security concerns include honesty, integrity, availability,
verification, permission, and confidentiality.
As a result, the risk of data compromise is increasing, and it may be classified into two categories: vital data and
archival material. Important information is information that a subscriber needs at any given time and would be irritated
by any halt or disappearance. Furthermore, archival data is data that is extremely rare in its entirety, and typically in
a non-critical moment. As a result, a gap in it won't be able to be regarded as a major issue.
When using the internet-cloud platform, data security and privacy must be prioritized. Data loss or disclosure can
have a significant negative impact on a company's brand and confidence. Data leakage prevention is viewed as the
most pressing issue, accounting for 88 percent of main concerns. Similarly, data remoteness and privacy have a 92
Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002 991
Author name / Procedia Computer Science 00 (2019) 000–000 3
percent impact on security issues. Data protection, dependability, honesty, availability, authentication, and secrecy, as
well as a lack of resources and skills, are among the most serious security challenges in cloud computing [4].
• Integrity: Data in the cloud can be harmed as a result of data transmission to cloud storage. Because the
information and calculations are outsourced to a web server, the data's validity should be monitored and
maintained at all times to ensure that the information and calculations interact. Data integrity refers to the
protection of records from tampering. It is necessary to make some adjustments to the specifics.
• Availability: The ability of a cloud subscriber to receive critical data at any time is referred to as
accessibility. The ability to keep access to cloud computing resources at any moment is a crucial concern
for each firm. A system is called available when an authorized individual can use and control the device
at any time and preserve data.
• Vendor lock-in: Companies that employ cloud-based services frequently opt to change their Cloud Service
Provider (CSP) and go with a different one. This could be an explanation since the CSP will no longer
adapt to the tenant's needs, regardless of whether there are improvements or upgrades in the services that
the client does not expect. They are unable to meet the needs of clients or any other trigger that causes the
customer to migrate to a different CSP, but they are unable to exit this situation, which is known as Vendor
Lock-in.
• Data Security: To increase the security of cloud computing, it is necessary to provide encryption,
certification, and intrusion detection for information maintained in the cloud.
• Interoperability: This is defined as the willingness of two or more processes to collaborate to share and
use data. Companies will be unable to combine their information technology network in the cloud to obtain
efficiency and cost savings if these systems are not integrated. There are also cloud-computing networks
that are designed as closed systems that do not link.
Due to the quick development of information technology, there is a significant amount of data outsourcing to cloud
servers, and multiple attacks will jeopardize the secrecy of cloud data. Before being transferred to the cloud, user data
is regularly encrypted in order to prevent data leakage and guarantee data security. The fact that ciphertext is kept on
external servers prevents the use of standard search algorithms. It is required to apply an appropriate searchable
encryption technique to find the target data [5].
To address these issues, searchable encryption (SE) technology can provide data privacy and availability while also
allowing ciphertext data to be queried and retrieved. A searchable encryption approach often consists of encryption,
tokens, search, and decryption.
• Encryption: After encrypting the data and building the index structure, the user uploads the ciphertext and the
index structure to the server.
• Token: Users create a trapdoor for keywords using a key, and the token is required to keep any keyword
information private.
• Keyword search: The server runs the search algorithm using the keyword and returns the cypher text with the
matching keywords. Just the keyword information in the ciphertext must be obtained by the server.
• Decryption: Users decrypt the server's encrypted files with the help of the key to get the search results. Searchable
encryption systems can be divided into two categories:
i) Searchable symmetric encryption (SSE): SSE is a method for retrieving ciphertext that is based on
asymmetric encryption. Both data owners and users exchange the vital information.
ii) Searchable asymmetric encryption (SAE), a form of public key encryption that is suitable for one-to-many
data sharing scenarios. To guarantee its security, many hypotheses, including Decision Bilinear Diffie-
Hellman, is applied (DBH).
The separation of public and private keys makes the SAE technique particularly suitable for multi-user data sharing
systems, despite its typical inefficiency and reliance on bilinear pairings, which results in significant algorithm
complexity. The main motivation of this research is to develop complete data encryption and retrieval security system
in the cloud platform. It is known that the existing cloud system is not systematic. The development of Searchable
Encryption (SE), the accepted cloud data protection scheme gives end-to-end security and privacy. The devised
encryption system assures privacy and security. Further, research has been done using query encryption for document
that encrypts data in the cloud. The document is given a keyword, and during decryption, the ranking has been done
using ANN and extracting the best appropriate document based on the keyword.
992 Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002
4 Author name / Procedia Computer Science 00 (2019) 000–000
Based on contextual understanding and justifying the contemporary importance of the concern, this research is
developed as an analysis to identify the shortages and scopes of various Encryption Schemes used for data security in
a Cloud Environment.
The study's main objective is to determine an optimal data security and retrieval scheme in cloud platform through:
• Comparative analysis of various encryption architectures based on their performances
• Enlist the merits and demerits of the available cloud data encryption and retrieval schemes
• Justify the efficacy of the optimal data encryption architecture for cloud platforms based on the above
evaluations
• Using a combination of HAC Tree, Neural Network for index generation, and Cosine Similarity to support
the encryption mechanism, recognizes the value of Cosine Multi-Keyword Searchable Index (CMSI).
• Introduction of Neuro-Rank Policy
The presented research is arranged as follows: Section 1 introduces the concept of cloud computing followed by
the challenges and methods to address the various challenges. Section 2 illustrates the literature review in which the
pros and cons of state of art techniques have been described. Further, the research gap is defined in section 3. The next
section elaborates on the research methodology and experimental study. Section 5 defines the simulated results and is
lastly concluded in section 6.
2. Literature Review
Several scholarly works are done to assess, compare and refine the encryption architectures that are applied for
cloud computing data security purposes. Some significant studies are given in the following table 1.
3. Research Gap
The review of literature presented in the previous section shows a significant number of approaches developed and
tested over the past years (they are done even before the period covered in this study) to ensure privacy while using
cloud-based on trust, encryption and access control. However, none of them formally claimed to be universal and
eligible for providing complete data encryption and retrieval security in the cloud platform. Furthermore, the systems
are scattered and mostly not systematic.
Searchable Encryption (SE), the accepted cloud data protection scheme offers the privacy of the data when data is
transferred from local storage. Although, privacy is assured, it further complicates the taskthe at server level. For
traditional query-based search and retrieval processes, content and keywords are revealed. Therefore, encrypted search
also inhibits the searching operations.
Trapdoor-based user queries as experimented with within [6], [7], and [8] cannot fully ensure data protection from
harmful attackers, where they can guess the keyword through the use of a constant trapdoor. Thus, the attackers can
identify the frequent trapdoors. Although [8] has enabled authenticated keyword search as an improvement over the
PEKS scheme, the model is vulnerable to deliberate targeted attacks. Also, these models are high on overhead costs.
Multiple keyword-based searches are secure, faster, and low cost than single keyword-based searches [11]. These
tools, such as the model used in [9] also reduce query representation and its expression problem (typos, spelling
mistakes, etc.).
Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002 993
Author name / Procedia Computer Science 00 (2019) 000–000 5
Hui Yin et 2017 They suggested a search strategy that Ensures user's privacy of query Time-Consuming to do the search
al. [6] boosted privacy by allowing the data to content; Secure Search Scheme on a cloud server than KNN and
generate a different random query SSE
trapdoor each time. We create a secure
indexing for each data frame using the
Bloom filter and the bilinear pairing
operation, allowing the cloud to do a
search without receiving any helpful
information.
Tahir et al. 2017 By precisely describing keyword-trapdoor Lightweight; Efficient in Security concerns exist on
[7] in distinguishability and trapdoor-index performing large-scale data search. leakages through Index table and
table in distinguishability, they enumerate search outcome; although no
the characteristics of a "secure" ranked SE information is exposed of data
system. To counter passive attacks, they outsourcing; works ONLY with a
developed and demonstrated a novel single keyword
Ranked based SSE that is entirely based
on a probabilistic encryption algorithm.
Huang et 2017 Presented the idea of Public-key The scheme is comparable in Bears security risk in a real-life
al. [8] Authenticated Encryption with Term efficiency with Boneh et al.’s condition where the attacker
Search (PAEKS), which allowed the data scheme. chooses a user to collect data
sender to both encrypt and authenticate a information.
keyword. The verification would be
persuaded that the sender alone is
responsible for creating the encrypted
keyword. Based on straightforward and
static presumptions, the random oracle
model's security was examined using the
provided security models.
Poh et al. 2017 Presented a comparative study on Most models used a combination Practical use of external devices to
[9] available Searchable Symmetric of index tables and trees for reduce leakages is not properly
Encryption (SSE) models to classify their reliable updates and storage; analyzed; Mostly uses Index
features and evaluate the model efficacy External mechanisms such as Tables that are prone to leakage
Oblivious RAM can be employed risk; Search is time-consuming;
to reduce leakages. Not a preferred option where I/O
access is essential
Ilakiya 2019 OTP-based Secured Information Retrieval Data will be protected by using High storage is required for this
et.al. [10] from the Cloud Using Human Voice. multi authentications. kind of authentication. This kind of
authentication can also be
extraneously influenced by one
sore throat and cold.
Sun [5] 2019 Encryption and decryption of images More secure in the case of Increase the communication range
using classical and quantum cryptography. exchanging multimedia data. and bit transfer rate.
Shan Jiang 2019 Proposed a multi-keyword search protocol Low in computational cost as most Unreliable and not feasible for
et al.[11] with bloom filter support that is more of the data are excluded for the use large-scale search; Blockchain data
effective and protects user privacy. In the of low-frequency filter; efficient can be mishandled by attackers;
protocol, a multi-keyword search database with dynamic update Risk of data loss
operation filtered the database using a option; faster and safer than single-
low-frequency term chosen using a bloom keyword based blockchain models
filter. It was suggested to utilise
pseudorandom tags to make it easier to
finish each search operation in only one go
994 Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002
6 Author name / Procedia Computer Science 00 (2019) 000–000
Malhotra et 2019 The study demonstrates brand-new secure Multi-keyword-based model and so Model accuracy may vary based on
al. [12] storage and the document ranking system better in search flexibility; Can be real-time database configuration
for the cloud. The data is encrypted based enhanced with other machine
on the correlation between the data files learning algorithms, such as SVM,
determined by cosine similarity because ANN
no previous reference for any data is
retained on the server. Through the use of
supervised machine learning, the retrieved
data is ranked.
Islam et al. 2019 Set up a secure authentication system and Better data security from hacking Time-consuming
[13] appropriate cryptography for cloud
computing. Auto encryption and KEYs
changing processes were part of the
operation in the cloud end. Customers
were initially not sent newly generated
KEYs. In order to authenticate users, three
actions would be taken. The encryption
procedure can be started manually by
Cloud Service Providers (CSP) at any
time or automatically once users log out.
Suneetha et 2019 ANN was integrated to enhance the Focussed on ensuring data Can face the risk of data loss in real
al. [14] security and confidentiality of the cloud confidentiality practice with unskilled/improper
computing environment. The work handling
involved a dynamic hashing component
for the storage of the sensitive data.
E. Nirmala 2021 Constructed a keyword searching Binary Better in security and reliability; Model chiefly focussed on text-
et al. [15] Tree algorithm added with multiple less time consuming; more based search
corrections features based on ranking. compatible than traditional
Fuzzy Gramm was used to address methods; workable with
spelling errors while ancestral relations single/multiple keywords
can be found based on stemming
procedure.
Tyagi et al. 2021 AES and Fernet were used to present to Effective in Image Encryption in AES, when adopted/utilized on
[16] double the encryption level. Along with Cloud Environment Solid-State-Drives (SSDs)
CNN auto-encoders were also used to it’s identified as less foolproof and
protect the data that is available in the thus adopts a hybrid
form of images in the cloud. model is recommended for
confidential datasets.
For Fernet, the key disadvantage is
that the key could be obtained by
third parties while transferring to
the receiver’s end which is
highly-risky and a huge drawback
in Fernet and other
symmetric cryptography.
Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002 995
Author name / Procedia Computer Science 00 (2019) 000–000 7
Sana et al. 2021 A safe and secure data communication The model can work for speech and The model is operated with
[17] was assured with the secure design based voice recognition; less time- homomorphic encoding and so
on ANN and encryption techniques. Here, consuming; better in data retrieval needs to take care of the noise
third could access the data in encrypted accuracy reduction measures; it is costly;
form and thus data is not disclosed to high in computational expenses;
ensure privacy. The work involved Matrix MORE encryption used in the
Operation-based Randomization and model may pose a security risk
Encipherment (MORE) along with neural
network architecture.
Zulifqar [3] 2021 The work have presented a Verifiable Less complexity of computation. Perform better only with fewer
public key encryption in a multi-user search keywords.
cloud platform to represent a
homomorphic encryption.
Bernardo 2021 Analysed the fundamental concepts of Easy to deploy; Enhances Difficulties exist in performance
Pulido- Fully Homomorphic Encryption (FHE) reliability analysis, bootstrapping, and
Gaytan et revolving around a cloud environment overhead.
al. [18] with discussion on practical
implementations, advantages, limitations,
practical implications in concern to neural
networks.
Ma et al. 2022 Designing the hybrid encryption The proposed model improved There is a small reduction in
[19] technique using the encrypted images and DenseNet model is 8 to 9 times accuracy due to the adoption of the
DenseNet model has been used for a fine- smaller than the standard encryption model.
tuning ad a feature extractor was convolution method.
constructed to improve the performance.
Zhang et 2022 An encrypted retrieval scheme was The proposed system protects data There is a leakage of cipher text
al. [20] proposed considering the multiuser search privacy. location and a fuzzy keyword
encryption model. The authors used the search query was not used for
LSTM model to extract the semantic efficient encryption,
features.
Wang et al. 2022 A novel searchable technique has been The system's computational and The time to execute the ciphertext
[21] developed using the Linear Secret Sharing storage efficiency is high. transfer is more in comparison to
scheme. The authors used the various other search techniques.
search mechanisms and used 0,1 coding
theory.
Single Keyword-based searches are usually found to be time-consuming, costly (from peer-stakeholder charges,
such as computational resources, bandwidths, etc.), and prone to attacks when repeatedly used. The attacker can
retrieve them by guessing the common encryptions.
Index table-based traditional searchable symmetric encryption tools in [9] have a high risk of data leakage. Ranked
Searches are considered efficient and usable for large-scale data over cloud platforms as [7] and [15] based on their
reliability and privacy preservation. Asymmetric Key Encryption tools [10] are considered more secure than
Symmetric Encryption tools. However, they are time-consuming and costly to develop and maintain.
Data retrieval systems with Bloom Filter reduce the search time as the method can directly search the index in place
of scanning the whole cipher text. However, bloom filter is also associated with a limitation [11] that they have false
positive probability and so are unreliable.
Homomorphic Encryptions [17], [18] that are currently implemented in cloud data security are reliable applications.
But they are costly and most variable in their performances in real-life conditions. The Machine Learning cloud data
storage and retrieval tools as [12] – [17] are efficient and more flexible in terms of search types and can be utilized
for image and voice recognition. However, searchable encryption gaining attention in terms of flexibility [22].
They reduce computational time and privacy-preserving. But, with the rise of deep learning-based cyber-attacks,
these tools are found to be exposed to the latest malicious intrusions. Thus, for real-world applications these tools
996 Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002
8 Author name / Procedia Computer Science 00 (2019) 000–000
need to be equipped. Further, the latest technological advances involving neural based processing architectures and
block chain assure promising future.
A thorough analysis is required on the existing research and suitable adaptability to prevent the data-
stealing/damage that is commonly occurring due to cloud platform limitations or cyber-attacks.
4. Methodology
The primary goal of this research is to provide a methodology for document or query encryption that encrypts data
in the cloud. The document is given a keyword, and during decryption, ANN is used to rank and extract the best
appropriate document based on the keyword. Certain parameters are computed to determine whether a document is
approved or denied for a certain keyword.
Term frequency (TF) and Inverse document frequency (IDF) are utilized for the computation of the HAC index in
the present section. Here, TF represents the ratio of the occurrence of the term to the total number of terms in the
dataset. And IDF is the ratio of one term occurrence in the present document to the total number of occurrences in
another document. HAC is calculated with two components M in the numerator and Q in the denominator and it is
mathematically represented by the following equations
TF
M= (1)
k 2
t =1
(TFe )
n
J i count
Here, TF = i =1
(2)
n
J=current term for the processing
IDFe
Q= (3)
(TF ) e
2
TFpresent
Here, IDF = (4)
t =1TFother
k
e= number of elevations
M
HACindex = (5)
Q
Machine learning is utilized for the ranking of SE keywords. This Machine Learning architecture perform training
based on the data extracted using TF, IDF, M, and Q. The three layered neural network architecture is shown in Fig.
1. The four input values are presented by P1, P2, P3 and P4 to generate the output using Machine Learning index
which ranks the document.
Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002 997
Author name / Procedia Computer Science 00 (2019) 000–000 9
Where,
docxi = docxi ,1 , docxi ,2 ......docxi ,m
chk = chk ,1 , chk ,2 ,............, chk ,m
Search
= Index log( HAC + Neural Index) (7)
Here,
Upper Boundary (UB)= Nei + Nei .30 / / where Nei is the stored index, A 30% upper margin is considered.
Lower Boundary (LB) = Nei − Nei .30 / / 30% lower margin
If
SI LB, and SI UB then, RL.append which is adding the recommendation value otherwise reject the
document. The process of the proposed model is as in the below displayed flow chart in Fig. 2.
998 Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002
10 Author name / Procedia Computer Science 00 (2019) 000–000
5. Results
In this section, the list of keywords that are mentioned for the search word is displayed and all the parameters get
calculated. The results for recommendation are also tabulated below.
The tweets that are used as the keywords are summarized in Table 2. The search term chosen is “Sheenam requires
a dentist with knowledge of tooth gums”.
‘canal’ ‘afford’ ‘dentist’ ‘root’ ‘teeth’ ‘advil’ ‘cavity’ ‘solution’ Work Root Enduranc
e
‘tooth’ ‘dentist’ ‘afford’ ‘glasses’ Ignore ‘pain’ ‘jobs’ ‘money’ ‘hurt’ Whiskey Teeth
‘filling’ ‘afford’ ‘insuranc ‘dentist’ Appoi ‘teeth’ ‘genes’ Afford Fee Consump Gum
e’ ntment tion
Cricket dentist pain teeth yard list runs wicket toss win loose
Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002 999
Author name / Procedia Computer Science 00 (2019) 000–000 11
The generated indexes have been used for the storage and processing of the documents against their ground truth
using Neural Networks (NN) having feed forward orientation methods. To evaluate the performance, true positive
rate (TPR) along with false positive rate (FPR) has been calculated to determine the significance of the document to
its relative class. In addition to that, a computation Cost Ratio (CCR) is also calculated for the proposed algorithm to
the indexes generated by HAC only and Neural only. The CCR is calculated with the NN classification algorithm.
The evaluation is done based on several relevant documents searched in the given interval of time that is being utilized
by the proposed algorithm. To check the performance, a time of 60 seconds is supplied to search and list data from
various categories.
As shown in table 4, the evaluations have been done using various data patterns that are associated with the dataset.
A total of 1000 data elements have been supplied and the proposed work algorithm performs significantly well when
1000 Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002
12 Author name / Procedia Computer Science 00 (2019) 000–000
it comes to TPR and FPR. The maximum TPR attained by the proposed algorithm is .95845908 at 100 data elements
viz. 100 searches containing tweets from various categories. The overall TPR of the proposed algorithm lies between
.92-.96 and the average is 0.938. The other state of art algorithms like multi-class SVM if combined with the proposed
index structure, result way behind the proposed algorithm. The maximum attained TPR with multi-class SVM is .8852
whereas the Naïve Bayes algorithm performs marginally better than multi-class SVM. The maximum attained TPR
for Naïve Bayes is 0.8952. As illustrated earlier, a total of 60 simulation seconds have been utilized against passed
relevant queries and a total number of true responses have been noted which is shown in table 5. The count values
obtained using the proposed, HAC and Neural Index are shown in Table 5.
For the same simulation period, a total of 760 true positives i.e true data against its supplied label is indexed whereas
it goes down with Hac and Neural Index alone. To calculate the CCR, the proposed algorithm divides the minimum
attained count by self-attain count viz. the CCR for the proposed will be min (All count viz. 530)/(Self Count viz.
760)=0.6973. A graphical representation for the same is shown in Fig. 3.
6. Conclusion
This research provides a cloud data encryption and retrieval model based on Symmetric Searchable Encryption
using Machine Learning. The proposed model is used to mitigate this risk and contribute to the data security solutions
in cloud data storage and retrieval systems. The suggested model uses Artificial Neural Networks that improve data
security in conjunction with an effective keyword ranking technique. Further, the authors have calculated the HAC
index and the Neural index for the query or the documents, and the search index will be created based on these two
factors. If the Search index of a document falls within the set boundaries, the document is approved; if the Search
index of a document falls outside or below the stated limits, the document is refused. The comparative analysis of the
proposed work is performed against multi-class SVM and Naïve Bayes. It was observed that even for 100 searches
the proposed work exhibited high TPR along with low FPR in comparison to multi-class SVM and Naïve Bayes along
with a minimal CCR of 0.6973. These outcomes contribute to the encryption of documents in the cloud with great
security, as well as the retrieval of data based on a specific keyword that has been assigned. As a result, this study aids
in the prevention of uncertainties such as cyber-attack plans and cloud storage system errors. In the future, an attempt
has been made to improve the retrieval performance and enhance the data security model using the fuzzy model.
Acknowledgments
References
[1] Tari, Yi, Premarathne, Bertok, and Khalil. (2015) “Security and privacy in cloud computing: vision, trends, and challenges.” IEEE Cloud
Computing, 2(2): 30-38.
[2] Ratanghayra. (2017) “Review on Dynamic Multi-Keyword Ranked Search over encrypted mobile cloud data.” IJNRD-International Journal
of Novel Research and Development (IJNRD), 2(12): 8-10.
[3] Hashizume, Rosado, Fernández-Medina, and Fernandez. (2013) “An analysis of security issues for cloud computing.” Journal of internet
services and applications, 4(1): 1-13.
[4] Zulifqar, Anayat, Kharal, (2021) “A Review of Data Security Challenges and their Solutions in Cloud Computing.” International Journal of
Information Engineering & Electronic Business, 13(3): 32-41.
[5] Sun. (2019) “Privacy protection and data security in cloud computing: a survey, challenges, and solutions.” IEEE Access, 7: 147420-147452.
[6] Hui Yin, Zheng Qin, Lu Omang, and Keqin Li. (2017) “A Query Privacy-Enhanced And Secure Search Scheme Over Encrypted Data In
Cloud Computing.” Journal of Computer and System Sciences, 2 (90): 14-27.
[7] Tahir, Ruj, Rahulamathavan, Rajarajan, and Glackin. (2017) “A new secure and lightweight searchable encryption scheme over encrypted
cloud data.” IEEE Transactions on Emerging Topics in Computing, 7(4): 530-544.
[8] Huang, and Li. (2017) “An efficient public-key searchable encryption scheme secure against inside keyword guessing attacks.” Information
Sciences, 403: 1-14.
[9] Poh, Chin, Yau, Choo, and Mohamad. (2017) “Searchable symmetric encryption: designs and challenges.”ACM Computing Surveys (CSUR),
50(3): 1-37.
[10] Ilakiya, Vijithra, Kuppusamy, and Mahalakshmi. (2019) “Impact of Asymmetric Encryption in Cloud Computing: A Study.” International
Journal of Computer Sciences and Engineering, 7(3): 32-43.
[11] Jiang, Cao, McCann, Yang, Liu, Wang, and Deng. (2019) “Privacy-preserving and efficient multi-keyword search over encrypted data on a
blockchain.” In 2019 IEEE International Conference on Blockchain (Blockchain), IEEE: 405-410.
[12] Malhotra and Singh. (2019) “An Optimized Solution for Ranking Based On Data Complexity.” International Journal of Innovative
Technology and Exploring Engineering(IJITEE), 8(11): 41-49.
[13] Islam, Chaudhury, and Islam. (2019) “A simple and secured cryptography system of cloud computing.” In 2019 IEEE Canadian Conference
of Electrical and Computer Engineering (CCECE), IEEE: 1-3.
[14] Suneetha, Kishore, Singh, (2019) “A Security Model Using Artificial Neural Networks and Database Fragmentation in Cl Environment”,
International Journal of Recent Technology and Engineering (IJRTE) 8(2): 34-43.
[15] Nirmala, Muthurajkumar, and Subitha. (2021) “An Efficient Privacy-Preserving Ranked Keyword Search Method.” In IOP Conference
Series: Materials Science and Engineering, 104(1): 102-112.
[16] Tyagi. (2021) “Enhancing Security of Cloud Data through Encryption with AES and Fernet Algorithm through Convolutional-Neural-
Networks (CNN).” International Journal of Computer Networks and Applications, 8(4): 288-299.
1002 Sheenam Malhotra et al. / Procedia Computer Science 218 (2023) 989–1002
14 Author name / Procedia Computer Science 00 (2019) 000–000
[17] Sana, Li, Javaid, Liaqat, and Ali. (2021) “Enhanced Security in Cloud Computing Using Neural Network and Encryption.” IEEE Access, 9:
145785-145799.
[18] Pulido-Gaytan, Tchernykh, Cortés-Mendoza, Babenko, Radchenko, Avetisyan, and Drozdov. (2021) “Privacy-preserving neural networks
with Homomorphic encryption: Challenges and opportunities.” Peer-to-Peer Networking and Applications, 14(3): 1666-1691.
[19] Ma, Zhou, Qin, Xiang, Tan, and Cai, (2022) “A privacy-preserving content-based image retrieval method based on deep learning in cloud
computing.” Expert Systems with Applications, 2(3):117508.
[20] Zhang, Qiuyu, Minrui Fu, Yibo Huang, and Zhenyu Zhao (2022) "Encrypted Speech Retrieval Scheme Based on Multiuser Searchable
Encryption in Cloud Storage." Security and Communication Networks.
[21] Wang, Haiyan, Yuan Li, Willy Susilo, Dung Hoang Duong, and Fucai Luo(2022) "A fast and flexible attribute-based searchable encryption
scheme supporting multi-search mechanism in cloud computing." Computer Standards & Interfaces: 82, 103-115.
[22] Wang, Q., He, M., Du, M., Chow, S.S., Lai, R.W. and Zou, Q., (2016) “Searchable encryption over feature-rich data”, IEEE Transactions on
Dependable and Secure Computing, 15(3), 496-510.