✅
International Publishing Platform
Serving Researchers Since 2012

A Proposed Methodology to Improve the Integrity and Confidentiality of Cloud Computing Services

DOI : 10.5281/zenodo.24361197
Download Full-Text PDF Cite this Publication

Text Only Version

A Proposed Methodology to Improve the Integrity and Confidentiality of Cloud Computing Services

Mohammad Yaser Hashisho

Department of Computer Systems and Networks Faculty of Informatics Engineering, University of Aleppo, Syria

Souheil Khawatmi

Department of Computer Systems and Networks Faculty of Informatics Engineering, University of Aleppo, Syria

Abstract – Cloud storage has to keep outsourced data confidential and intact without giving up the processing speed and the low resource use that make the cloud attractive. Symmetric ciphers used alone are fast but leave the secret key exposed while it is shared, and schemes that add a blockchain ledger to guarantee integrity pay for it with long delays and heavy use of server resources. This paper proposes an authenticated hybrid cryptosystem for data at rest that encrypts file chunks with the Advanced Encryption Standard, protects the chunk keys with elliptic curve cryptography on the prime256v1 curve, and verifies every chunk with a hash based message authentication code, while the encrypted keys are isolated in a separate key vault. The methodology was implemented in the CloudSim simulator and compared with single symmetric ciphers, with a recent hybrid that pairs the same two algorithms with a plain hash, with a blockchain based hybrid, and with a hybrid built on the Rivest Shamir Adleman algorithm. It encrypted 40 percent faster than the hash based hybrid. Against the blockchain based hybrid it reduced encryption time by up to 89 percent and decryption time by up to 70 percent for files of 30 to 100 megabytes, lowered power draw by 15 to 29 percent, and kept the avalanche effect above 50 percent. For payloads of 64 to 288 bytes encryption took a constant 3.5 milliseconds, more than 99 percent below the compared hybrid. The results show that integrity can be verified in the cloud without a blockchain at a small and predictable cost.

Keywords – cloud computing; hybrid encryption; elliptic curve cryptography; prime256v1; AES; HMAC; avalanche effect; CloudSim

  1. INTRODUCTION

    Cloud computing delivers software, platforms and infrastructure as services over the Internet, so that organizations and individuals can store, process and share data from any place without owning the underlying hardware [1]. Storage is the most widely used of these services: it acts both as a large backup and as a primary medium, and it removes the need to keep data on a personal device. The same property is the main obstacle to adoption. Data placed on servers that the owner does not control can be read, sold or modified by intruders, and in some cases by the administrators of the provider, and the surveys of cloud security list data protection ahead of every other challenge [2].

    The technical problem has three parts. First, sensitive data on shared external storage is exposed to theft and to unauthorized modification. Second, a symmetric cipher such as the Advanced Encryption Standard (AES) encrypts large volumes quickly, but used alone it offers no safe way to distribute the secret key between the user and the cloud server.

    Third, a conventional asymmetric cipher such as Rivest- Shamir-Adleman (RSA) solves key distribution only with keys of 2048 or 3072 bits, which makes it slow and limits it to very small blocks. Recent proposals close the integrity gap by recording data fingerprints on a blockchain, but the consensus and ledger operations add a delay and a resource load that real- time storage cannot afford.

    This paper proposes a methodology that secures data at rest without that load. Each file is divided into independent chunks; every chunk is encrypted with AES under its own key and authenticated with a hash-based message authentication code (HMAC); the per-chunk keys are encrypted with elliptic curve cryptography (ECC) on the prime256v1 curve and kept in a key vault that is separate from the data servers. The contributions are:

    • an authenticated hybrid design (AES, ECC prime256v1 and HMAC-SHA256) in which ciphertext, tags and keys are stored apart, so that a breach of one store does not expose the data;

    • an implementation in CloudSim with an evaluation of execution time, power and energy, throughput, storage overhead and avalanche effect on files of 30 to 100 MB and packets of 64 to 288 bytes;

    • a comparison with single symmetric ciphers, with an ECC-AES-SHA hybrid, with a blockchain-enabled AES-ECC hybrid and with an RSA-AES hybrid, together with recommended chunking thresholds and the scope of suitable applications.

    Section II reviews related work. Section III describes the proposed methodology and Section IV the experimental setup. Section V presents and discusses the results, and Section VI concludes the paper.

  2. RELATED WORK

    1. Security of Data Stored in the Cloud

      Saroj et al. [3] used threshold cryptography: the data owner divides users into groups, gives each group one decryption key and each member only a share of it, and controls access through a capability list, while a one-time session key agreed with Diffie-Hellman protects the transfer. Harikrishna and Amuthan [4] described key generation, digital signature and signature verification with an access list and Diffie-Hellman key exchange; the approach remains exposed to denial-of- service attacks. Chandu and Kata [5] combined block and stream ciphers for confidentiality with hashing for integrity, but

      did not address user verification. Ziglari and Yahya [6] placed an encryption service between data processing and storage and hid the key from both; the repeated transfers between services are the weakness of the model. In [7] all files are merged into one container that is encrypted, fragmented and uploaded with a protected storage key, which forces a user to download and decrypt the whole container to recover a single file.

      Singh and Singh [8] encrypted with AES under a key derived from the user password with HMAC and authenticated the user with a digital signature, yet did not achieve data integrity. Rewagad and Pawar [9] combined a hash-based digital signature, Diffie-Hellman key exchange and AES, a three-way protection that keeps speed acceptable. The survey in [10] covers device-based identification, attribute-based encryption and blockchain for privacy, and notes that blockchain needs very large resources.

    2. Selection of the Cryptographic Algorithms

    Table I summarizes the studies that guided the choice of algorithms. The reviews in [11] and [12] found AES more secure and faster than DES, 3DES, Twofish and RSA, and the survey of cryptanalytic attacks in [13] shows that brute-force, ciphertext-only, side-channel, boomerang, power-analysis and algebraic attacks remain theoretical or implementation- dependent against the full cipher. Selvi and Sakthivel [14] built a hybrid of ECC, AES and SHA and showed that it gives stronger encryption with less computing power and bandwidth than 3DES and AES; the size of an ECC key grows linearly with the equivalent symmetric key, while an RSA key grows far faster. Their scheme, however, fixes no standard curve parameter, verifies integrity with an unkeyed hash and does not plan how files are partitioned. Bakro et al. [15] combined AES and ECC with a blockchain and reported a high level of security, at the price of the execution time and resources that the ledger consumes; the file is also split into two parts, which loads memory and the network. Akteret al. [16] encrypted the plaintext with 128-bit AES and the AES key with RSA; a 2048-bit RSA key can only encrypt a few hundred bytes and the operation is slow.

    Ref.

    Algorithms

    Contribution

    Limitation

    [11]

    AES, DES, RSA

    Analytical comparison to identify the best cipher for protecting

    information

    Theoretical study with no implementation in a cloud environment

    [12]

    Enhanced AES (14 rounds), DES, 3DES,

    Twofish

    More rounds to raise security and reduce processing time

    Comparison limited to small files

    [13]

    AES

    (cryptanalysis)

    Survey of side-channel, brute-force and algebraic attacks on

    AES

    Analytical survey, no proposed model

    [14]

    ECC, AES, SHA

    Equivalent security with smaller keys and lower energy

    No standard curve parameter; keys not

    protected in an ideal way

    [15]

    AES, ECC,

    blockchain

    High security level for the cloud storage infrastructure

    Large increase in execution time and

    resource use caused by the blockchain

    [16]

    AES, RSA

    Plaintext encrypted with AES and the AES key with RSA

    RSA is slow and limited to a few hundred bytes

    TABLE I. Studies Used to Select the Algorithms

    Two gaps follow from this review. Hybrids that protect the key with RSA inherit its cost, and hybrids that rely on a blockchain for integrity are too slow for real-time storage. The methodology below keeps the AES-ECC combination of [14] and [15], fixes the curve to a standard parameter, and replaces the plain hash and the ledger with keyed authentication tags and isolated key storage.

  3. PROPOSED METHODOLOGY

    1. Design Overview

      The methodology targets the three security goals of confidentiality, integrity and authenticated key exchange with one standardized primitive for each. AES provides confidentiality of the stored chunks. HMAC with SHA-256 provides integrity and detects any modification of a chunk. ECC on the prime256v1 curve protects and exchanges the keys; it was preferred to RSA because it gives equivalent security with far smaller keys and lighter operations [14], [17]. The design applies defense in depth: a file is split into chunks that are protected independently, and ciphertext, authentication tags and keys are held in three different places. Fig. 1 shows the storage and retrieval phases.

      Fig. 1. Storage and retrieval phases of the proposed methodology.

    2. Storage Phase

      The file is divided into chunks of fixed size (1 MB by default). For chunk i the client generates a random 256-bit encryption key KAES,i, a random 256-bit authentication key KHMAC,i and a 128-bit nonce. The chunk is encrypted with AES

      in cipher block chaining mode, and a tag is computed over the ciphertext, the nonce and the chunk index:

      Tagi = HMAC-SHA256(KHMAC,i, Ci || Noncei || i) (1)

      The original chunk is then replaced by its ciphertext. The keys of the chunk are bundled into a key package,

      18033-3 [22]. Neither AES nor ECC alone satisfies all of them, whereas the hybrid does, at a 128-bit security level.

      TABLE II. Standards Satisfied by the Hybrid and by Its Components

      Standard

      Scope

      Hybrid

      AES

      ECC

      NIST SP 800-38A [18]

      AES-CBC mode of operation

      Yes

      Yes

      No

      NIST SP 800-56A [19]

      ECC key establishment

      Yes

      No

      Yes

      RFC 2104 [21]

      HMAC construction

      Yes

      No

      No

      ISO/IEC 18033-3

      [22]

      International AES standard

      Yes

      Yes

      No

      KP = (Chunk_id, K , K , Nonce ) (2)

      i AES,i HMAC,i i

      which is encrypted with the ECC public key of the data owner. The wrapping key is obtained from an elliptic curve Diffie-Hellman exchange on prime256v1 followed by HKDF- SHA256, so no secret key ever travels over the channel. The ciphertext chunks are uploaded to the cloud store, the encrypted key packages are written to the key vault and the tags are kept in a database for later verification. Fig. 2 lists the procedure.

      Input: file F, owner public key PK; chunk size s = 1 MB 1: chunks Split(F, s); n number of chunks

      2: for i = 1 to n do

      3: KAES,i Random(256 bits); KHMAC,i Random(256 bits) 4: Noncei Random(128 bits)

      5: Ci AES-Encrypt(chunks[i], KAES,i, Noncei)

      6: Tagi HMAC-SHA256(KHMAC,i, Ci || Noncei || i)

      7: KPi (Chunk_id, KAES,i, KHMAC,i, Noncei, timestamp) 8: EKPi ECC-Encrypt(KPi, PK)

      9: Metai (file ID, i, n, timestamp) 10: end for

      11: Upload {Ci} to cloud storage; store {Tagi, Metai} in the tag database 12: Store {EKPi} and the package metadata in the key vault

Fig. 2. Pseudocode of the storage procedure.

  1. Retrieval and Verification

    On an access request the stored ciphertext and its tag are read, the key package is recovered from the vault with the ECC private key of the owner, and the tag is recomputed. If the two tags are equal the chunk is decrypted and delivered; otherwise the request is rejected and the owner is alerted. Retrieval needs no new random numbers and no new ephemeral keys: a single scalar multiplication on the curve recovers the AES key, after which the blocks are decrypted directly.

  2. Key Vault and Package Metadata

    The key vault is a dedicated secure store for secret keys and certificates, separate from the general database. It holds only the ECC-encrypted key packages, so an attacker who breaks into the vault still needs the private key, which never leaves the data owner. An attacker who breaks into the main storage server obtains ciphertext that cannot be decrypted. Tags are stored apart from the chunks for the same reason: a tag kept beside its chunk would let an attacker replace the chunk and recompute a matching tag. Each package carries metadata: package identifier, identifier of the original file, number of chunks, version of the cipher suite (AES-256 with ECC-256), creation date and state (active or invalid).

  3. Compliance With Standards

Every component follows a published standard, as listed in Table II. AES in cipher block chaining mode follows NIST SP 800-38A [18]; key establishment by elliptic curve Diffie- Hellman follows NIST SP 800-56A [19]; the use of HMAC follows NIST SP 800-107 [20] and its construction RFC 2104

[21]; and AES is also an international standard under ISO/IEC

  1. EXPERIMENTAL SETUP

    The methodology was implemented in Java inside CloudSim, a widely used simulator that models data centers, processors, memory, network and storage and reports the results of each run [23]. The AES and ECC implementations were both integrated in Java because the simulator does not support components written in different languages. Table III gives the simulation parameters.

    TABLE III. Simulation Parameters

    Parameter

    Value

    Simulator

    CloudSim 3.0.3

    Programming language

    Java (JDK 11)

    Elliptic curve

    prime256v1

    AES key size

    128 192 and 256 bits

    ECC key size

    256 bits

    Integrity function

    HMAC-SHA256

    Cloud files

    1 MB to 1000 MB

    Small payloads

    64 to 288 bytes

    Host

    Windows 10, Core i5, 8 GB RAM, 256 GB SSD and 1 TB hard disk

    Three comparisons were made: with the ECC-AES-SHA hybrid of [14] and the single ciphers AES, 3DES and Blowfish reported with it, at a 256-bit key on a 5 MB file; with the blockchain-enabled AES-ECC hybrid of [15] on files of 30, 60 and 100 MB; and with the RSA-AES hybrid of [16] on payloads of 64 to 288 bytes. The scheme of [14] checks integrity with plain SHA-256, uses a generic curve with no accelerated standard parameter and runs as a web application inside the simulator; the system of [15] likewise uses a generic 256-bit curve and a web application connected to a local server. The metrics are encryption and decryption time, power consumption, throughput, size of the encrypted file and avalanche effect. The avalanche effect is the share of ciphertext bits that change when one bit of the key or of the plaintext is flipped,

    AE = (Nflipped / Ntotal) × 100% (3) with an ideal value of 50% and an accepted range of 45%

    to 55%. Throughput is the amount of data processed per unit of

    time,

    TP = S / T (4)

    where S is the data size in MB and T the execution time in seconds. The energy of one operation is obtained from the measured average power P and the execution time as E = P × T.

  2. RESULTS AND DISCUSSION

    1. Comparison With the ECC-AES-SHA Hybrid and Single Ciphers

      Table IV and Fig. 3 compare the proposed hybrid with the ECC-AES-SHA hybrid of [14] and with AES, 3DES and Blowfish at a 256-bit key. The proposed methodology completes encryption, authentication and key exchange in 4.99 ms, against 8.33 ms for the hybrid of [14]: a saving of 3.34 ms, or 40.10%, which is a speed-up of about 1.67 times. It also draws 3.95 W instead of 4.15 W. The gain comes from the accelerated standard curve prime256v1 and from processing independent chunks in memory, whereas [14] relies on a generic, unoptimized curve equation. Compared with AES alone, the elliptic curve and authentication layers add about 2 ms and 0.87 W, and the hybrid remains faster than 3DES and lighter than 3DES and Blowfish. This is the price of removing the key-sharing weakness of a symmetric cipher used alone.

      TABLE IV. Encryption Time and Power Consumption at a 256-Bit Key (5 MB File)

      Scheme

      Encryption time (ms)

      Power (W)

      Proposed hybrid

      4.99

      3.95

      ECC-AES-SHA hybrid [14]

      8.33

      4.15

      AES

      3.00

      3.08

      3DES

      8.00

      4.65

      Blowfish

      4.00

      4.25

      Fig. 3. Encryption time and power of the proposed hybrid, the ECC-AES- SHA hybrid and single ciphers at a 256-bit key.

      Table V gives the avalanche effect in a 128-bit block. A one-bit change of the key flips 67 bits (52.34%) in the proposed hybrid and 63 bits (49.22%) in [14]; a one-bit change of the plaintext flips 65 bits (50.78%) against 62 bits (48.44%). The higher key sensitivity follows from scalar multiplication on the standard curve combined with an independent key for every chunk, where [14] derives its key directly through SHA- 256.

      TABLE V. Avalanche Effect Compared With the ECC-AES-SHA Hybrid [14]

      Test

      Hybrid

      [14]

      1-bit key change: bits flipped

      67 of 128

      63 of 128

      1-bit key change: avalanche effect

      52.34%

      49.22%

      1-bit plaintext change: bits flipped

      65 of 128

      62 of 128

      1-bit plaintext change: avalanche effect

      50.78%

      48.44%

    2. Comparison With the Blockchain-Enabled Hybrid

      Table VI compares the proposed methodology with the AES-ECC-blockchain system of [15]. The reference system carries a nearly constant penalty of about 1000 ms at every file size, the cost of the consensus and ledger operations. The proposed system encrypts a 30 MB file in 120 ms instead of 1130 ms, a reduction of 89.38%; the reduction is 80.95% at 60

      MB and 71.43% at 100 MB (Fig. 4). Decryption is 70%, 66% and 68.89% faster at the three sizes, about two thirds on average, because the reference system must query the ledger and validate the fingerprints and the smart-contract permissions before any block is released. In the proposed system decryption is also 25% to 30% faster than encryption of the same file, since no random values or ephemeral keys are generated on retrieval.

      TABLE VI. Performance Compared With the Blockchain-Enabled Hybrid [15]

      Metric

      30 MB

      60 MB

      100 MB

      Hybrid

      [15]

      Hybrid

      [15]

      Hybrid

      [15]

      Encryption time (ms)

      120

      1130

      240

      1260

      400

      1400

      Decryption time (ms)

      90

      300

      170

      500

      280

      900

      Power (W)

      17

      24

      26

      35

      40

      47

      Energy (J)a

      2.04

      27.12

      6.24

      44.10

      16.00

      65.80

      Storage overhead (KB)

      4

      9

      8

      13

      13

      18

      a. Derived as average power multiplied by encryption time.

      Fig. 4. Encryption and decryption time compared with the blockchain- enabled hybrid.

      The reference system draws about 7.67 W more on average, so the proposed system saves 29.17%, 25.71% and 14.89% of the power at 30, 60 and 100 MB. Because it also finishes much sooner, the energy spent per encryption falls by more than 75% at every size (Fig. 5), which suits green cloud computing and

      battery-powered edge devices: the processor and the network interface are freed from mining and ledger synchronization.

      Fig. 5. Average power and energy per encryption compared with the blockchain-enabled hybrid.

      The encrypted file grows by 4, 8 and 13 KB for the three sizes, against 9, 13 and 18 KB in the reference system (Fig. 6). Removing the mandatory block headers, about 5 KB per file, cuts the added storage by 55.56%, 38.46% and 27.78%.

      Fig. 6. Size added to the encrypted file compared with the blockchain- enabled hybrid.

      Table VII collects the remaining indicators. Cryptographic throughput reaches 250 MB/s for encryption and 350 MB/s for decryption, 13.64% and 9.38% above the 220 MB/s and 320 MB/s of the cryptographic stage of the reference system, taken without its ledger operations. The avalanche effect is 52.34% for a one-bit key change and 50.78% for a one-bit plaintext change in a 128-bit block, against 50.78% and 49.22%. A one- bit change in the secret scalar moves the resulting curve point completely, and the key derivation function and the AES key expansion then spread the change over more than half of the ciphertext bits. Leaving out the blockchain therefore takes nothing from the statistical strength of the ciphertext against linear and differential cryptanalysis. Integrity is verified in about 0.05 ms, against 200 to 500 ms through the ledger, and verification works without a network connection. The blockchain keeps one advantage: it prevents tampering by the cloud provider itself, which keyed tags alone do not.

      TABLE VII. Further Indicators Compared With the Blockchain- Enabled Hybrid [15]

      Indicator

      Hybrid

      [15]

      Avalanche effect, 1-bit key change

      52.34% (67 of 128

      bits)

      50.78% (65 of 128

      bits)

      Avalanche effect, 1-bit plaintext change

      50.78% (65 of 128

      bits)

      49.22% (63 of 128

      bits)

      Cryptographic throughput, encryption

      250 MB/s

      220 MB/s

      Cryptographic throughput, decryption

      350 MB/s

      320 MB/s

      Integrity verification time

      0.05 ms

      200 to 500 ms

      Detects modification

      Yes

      Yes

      Prevents tampering by the provider

      No

      Yes

      Additional storage cost

      Very low

      Very high

      Suitable for large files

      Yes

      Limited by block size

      Verification without network access

      Yes

      No

    3. Comparison With the RSA-AES Hybrid

      Table VIII and Fig. 7 compare the methodology with the RSA-AES hybrid of [16] on small payloads, the typical load of Internet of Things devices. Encryption takes 3.5 ms instead of about 1047 ms, an improvement of 99.66%, and decryption 1.2 ms instead of about 54 ms, an improvement of 97.78%. Throughput is roughly 300 times higher at every size. The time of the proposed system does not change with the payload because AES processes 16-byte blocks: 64 bytes are 4 blocks and 288 bytes are 18 blocks, which take microseconds. The measured 3.5 ms is in practice the fixed cost of establishing the session and deriving the shared key by one scalar multiplication on prime256v1.

      TABLE VIII. Results for Small Payloads Compared With the RSA- AES Hybrid [16]

      Size

      Encryption (ms)

      Decryption (ms)

      Throughput (KB/s)

      (bytes)

      Hybrid

      [16]

      Hybrid

      [16]

      Hybrid

      [16]

      64

      3.5

      1047.07

      1.2

      53.90

      17.4

      0.058

      96

      3.5

      1047.08

      1.2

      53.97

      26.1

      0.087

      144

      3.5

      1047.10

      1.2

      53.99

      39.1

      0.131

      176

      3.5

      1047.20

      1.2

      53.10

      47.8

      0.160

      208

      3.5

      1047.30

      1.2

      53.18

      56.3

      0.189

      240

      3.5

      1047.40

      1.2

      53.25

      65.0

      0.218

      288

      3.5

      1047.50

      1.2

      53.44

      77.9

      0.262

      Fig. 7. Encryption and decryption time for small payloads compared with the RSA-AES hybrid.

    4. Security Analysis

      Table IX lists the attacks on files stored in the cloud that each scheme withstands. The hybrid covers unauthorized access, storage misconfiguration, hash collision, unauthorized modification, side-channel and cloud man-in-the-middle attacks, a set that neither AES nor ECC covers in full alone, and it resists replay under conditions. Two cases remain open: recovery of deleted data and a malicious internal administrator, both of which the blockchain-based system handles.

      Attack

      Hybrid

      AES

      ECC

      [15]

      Unauthorized access

      Yes

      Yes

      No

      Yes

      Recovery of deleted data

      No

      No

      No

      Yes

      Storage misconfiguration

      Yes

      Yes

      No

      Yes

      Malicious internal administrator

      No

      No

      No

      Yes

      Replay

      Partial

      Partial

      No

      No

      Hash collision

      Yes

      No

      No

      Yes

      Unauthorized modification

      Yes

      Yes

      No

      Yes

      Side channel

      Yes

      No

      Yes

      Yes

      Cloud man-in-the-middle

      Yes

      Yes

      No

      Yes

      TABLE IX. Attacks Withstood by Each Scheme

      systems, confidential government documents, personal data on mobile phones, and cloud-connected Internet of Things devices with limited resources. It is not intended for hard real-time applications such as autonomous driving, robotic surgery, competitive online gaming and drones, nor for decentralized trustless environments that require non-repudiation.

  3. CONCLUSION

This paper presented an authenticated hybrid methodology for cloud data at rest that combines AES for confidentiality, ECC on prime256v1 for key protection and HMAC-SHA256 for integrity, with chunk-level isolation and a key vault separated from the data servers. In simulation it encrypted 40.10% faster than a recent ECC-AES-SHA hybrid and added about 2 ms to AES alone while removing the key-sharing weakness; it cut encryption time by up to 89.38%, decryption time by up to 70% and energy by more than 75% relative to a blockchain-enabled hybrid; it reduced the added storage by 27.78% to 55.56%; and it encrypted small payloads more than 99% faster than an RSA-AES hybrid, with an avalanche effect above 50% throughout. Integrity of cloud data can thus be verified in a fraction of a millisecond without the cost of a ledger.

The methodology has limits. It costs more time and power than a symmetric cipher alone, and it does not protect against a malicious internal administrator or the recovery of deleted data. Future work will combine the design with post-quantum key encapsulation such as ML-KEM alongside prime256v1, study search and indexing over encrypted data with homomorphic encryption, and offload the curve multiplications and AES rounds to graphics processors and FPGAs for very large data sets.

  1. Recommended Thresholds and Scope

Chunking multiplies the number of keys and tags, so its parameters must be bounded. A chunk should not be smaller than 1 MB, a file should not be split into more than 10 000 chunks, and files above 10 GB need additional optimization of key generation. Table X gives the recommended settings by file size.

TABLE X. Recommended Chunking by File Size

File size

Split

Chunk size

Chunks

Assessment

Below 1 MB

No

Whole file

1

Avoids splitting cost

1 MB to 100 MB

Yes

1 MB

1 to 100

Ideal

100 MB to 1 GB

Yes

5 MB

20 to 200

Good

1 GB to 10 GB

Yes

10 MB

100 to 1000

Acceptable

With a constant 3.5 ms for small packets and 250 MB/s for large files, the methodology suits electronic health records and medical images, real-time financial and digital banking

REFERENCES

  1. F. S. Al-Anzi, S. K. Yadav, and J. Soni, Cloud computing: Security model comprising governance, risk management and compliance, Indira Gandhi Delhi Technical University, Delhi, India, 2015.

  2. F. Shahzad, State-of-the-art survey on cloud computing security challenges, approaches and solutions, in Proc. 6th Int. Symp. Applications of Ad hoc and Sensor Networks, 2014, pp. 357362.

  3. S. K. Saroj, S. K. Chauhan, A. K. Sharma, and S. Vats, Threshold cryptography based data security in cloud computing, in Proc. IEEE Int. Conf. Computational Intelligence and Communication Technology, 2015.

  4. P. Harikrishna and A. Amuthan, A survey of testing as a service in cloud computing, in Proc. Int. Conf. Computer Communication and Informatics, Coimbatore, India, Jan. 2016.

  5. P. M. S. S. Chandu and D. Kata, Integrating and enhancing the quality of services in cloud computing with software testing, Sathyabama University, Chennai, India, 2016.

  6. H. Ziglari and S. Yahya, Deployment models: Enhancing security in cloud computing environment, in Proc. 22nd Asia-Pacific Conf. Communications (APCC), 2016.

  7. R. Salloum, A proposed study to improve the security of data stored in cloud computing, M.S. thesis, Univ. of Aleppo, Aleppo, Syria, 2017 (in Arabic).

  8. N. Singh and N. K. Singh, Information security in cloud computing using encryption techniques, Int. J. Scientific and Engineering Research, vol. 5, no. 4, 2019.

  9. P. Rewagad and Y. Pawar, Use of digital signature with Diffie Hellman key exchange and AES encryption algorithm to enhance data security in cloud computing, in Proc. Int. Conf. Communication Systems and Network Technologies (CSNT), 2013, pp. 437439.

  10. A. I. Tahirkheli et al., A survey on modern cloud computing security over smart city networks: Threats, vulnerabilities, consequences,

    countermeasures, and challenges, Electronics, vol. 10, no. 15, art. 1811, 2021.

  11. R. Sood and H. Kaur, A literature review on RSA, DES and AES encryption algorithms, in Emerging Trends in Engineering and Management, V. Dhiman and P. Dhand, Eds. India: SCRS, 2023, pp. 5763.

  12. N. A. M. Ariffin and A. Y. A. Ashawesh, Enhanced AES algorithm based on 14 rounds in securing data and minimizing processing time, J. Physics: Conf. Series, vol. 1793, art. 012066, 2021.

  13. H. Zodpe and A. Shaikh, A survey on various cryptanalytic attacks on the AES algorithm, Int. J. Next-Generation Computing, vol. 12, no. 2,

    pp. 115123, April 2021.

  14. P. Selvi and S. Sakthivel, A hybrid ECC-AES encryption framework for secure and efficient cloud-based data protection, Scientific Reports, vol. 15, art. 30867, 2025.

  15. M. Bakro, S. K. Bisoy, A. K. Patel, and M. A. Naal, Hybrid blockchain-enabled security in cloud storage infrastructure using ECC and AES algorithms, in Blockchain Based Internet of Things. Singapore: Springer, 2022.

  16. R. Akter, M. A. R. Khan, F. Rahman, S. J. Soheli, and N. J. Suha, RSA and AES based hybrid encryption technique for enhancing data security in cloud computing, Int. J. Computational and Applied Mathematics and Computer Science, vol. 3, pp. 6071, 2023.

  17. J. Joy and S. Devaraju, Securing cloud data through efficient using ECC algorithm, Educational Administration: Theory and Practice, vol. 30, no. 5, pp. 94219429, 2024.

  18. M. Dworkin, Recommendation for block cipher modes of operation: Methods and techniques, NIST Special Publication 800-38A, December 2001.

  19. E. Barker, L. Chen, A. Roginsky, A. Vassilev, and R. Davis, Recommendation for pair-wise key-establishment schemes using discrete logarithm cryptography, NIST Special Publication 800-56A, Rev. 3, April 2018.

  20. Q. Dang, Recommendation for applications using approved hash algorithms, NIST Special Publication 800-107, Rev. 1, August 2012.

  21. H. Krawczyk, M. Bellare, and R. Canetti, HMAC: Keyed-hashing for message authentication, IETF RFC 2104, February 1997.

  22. Information Technology – Security Techniques – Encryption Algorithms

    – Part 3: Block Ciphers, ISO/IEC 18033-3, 2010.

  23. K. Bahwaireth, L. Tawalbeh, E. Benkhelifa, Y. Jararweh, and M. A. Tawalbeh, Experimental comparison of simulation tools for efficient cloud and mobile cloud computing applications, EURASIP J. Information Security, vol. 2016, art. 15, 2016.