Cloud security is a major concern in the field of computer science. The cloud is not only responsible for storing data but also offers services such as Infrastructure as a Service (IaaS) and Platform as a Service (PaaS), which provide both platforms and complete infrastructure solutions. Additionally, cloud services enable organizations to access resources both internally and externally through private, public, and hybrid cloud models (Pandey and Sharma 2017).

Over the past few years, there has been a significant increase in the volume of cloud computing utilization, particularly in the use of public cloud. This growth has increased the potentially vulnerable amount of sensitive data. More and more security risks and issues arise, particularly in public cloud environments due to its multi-tenancy and virtualization features (Wang et al., 2018).

One of the main security problems the adoption of cloud computing raises is protecting sensitive data in public cloud infrastructure. The issue of secure data storage in cloud computing has become a significant concern as enterprises find it difficult to ensure the availability, confidentiality, and integrity of their information assets when entrusting them to outside cloud service providers (Stergiou et al., 2020).

Cryptography is a crucial element in achieving this goal (Wang et al., 2020). Using cryptographic technology, which involves the encryption and decryption process, it is possible to ensure the confidentiality of data transmitted over a network (Hossain et al., 2019). This paper presents the improved, secure, and integrated encryption model, SHIELD to address the intricate security needs of public cloud computing environments regarding sensitive data storage.

Riyaz Fathima Abdul and Saravanan Arumugam (2025) address the challenge of strengthening data security against cryptanalytic attacks while preserving data integrity by proposing a hybrid encryption scheme combining RSA, ChaCha20-Poly1305, and a shuffling mechanism. Although effective in enhancing robustness, the model’s high computational overhead results in slow execution and low throughput, making it less suitable for real-time or resource-constrained environments, and it lacks post-quantum protection as well as extensive real-world testing.

Shukla et al. (2024) address the need for enhanced security in Wireless Sensor Networks (WSNs) by proposing the Three Phase Hybrid Cryptographic (TPHC) algorithm, which combines AES, DES, and a modified RSA to provide confidentiality and authentication, alongside a custom hashing algorithm to ensure data integrity through one-wayness, collision resistance, second pre-image resistance, and a strong avalanche effect. The TPHC outperforms existing WSN algorithms in computation time, encrypted and digested data size, and energy usage. However, its use of three encryption algorithms introduces computational complexity, the modified RSA remains resource-intensive for low-power nodes, and the model lacks evaluation for scalability, diverse real-world deployments, and post-quantum readiness.

Almalawi et al. (2024) address data security in decentralized Edge AI networks with a hybrid mechanism combining Ant Lion Optimization (ALO) and Diffie–Hellman-based Twofish cryptography (DHT), enhanced by an Autoencoder for malicious data detection. Achieving 99.45% accuracy, 2 s execution time, 0.8 s delay, and 3.2 mJ energy use, the method offers strong security but suffers from computational overhead, high cost, and limited scalability, requiring optimization for resource-constrained and dynamic environments.

Anjana (2024) addresses the challenge of securing data confidentiality and integrity in cloud computing by proposing an enhanced three-layer cryptographic algorithm that combines AES and ChaCha20 for hybrid encryption with SHA-3 hashing for message integrity verification. The study evaluates the scheme’s encryption strength, computational efficiency, and resistance to attacks, showing its suitability for mitigating common cloud security threats while offering insights to inform best practices in cloud data protection. However, it does not address performance trade-offs in large-scale deployments, lacks real-world implementation testing, introduces potential computational complexity from its multi-layer design, and omits consideration of post-quantum readiness.

Rath et al. (2024) presented a hybrid model for file encryption that combined a modified AES with masking and RSA algorithms. This dynamic AES-RSA algorithm, featuring masked key management, offered enhanced security and resistance against cryptographic attacks.

In the proposed encryption system, Sabitha et al. (2023) used a hybrid methodology that combined elliptical and parametric cryptography techniques. Authorization, key generation, and data encryption are the three separate security checkpoints that make up the system.

Chatterjee et al. (2023) used steganography and cryptography techniques to solve cloud storage problems. Block-per-block encryption was done using AES, RC6, Blowfish, and BRA. Sensitive data was safeguarded using the LSB method. The SHA1 hash algorithm was used to validate data security.

Ranganatha and Sujatha (2023), in their proposed method, created keys by utilising a lightweight Edwards curve. The created private keys were changed using the user's Identity-Based Encryption. The Advanced Encryption Standard (AES) encryption process was accelerated by using the suggested key reduction technique to further shorten the keys. Diffie-Hellman key exchange was used to exchange the public keys.

Using the hybrid approach, William et al. (2022) combined a hash function, an asymmetric algorithm (ECC), and a symmetric algorithm (AES) (SHA256). In this instance, digital data was encrypted using the mathematical formula SHA-256. It was employed to guarantee the integrity of the data. The proposed hybrid solution claimed to be comparable to an existing method that encrypts text and images using the AES algorithm. In comparison with the previously discussed approaches, the proposed strategy was more efficient since it was more effective at text encryption. Future advances could result in a reduction of the time needed to encrypt and decode a photograph.

A multilevel approach with three encryption levels was suggested by Binita and Blessy (2022). Data Encryption Standard (DES) was used at the first level, mixed transposition was used at the second level, and Blowfish was used at the third level to achieve this. Since deciphering an algorithm took time, data security increased with the number of encryption levels.

A. Algorithms Used in our Experiment.

(i) Advanced Encryption Standard (AES). The Advanced Encryption Standard (AES) algorithm, created in 1998 by Joan Daemen and Vincent Rijmen, is a symmetric key block cipher. It supports data and key lengths of 128, 192, and 256 bits. AES operates on a 128-bit data length, organized into four operational blocks as an array of bytes arranged in a 4×4 matrix known as the state. These blocks undergo various transformations during rounds. The number of rounds (N) for full encryption varies: 10 rounds for a key length of 128, 12 rounds for 192, and 14 rounds for 256. Each AES round employs permutation and substitution networks and is adaptable for hardware and software implementations (Priyadarshini et al., 2016).

(ii) RSA Algorithm. Founded in 1977, RSA is a public key cryptosystem, an asymmetric cryptographic algorithm named after its founders Rivest, Shamir, and Adelman. It generates two keys: a public key for encryption and a private key for decryption. The RSA algorithm consists of three steps: key generation, encryption, and decryption. Key generation creates the keys used for data encryption and decryption, encryption transforms plaintext into ciphertext, and decryption reverses this process. RSA relies on the factoring problem, finding the product of two large prime numbers. Key sizes range from 1024 to 4096 bits (Priyadarshini et al., 2016).

METHODS 

This study adopted the mixed methodology that comprises SSADM and the experimental method. The Structured Systems Analysis and Design Methodology was used to define, ideate, and prototype our solution, while the experimental method involved the investigation and comparison of the variable's impacts based on various parameters to obtain solutions and analyse findings based on experiment outcomes.

A. Evaluation Parameters 

It's critical to assess the encryption algorithms to make sure they can handle a range of data kinds, workloads, and usage scenarios with efficiency, scalability, and practicality. The behaviour and results of the algorithm were measured in this study using empirical approaches that involved real-world experiments or simulations, giving researchers direct feedback and insights into the functioning and performance of the system. Python 3.12 was utilized for implementation, enabling the testing of different data types and sizes for encryption and decryption techniques. The following are a few metrics that are used to evaluate performance:

(i) Encryption Time. Encryption time is the time that an encryption algorithm takes to produce a cipher text from a plain text. We measured the encryption time in seconds in our experiment. The system's performance is affected by the encryption time. For the system to be fast and responsive, the encryption time must be less.

(ii) Decryption Time. The time taken to produce plain text from cipher text is called decryption time. To improve system responsiveness and speed, the decryption time should be less. System performance is impacted by the decryption time. We measured the decryption time in seconds in our experiment.

(iii) Avalanche Effect. This is a desirable feature of cryptographic algorithms, where a change in a single bit of ciphertext should lead to changes in multiple bits of the plaintext message, or vice versa. To calculate it: Avalanche Effect = (Number of Changed Bits in Ciphertext) / (Number of Bits in Ciphertext). A robust cipher should always achieve an avalanche effect greater than 50%.

In this section, we discussed the results obtained based on three parameters.

Experiment/Case Study 1. The encryption time for each of the different techniques is shown in Table 1.

Table 1: Encryption time result (seconds).

Experiment / Case Study 2. 

Table 2 displays the decryption time for all the various techniques.

Table 2: Decryption time result (seconds).

Experiment / Case Study 3.

Table 3: Avalanche Effect (%).

Analysis of the Results. Table 4 presents a comparison of the performance evaluation among all the encryption methods used in the experiments. The result analysis is based on the average encryption time, average decryption time and average avalanche effect. The files used in the tests vary from 23kb to 100mb. Graphs of both the algorithms along with their tables are shown below.

Table 4: Performance Comparison among all the encryption methods for encryption and decryption time.

(i) Analysis and evaluation based on Encryption Time. The time taken to execute SHIELD encryption is shown in a graphic form in Fig. 1. The encryption time is displayed on the Y axis in seconds, and the X axis displays algorithms. The encryption time for AES was 46.138, that of RSA was 78.409, and SHIELD has an encryption time of 44.339. SHIELD has a lower encryption time compared to both AES and RSA. So, in terms of encryption time, SHIELD shows approximately a 3.90% improvement compared to AES and approximately a 43.45% improvement compared to RSA. Therefore, SHIELD shows an average encryption time improvement of approximately 28.8% over the combined average encryption times of AES and RSA. This single percentage represents the overall improvement in encryption time achieved by SHIELD compared to using AES and RSA separately. 

Fig. 1. Encryption time of existing methods and SHIELD model.

(ii) Analysis and evaluation based on Decryption Time. The time taken to execute SHIELD decryption is shown in a graphic form in Figure 2. The decryption time is displayed on the Y axis in seconds, and the X axis displays the algorithms. The decryption time for AES was 44.807, that of RSA was 224.725, and SHIELD has a decryption time of 32.043 indicating that SHIELD has a significantly lower decryption time compared to both AES and RSA. So, in terms of decryption time, SHIELD shows approximately a 28.49% improvement compared to AES and approximately an 85.76% improvement compared to RSA. Therefore, SHIELD shows an average decryption time improvement of approximately 76.2% over the combined average decryption times of AES and RSA. This single percentage represents the overall improvement in decryption time achieved by SHIELD compared to using AES and RSA separately.

Fig. 2. Decryption time of existing methods and SHIELD model.

(iii) Analysis and Evaluation Based on Average Avalanche effect (%). A minimum of 50% avalanche effect is required to classify the algorithm as very secure. The avalanche effect as shown in a graphical representation in figure 3 indicates that all the three methods (AES, RSA, SHIELD) show very similar avalanche effects for encryption, with SHIELD slightly higher than AES and RSA, indicating a strong and comparable level of security in terms of the avalanche effect. The model achieved an Avalanche effect exceeding 50% which shows it achieved the desired level of unpredictability in the output ciphertext to provide strong encryption as compared to AES and RSA.

Fig. 3. Average Avalanche effect (%) of existing methods and SHIELD model.

Future research should focus on testing the SHIELD model with major public cloud service providers such as AWS, Azure, and Google Cloud Platform to assess compatibility, performance, and cross-platform effectiveness. Security and privacy enhancements should be pursued through penetration testing, vulnerability analysis, and compliance checks. Performance evaluation should extend to real-world user scenarios, larger datasets, and advanced simulations on higher system configurations to determine scalability and efficiency. Comparative testing against established benchmarks and across different hardware environments will provide deeper insights into its practical applicability and potential for optimization.