Contents Introduction
Cloud computing has transformed the information technology landscape by offering scalable, flexible, and cost-effective solutions for data storage, management, and sharing. Operating across multiple layers—Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS)—cloud platforms provide a wide range of services, from basic infrastructure management to sophisticated applications delivered over the internet. The adoption of cloud computing has permeated various sectors, including healthcare, finance, and entertainment. For example, NETFLIX utilizes cloud technology to stream services to millions, highlighting cloud computing significance in everyday applications. The rapid expansion of this technology has led to industry giants like Google, Meta, and Apple dominating the market with their scalable solutions, while also creating opportunities for smaller providers to enter the competitive landscape [1]. As the cloud user base expands, new opportunities arise for smaller cloud providers to enter the market, fostering a more competitive and innovative ecosystem. This expansion, fueled by the cloud sustainable and accessible nature, continues to attract more users globally. As the cloud user base expands, it creates opportunities for smaller cloud providers to emerge. In cloud services, job allocation typically involves an exchange between the service provider and the service seeker, as depicted in Figure 1. This figure presents a focused view of a cloud infrastructure segment, emphasizing the mechanisms of job provisioning and virtual resource allocation. It illustrates the key components and interactions involved in managing user requests effectively while optimizing resource distribution within the cloud environment. The depiction serves to highlight the integration of algorithmic strategies designed to enhance both privacy preservation and job allocation efficiency, thereby providing a clear understanding of the proposed frameworks operational context.However, this proliferation of cloud services also introduces significant challenges, particularly in the areas of privacy, trust, and resource management. The ability to address these challenges is crucial for maintaining the security and efficiency of cloud-based systems, which are now integral to the functioning of modern society. Cloud computing has revolutionized information technology by providing scalable, flexible, and cost-effective solutions for data storage, management, and sharing. While its benefits are substantial, significant challenges related to privacy, trust, and resource management remain. Addressing these issues is essential for maintaining the security and efficiency of cloud-based systems. Recent advancements have focused on developing privacy-preserving techniques. For example, [2] proposed a hybrid approach that combines k-anonymity with differential privacy to enhance data protection, while [3] integrated homomorphic encryption with k-anonymity for secure computations on encrypted data. In terms of trust management, [4] introduced a framework that combines reputation systems with blockchain technology for secure trust assessment. Reference [5] developed a trust model using machine learning to evaluate trustworthiness based on historical interactions. For resource management, [6] introduced a reinforcement learning approach for dynamic provisioning, adjusting allocations based on real-time patterns, while [7] proposed a framework that integrates predictive analytics to enhance performance. These developments underscore the need for adaptive, data-driven approaches in cloud environments.
A. Need and Research Motivation
The proliferation of cloud computing has transformed how organizations and individuals manage, store, and share data. With its inherent flexibility and scalability, cloud technology has become essential across various industries, including healthcare, finance, and education. However, as more entities migrate to cloud environments, significant challenges regarding privacy, trust, and resource management have emerged, necessitating a comprehensive research approach. Privacy concerns are paramount as sensitive data stored in cloud environments is susceptible to unauthorized access and potential breaches. Current privacy-preserving methods often focus on protecting either data providers or seekers, leading to vulnerabilities. To address this, there is a pressing need for comprehensive solutions that ensure the anonymity of all parties involved, such as the application of k-anonymity. Trust is another vital element in cloud interactions. Users require confidence that their data will be managed securely and responsibly by service providers. However, many existing trust management frameworks fall short, failing to adapt to the dynamic nature of cloud environments. The development of advanced algorithms for trust assessment is essential to foster a reliable and transparent cloud ecosystem. Resource allocation in cloud computing also presents significant challenges, particularly due to fluctuating workloads and varying demands. Inefficient allocation can result in wasted resources and increased operational costs. A more dynamic and time-sensitive approach is necessary to optimize resource distribution and enhance overall system performance.
This research seeks to tackle these interconnected challenges by proposing a novel hybrid framework that enhances privacy through k-anonymity, refines trust generation mechanisms, and optimizes resource allocation strategies. By addressing these issues comprehensively, the study aims to improve the security, efficiency, and user confidence in cloud services, ultimately contributing to the advancement of secure cloud-based systems.
B. Problem Statement
The existing study of research in cloud-based data sharing highlights several significant challenges that the proposed algorithm seeks to address comprehensively. A primary concern is the inadequacy of current privacy-preserving techniques, which often focus solely on protecting either data producers or data seekers. This narrow approach increases the risk of re-identification and data misuse, as it leaves one party vulnerable. The proposed algorithm employs k-anonymity, ensuring that all participants in data transactions are anonymized, thus fortifying privacy for both data producers and seekers. In addition, many contemporary trust management frameworks rely on static assessment methods that fail to account for the dynamic nature of cloud environments. These rigid approaches lead to unreliable trust evaluations, eroding confidence in cloud services. To combat this, the proposed algorithm integrates an enhanced Firefly algorithm for trust generation, which adapts to real-time data interactions and continuously refines trust metrics based on user behavior and historical interactions. This adaptability fosters a more secure and trustworthy environment for data sharing. Resource allocation poses another critical challenge, with traditional methods often struggling to efficiently distribute resources in response to fluctuating cloud workloads. The proposed solution introduces a T-MBFD algorithm that dynamically adjusts resource allocations based on real-time demand and usage patterns, thus optimizing resource utilization and reducing operational costs.
Furthermore, previous approaches frequently isolate privacy and trust issues, failing to recognize their interconnectedness and the necessity for a holistic solution. The proposed algorithm addresses this by integrating privacy preservation, trust evaluation, and resource allocation into a unified framework, promoting seamless interactions and enhancing the overall security of cloud data sharing. Lastly, many existing studies do not leverage modern machine learning techniques to enhance the effectiveness of privacy and trust mechanisms. The proposed algorithm employs advanced machine learning models to analyze data patterns and improve k-anonymity measures, ensuring scalability and robustness when dealing with large datasets. Additionally, the framework incorporates comprehensive evaluation metrics, including execution time and computational efficiency, which provides a clearer understanding of performance across various dimensions. The proposed algorithm offers a robust solution to the multifaceted challenges present in cloud-based data sharing by ensuring comprehensive privacy protection, fostering reliable trust assessments, optimizing resource allocation, and leveraging modern machine learning techniques for enhanced performance and scalability.
C. Research Contribution and Novelty
This research significantly advances the field of cloud computing by providing a comprehensive framework that addresses the critical challenges of privacy, trust, and resource management in cloud-based data sharing. The key contributions and novel elements of this work are outlined below:
Innovative Hybrid Paradigm: This study introduces a groundbreaking hybrid framework that uniquely integrates k-anonymity with advanced computational algorithms, thereby establishing a robust mechanism for privacy protection and trust enhancement. By anonymizing both data producers and seekers, the proposed methodology effectively mitigates the risks associated with data re-identification, ensuring confidentiality in cloud interactions.
Dynamic Resource Allocation with T-MBFD: The research presents the T-MBFD algorithm, a pioneering solution designed for dynamic resource allocation in cloud environments. This algorithm intelligently adapts to real-time workload fluctuations, optimizing resource distribution and significantly reducing wastage. This innovation addresses a critical gap in existing methodologies that often fail to consider the dynamic nature of cloud resource demands.
Robust Trust Generation Mechanism: The study proposes an advanced trust evaluation framework that synergistically combines the optimized Levenberg algorithm with the Firefly Algorithm. This novel integration enhances the precision and reliability of trust assessments for cloud service providers, empowering service seekers to make informed decisions while reinforcing the integrity of cloud interactions.
Empirical Validation and Performance Metrics: Rigorous simulations conducted on a dataset comprising 95,000 records substantiate the frameworks robustness and scalability. The exceptional performance metrics achieved—average accuracy of 96.416%, F-measure of 0.976, precision of 0.958, and recall of 0.989—demonstrate the frameworks superiority in enhancing privacy protection and trust generation compared to contemporary approaches.
Comprehensive Benchmarking: The research provides a systematic comparative analysis against existing models, consistently illustrating the proposed frameworks enhanced capabilities in safeguarding privacy, assessing trust, and optimizing resource management. This benchmarking underscores the frameworks novelty and establishes it as a leading solution in the realm of cloud-based data sharing.
Contribution to Academic and Practical Knowledge: By addressing the intertwined challenges of privacy, trust, and resource management, this research enriches the academic discourse on secure cloud computing. The findings not only contribute valuable insights for researchers but also offer actionable solutions for practitioners seeking to implement secure and efficient cloud environments.
Through these contributions, this research not only enhances theoretical understanding but also lays a practical foundation for future advancements in secure cloud data sharing, ultimately fostering a more trustworthy and efficient cloud ecosystem.
D. Paper Organization
The remainder of this paper is structured as follows: Section II reviews related work in the field, focusing on existing privacy-preserving techniques and trust management frameworks. Section III details the methodology employed in this research, outlining the proposed hybrid model. Section IV presents the experimental setup and results, and Section V concludes the paper, discussing implications and future research directions. At the end of the paper is a list of references mentioned.
Literature Review
This section examines recent advancements in privacy and trust mechanisms within cloud computing, focusing on techniques such as machine learning-based privacy protection, blockchain-integrated security frameworks, and optimization algorithms for resource management. The aim is to identify current trends, challenges, and potential areas for future research. In paper [8] introduced a cryptographic game theory model enhancing security in cloud environments, outperforming traditional frameworks by integrating probabilistic elements. Reference [9] proposed a clustering technique for privacy protection in social media, achieving k-anonymity and superior performance regarding processing time and information loss minimization. Reference [10] introduced a hybrid (a, k)-anonymity framework for Device-to-Device (D2D) data, but its complexity in parameter setting poses challenges. Reference [11] developed a de-swinging k-anonymity strategy to mitigate location privacy risks from malicious nodes. Reference [12] created a neural network architecture that effectively maintains differential privacy, yet the approach dependence on data quality remains a concern. Reference [13] applied ML techniques to categorize quasi-identifiers, enhancing data anonymization but requiring further exploration of the sensitivity categorization process. the paper [14] proposed a blockchain-based e-government framework integrating artificial immune systems to improve security and privacy. However, the study does not address potential scalability issues in diverse government environments. Reference [15] explored information propagation patterns on Twitter and developed privacy-preserving technologies, yet challenges in heterogeneous data sources persist. The paper [16] proposes the BAW-MOP method, integrating blockchain and reinforcement neural networks to enhance IoT network performance and security. Simulations showed improvements in throughput, accuracy, delay, and energy efficiency but faced challenges with complexity, overhead, and data quality reliance. In [17] a new cloud scheduling technique using Sequential Minimal Optimization (SMO) optimizes task allocation to VMs with deadline constraints. Simulations using CloudSim 3.0.3 demonstrated its benefits over PSO and BAT algorithms, though it lacked real-world validation and scalability testing. Paper [18] reviews AI-based resource management in Fog/Edge computing, addressing provisioning, scheduling, and load balancing. While it offered optimization strategies, it highlighted challenges with heterogeneous resources and AI dependency. Future directions include 5G, IIoT, blockchain, and quantum computing integration. The study [19] introduced a decision tree-based multi-objective resource allocation strategy for cloud computing, but the complexity of implementation and training data quality may limit its practical applicability. The study [20] proposed an SMO-based scheduling algorithm for cloud workloads, improving resource allocation, makespan, and load balancing. While effective in simulations, challenges like parameter tuning and reliability remain for future research. Reference [21] developed the Chronological Sailfish Optimizer (CSFO) algorithm to enhance privacy protection, but the balance between utility and secrecy needs further investigation. Reference [22] aimed to improve data sanitization in the Industrial Internet of Things (IIoT) using a Grasshopper-Black Hole Optimization method, facing challenges in optimal performance. Reference [23] privacy-preserving system allows machine learning trainers to aggregate data without sharing local information, yet its accuracy compared to traditional methods could benefit from further refinement. Reference [24] proposed a federated learning framework to protect client data through differential privacy, though significant noise impacts model performance. Reference [25] introduced the MLPAM model for secure data sharing but lack empirical validation of its effectiveness. Reference [26] presented a privacy-preserving deep learning model that reduces storage overhead but introduces computational demands that may hinder efficiency. Reference [27] developed a privacy-preserving machine learning system that faces challenges in data exchange accuracy, while [28] established a classification architecture using homomorphic encryption, which may struggle with privacy-preserving truths. Finally, [29] proposed the CryptoDL architecture, addressing encrypted data challenges, yet practical encryption key use remains problematic. This paper [30] explores the Internet of Things (IoT) and its impact on technology and society, connecting devices like drones and medical equipment. While IoT improves sectors like healthcare and manufacturing, it also introduces significant cybersecurity challenges. The study highlights gaps in understanding these challenges, particularly in AI, blockchain, zero trust security, and IoT-5G integration. Focusing on the CIA (Confidentiality, Integrity, and Availability) security triangle, it analyzes cybersecurity applications and challenges within the IoT, detailing architectural layers and attack types. The paper assesses existing cybersecurity techniques, offering insights into current research trends in IoT security. This paper [31] examines the impact of cloud computing on resource management, highlighting its foundational principles and potential to revolutionize information technology. It discusses the challenges of adoption and the benefits, such as enhanced innovation, operational efficiency, and scalability. By addressing these challenges, businesses can leverage cloud computing effectively. The study also offers insights into emerging trends in the field. This study [32] addresses secured workflow allocation (SWA) in cloud computing, focusing on task assignment to meet quality-of-service (QoS) standards and balance workloads. It highlights security challenges in sensitive workflows, noting that security overhead can affect QoS performance. The paper reviews existing SWA approaches, presents a taxonomy of challenges, and identifies open issues and future research directions, offering valuable insights for researchers tackling security challenges in SWA.
The current research landscape faces several critical challenges (Table 1), including ensuring location privacy amidst malicious nodes, balancing privacy with the performance of neural networks, and addressing variability in the sensitivity of quasi-identifiers (QIDs). Furthermore, there are significant gaps in achieving interoperability and security within decentralized systems, protecting private information across heterogeneous data sources, and managing the complexity and overhead of autonomous systems. Other challenges include the lack of real-world validation, difficulties in managing heterogeneous resources and balancing utility with privacy in data management. This research addresses these issues by introducing a novel hybrid paradigm that combines k-anonymity with advanced algorithms. The framework features a time-aware Modified Best Fit Decreasing (T-MBFD) algorithm for optimized resource allocation and an updated Firefly algorithm for improved trust generation. By integrating these components, the research effectively enhances privacy protection, trust, and resource efficiency in cloud-based environments, providing a comprehensive solution to the identified gaps and advancing the state of secure and efficient data sharing.
Methodology
The research focuses on improving privacy and trust in cloud computing environments through a two-part strategy. Initially, it implements a privacy-preservation framework using k-anonymity to secure personally identifiable information and obscure user locations and reputations. Following this, an enhanced job allocation policy is introduced to streamline and optimize task distribution among network users, aiming to boost operational efficiency and effectiveness.
The proposed integrated privacy preservation and job allocation system is designed to meet essential user requirements in cloud computing environments, focusing on privacy, efficient resource utilization, scalability, and flexibility. A primary user requirement is robust privacy protection for sensitive data, such as location information and historical activity. To address this, the system incorporates k-anonymity as a privacy-preserving technique, ensuring that each users data is indistinguishable from at least k-1 others, effectively mitigating re-identification risks. This approach balances data security and system performance, chosen over more complex alternatives that could introduce significant computational overhead in real-time environments. For efficient job allocation, the system integrates the T-MBFD algorithm to optimize resource allocation across cloud service providers. T-MBFD minimizes resource wastage and maintains balanced workload distribution, ensuring that jobs are assigned to providers with the most suitable resources, reducing execution time and enhancing overall system performance. The system employs sorting and threshold-based allocation, prioritizing jobs based on user needs while respecting each provider maximum and minimum resource thresholds. Scalability and flexibility are supported by the systems matrix-based data structure, which allows for efficient sorting, partitioning, and anonymization of user data. This structured approach facilitates the quick computation required for k-anonymity, ensuring the system can handle large datasets without performance degradation. The design balances the privacy-performance trade-off by opting for k-anonymity, which provides sufficient privacy protection while maintaining necessary performance levels for real-time job allocation in dynamic cloud environments. The integration of k-anonymity with T-MBFD ensures user privacy is preserved and jobs are efficiently allocated, offering a comprehensive solution addressing both security and operational efficiency.
A. Dataset
The dataset for analysis is extracted from the social cloud [35]. The dataset, comprising 15,000 to 95,000 records, was obtained from http://www.social-iot.org/. Before analysis, preprocessing steps included data cleansing, removal of duplicates, and handling missing values. An open-source IoT dataset, including service context and associated information, was obtained from http://www.social-iot.org/. This dataset was the primary data source, and careful preprocessing was performed to ensure consistency and quality, including resolving missing values, removing duplicates, and cleaning the data.
B. Privacy Preservation and Job Allocation Technique
In a cloud service model with n users providing d services, where each user can act as both a provider and a seeker, privacy preservation and job allocation are critical. This unified algorithm combines k-anonymity for privacy protection and T-MBFD for efficient job allocation.
By integrating privacy protection through k-anonymity and optimized job allocation using the T-MBFD approach, the algorithm effectively addresses both privacy and efficiency. This comprehensive solution offers a robust framework for secure and effective cloud-based data and service management. The T-MBFD algorithm is a resource allocation method tailored for cloud computing environments, where time and resource efficiency are critical. Unlike traditional approaches such as First Come First Serve (FCFS) or Shortest Job First (SJF), T-MBFD accounts for the dynamic nature of cloud resources by integrating time-sensitivity into its decision-making. It ensures that jobs are assigned based on urgency while considering the availability of resources, optimizing both time and resource utilization. The algorithm also adapts to real-time changes in user demands and provider capacities, making it highly effective in environments with fluctuating workloads. This adaptability prevents resource wastage and avoids overloading service providers, maintaining a balanced workload across the cloud infrastructure. The updated Firefly algorithm, on the other hand, enhances traditional optimization by incorporating real-time feedback and a memory component. These modifications allow the algorithm to dynamically adjust its search patterns based on current resource availability and job requirements, which is essential in the cloud constantly changing conditions. The memory component further improves the algorithms efficiency by retaining information from previous iterations, accelerating convergence and avoiding redundant searches. Unlike the traditional Firefly algorithm, which can get stuck in local optima, the updated version is better suited for dynamic environments where continuous adjustments are necessary to optimize resource allocation [36], [37]. Both algorithms provide significant advantages in cloud computing environments, where dynamic workloads and real-time resource adjustments are critical for maintaining operational efficiency and ensuring timely job execution.
The proposed trust generation framework systematically models and allocates jobs in cloud computing environments as shown in Figure 2.The process begins with initializing the user profile, where detailed information about users—including their skills, preferences, and specific requirements—is collected to create comprehensive profiles. Following this, the user request is defined, specifying the service or task needed and ensuring alignment with user expectations [38]. Next, the request is disseminated through appropriate channels to maximize visibility among potential service providers. After receiving responses, the framework assesses the time requirements for each provider to estimate the time needed to complete the task, which is crucial for ensuring timely service delivery. In the job allocation phase, the framework employs the T-MBFD algorithm to assign jobs to providers based on their resource capacity and suitability for the task, optimizing execution time and resource utilization. Finally, the framework incorporates feedback and optimization, collecting performance data from completed jobs to refine future allocations and enhance user satisfaction. This structured approach ensures that the framework effectively preserves user privacy while optimizing resource allocation in a dynamic cloud environment, resulting in a scalable and efficient solution for cloud-based services.
The simulation of each task within the proposed methodology follows a systematic sequence of steps:
Initialize User Profile: The simulation begins with setting up the user profile. This involves collecting detailed information about the user, such as their skills, preferences, and specific requirements. Establishing a comprehensive user profile is essential for accurately understanding their needs.
Define User Request: Following user profile initialization, the next phase is to define the users request. This request specifies the task or service the user needs, whether it involves acquiring a product, soliciting a service, or seeking a solution to a problem. Articulating this request is key to aligning it with the users expectations.
Disseminate Request: Once the request is defined, it must be communicated to potential responders. Disseminating the request involves sharing it through suitable channels or platforms to reach a broad audience. This ensures that the request is visible to numerous potential responders who may be able to address it.
Assess Respondent Time Requirements: After the request is disseminated, interested parties will offer solutions or services. The next step involves assessing the time requirements for each respondent. This includes estimating the time each respondent will need to complete the task or fulfill the request, which helps evaluate their efficiency.
These steps outline initializing the user profile, defining and disseminating the user request, and evaluating potential responders based on their time requirements. This approach ensures a streamlined and effective process for addressing user needs and finding solutions.
C. Hybrid Trust Generation Method
The proposed research employs machine learning techniques to assess service providers using aggregated data collected after resource allocation. Specifically, it optimizes the Levenberg algorithm with the Firefly Algorithm. Integrating these two methods aims to enhance the performance and reliability of the evaluation process, providing a robust mechanism for accurately rating service providers. This approach leverages the strengths of the Levenberg algorithm for neural network training and the Firefly Algorithm for fine-tuning hyperparameters, resulting in improved accuracy and efficiency in the evaluation metrics [39].
Algorithm 1 Proposed Integrated Privacy Preservation and Job Allocation Technique
Require:
Ensure:
Initialize matrices
Combine these matrices into a dataset D where each record
for each record in
record.in_degree
record.out_degree
end for
for each group in groups do
while not all(record.in_degree = maxDegree and record.out_degree = maxDegree for each record in group) do
insert_dummy_records(group, maxDegree)
end while
end for
Divide
shuffle(
if not validate_k_anonymity(
end if
Output the k-anonymized dataset
Create user demands and broadcast them to potential providers: demands
broadcast_demands(demands)
for each job in J do
for each provider in P do
end for
end for
Apply T-MBFD, sort users and providers. Set resource utilization thresholds: MaxThreshold
for each user in
for each provider in
if provider_capacity(provider, MaxThreshold, MinThreshold) then
assign_job(user, provider,
update_provider_resources(provider,
break
end if
end for
end for
Update the resources consumed: update_provider_resources(provider, time_cost)
Adjust the list: update_provider_list(
while jobs_remaining(J) do
allocate_remaining_jobs()
end while
Provide the final list of job-provider assignments: return final_allocation
for each parameter in parameters do
parameter.normalized_value
end for
Finish
Algorithm 2 Pseudo Code of Proposed Trust Generation Technique
Require: max_iters: Maximum number of iterations,
solver_list: List of solver names,
hidden_layer_min_values: Minimum hidden layer sizes, hidden_layer_max_values: Maximum hidden layer sizes,
iter_max_list: Max iteration values for neural network training,
s_test: Test dataset
Ensure: firefly_intensity: Array storing accuracy and loss values for each iteration,
configurations: List storing the configurations used in each iteration
Begin
Initialize Parameters:
Initialize storage for results:
for iteration from 1 to total_iters do
Generate random indices:
Form a neural network using a Levenberg solver:
Formulate predictions using the neural network:
Modify the intensities of firefly:
Update configuration list:
add config to configurations
if iteration > 1 then
for i from 1 to iteration do
for j from 1 to iteration do
Collect the subset of fireflies:
end for
end for
end if
Determine distance d:
Update attraction index AI:
end for
End
The proposed method employs the Levenberg algorithm, enhanced by the Firefly Algorithm, to optimize hyperparameters in the training and classifying of scheduled tasks. A modified equation updates the firefly attraction index based on its distance and proximity to other fireflies with varying performance indices [40]. After each iteration of constructing and evaluating the Levenberg model, the algorithm adjusts the attraction index, making higher-performing models more attractive and less-performing ones less so. This dynamic adjustment aims to find the optimal hyperparameter configuration for the task scheduling architecture. The technique generates a rating based on actual user classification rates, ensuring accurate and efficient model performance. The methods effectiveness is evaluated using quantitative learning criteria, with detailed results in the subsequent section.
Results and Discussion
A. Simulation Environment
A detailed experimental framework was designed to thoroughly assess the efficacy of an innovative hybrid approach to address critical trust and privacy issues in cloud-based data sharing. The investigation began by recognizing the widespread adoption of cloud data exchange and emphasizing the vital need for robust privacy protections for data providers and users. The proposed hybrid model was implemented and evaluated using MATLAB, employing a multi-layered strategy to enhance privacy and trust. In this research, we chose MATLAB as the primary platform due to its user-friendly interface and powerful tools for data analysis, which facilitate rapid prototyping and algorithm development. MATLAB extensive built-in functions enable complex mathematical operations, making it suitable for validating our proposed frameworks effectiveness in a controlled environment. While platforms like HADOOP are ideal for large-scale distributed computing, our focus was on demonstrating the algorithms functionality and privacy preservation capabilities without the added complexity of distributed systems. This approach allowed us to emphasize the strengths of our framework, laying the groundwork for future evaluations in more complex cloud environments. Unlike conventional methods that focus solely on protecting the identity of one party, this model incorporates k-anonymity to ensure the privacy of both data providers and consumers. It also integrates an advanced version of the Firefly algorithm to improve trust assessment, offering a novel approach to this longstanding challenge. Additionally, the model optimizes resource allocation in cloud environments through a T-MBFD algorithm, aimed at increasing efficiency and resource utilization. The effectiveness of privacy protection and trust evaluation was measured using key metrics such as accuracy, precision, recall, and F-measure. The Enhanced Firefly algorithm was instrumental in assessing trust, while k-anonymity was vital for privacy protection. The simulations were conducted in MATLAB R2022a on a Windows 11 Home Single Language system featuring Dual Intel v4 processors (2.60 GHz) and 128 GB of RAM. The models parameters were optimized for peak performance, with privacy maintained through k-anonymity and resource allocation managed by the T-MBFD algorithm. Statistical analysis was employed to evaluate the significance of results across different experimental scenarios, with comparative analysis conducted against existing methodologies. The aim was to comprehensively assess the proposed hybrid approach, demonstrating its potential advantages through rigorous testing and analysis. An extensive experiment with a dataset of 95,000 records proved the scalability and resilience of our strategy. A closer look revealed some impressive findings, including average precision, recall, F-measure, and accuracy scores of 0.958, 0.989, 0.976, and 96.416%, accordingly. These findings unequivocally demonstrate the substantial advantages of our methodology, particularly in terms of accelerated trust development and improved privacy protection. To facilitate a rigorous comparative analysis, our proposed framework was evaluated using the same datasets as those utilized in previous studies [10], [19]. This deliberate choice ensures that any differences in performance metrics are attributable solely to the algorithms themselves, rather than discrepancies in the underlying data. By employing identical datasets, we not only enhance the reliability of our results but also provide a more transparent assessment of the advancements offered by our framework in the domains of privacy preservation and job allocation. This methodological alignment strengthens the credibility of our findings, illustrating how our approach enhances the existing body of knowledge within cloud computing environments. The proposed framework was evaluated within a simulated cloud environment, enabling a controlled assessment of its performance regarding privacy preservation and job allocation strategies. This simulation allowed us to create various user scenarios and workloads, facilitating a thorough analysis of the frameworks efficiency and responsiveness while minimizing the complexities associated with real-world cloud environments.
B. Comparative Performance Analysis
In assessing the suggested algorithm, precision—a crucial metric that gauges the accuracy of positive predictions—and recall—a statistic that measures the capacity to accurately identify pertinent instances—are essential. The findings show that our method routinely outperforms [10] and [19] in terms of precision. This implies that our approach retrieves relevant instances with greater accuracy, which is a major advantage in accuracy-centric data sharing scenarios. Our algorithm demonstrates superior recall performance, particularly when compared to [19], retrieving more pertinent instances effectively. While [10] exhibits a different balance between recall and precision, our approach optimizes both metrics in a way that enhances trust assessment, crucial for cloud-based data sharing. This demonstrates the robustness of the proposed solution, highlighting its capability to outperform in real-world applications where trust and efficiency are paramount.
Table 2 shows the results of simulation assessments carried out over a different number of records, from 15,000 to 95,000, with an emphasis on accuracy and recall metrics for the method [10] and the work that is now in use by [19]. Compared to the two methods, the outcomes match the evaluation metrics of our suggested trust creation approach. The metrics that are employed are recall, F-measure, and precision. Recall quantifies the proportion of retrieved relevant instances, while F-measure is the harmonic mean of precision and recall. Precision quantifies the percentage of relevant instances that are retrieved. The results show that our suggested approach performs more accurately in retrieving pertinent instances than [10] and [19]. Additionally, our system achieves a significantly higher recall compared to [19], efficiently retrieving more relevant instances. While [10] demonstrates strong recall performance, our proposed framework excels in providing a more balanced and robust solution. By optimizing both recall and precision, the proposed algorithm delivers enhanced accuracy in managing cloud-based tasks, surpassing the capabilities of previous models. This balance ensures that our method not only retrieves more relevant data but also maintains strong reliability in diverse cloud environments. The proposed algorithm increased precision and competitive recall show how well it works to create reliable relationships between data seekers and suppliers, which aligns with the primary objective of safe cloud data sharing, highlighted in earlier studies [10], [19]. The competitive recall of the suggested algorithm indicates that it is especially effective at safeguarding the privacy of both searchers and data providers, which is consistent with the literature focus on privacy protection in cloud-based data sharing [10], [19]. The suggested approach produces a favorable F-measure by effectively balancing precision and recall. While [19] performs somewhat better than [10], the proposed approach significantly improves over them, which aligns with the literature emphasis on striking a balance between these two criteria.
Table 3 provides the f-measure and accuracy outcomes for the proposed algorithms trust generation and privacy preservation, with computations conducted across various records ranging from 15,000 to 95,000. The results reveal that our proposed algorithm outperforms [19] in terms of the F-measure, demonstrating its effectiveness in achieving a strong balance between precision and recall. While [10] exhibit a slightly higher F-measure, this can be attributed to their use of more advanced feature selection techniques, which may enhance their algorithms performance. Figure 3 visually represents the robust performance of our approach, showcasing its capability to excel in cloud environments and highlighting its potential for further optimization to reach the high standards set by [10]. The proposed algorithm achieves an impressive F-measure, reflecting a well-balanced approach between recall and precision. While [10] demonstrate a slightly higher F-measure—likely due to their advanced feature selection methods—our approach significantly outperforms [19]. This highlights its effectiveness in maintaining the critical balance between precision and recall emphasized in the literature. Additionally, the enhanced average accuracy of our algorithm underscores its robustness in delivering reliable predictions, which is crucial for real-world applications where data integrity is vital. These findings affirm that our framework not only excels in theoretical metrics but also offers practical value in dynamic cloud environments, fostering user trust and operational efficiency.
The 3D mesh diagram (Figure 4) comprehensively visualizes the comparative analysis of various performance metrics across algorithmic approaches. Each axis of the diagram corresponds to a specific aspect of the study: the x-axis represents the performance metrics, including precision, recall, and F-measure, while the y-axis denotes the algorithmic approaches, encompassing the proposed method and two existing algorithms. The z-axis portrays the metric values obtained from the analysis. The smooth curves in the mesh diagram are derived from interpolated values, enhancing the visual representation of the data and facilitating a more intuitive understanding of the comparative performance across different algorithms. The colormap applied to the diagram further aids in discerning variations in metric values, with warmer colors indicating higher values and cooler colors denoting lower values.
This visualization is a valuable tool for researchers and practitioners in assessing the efficacy of algorithmic approaches in achieving desired performance outcomes. By examining the spatial distribution of metric values within the mesh, stakeholders can gain insights into the strengths and weaknesses of each algorithmic approach relative to specific performance metrics. Such insights are instrumental in guiding decision-making processes related to algorithm selection and optimization, ultimately contributing to advancing research and development efforts in computational science and data analytics. Figure 5 illustrates the accuracy of trust generation and privacy preservation, showcasing varying levels of accuracy corresponding to changes in the number of records. Notably, our proposed trust generation and privacy preservation consistently achieved higher accuracy throughout all record variations than both [10] and [19]. This suggests that the proposed approach demonstrated superior performance metrics with minimal associated complexity, affirming the effectiveness of our proposed methodology.
C. Robustness and Effectiveness Analysis
In summarizing the superior performance of our proposed methodology, a comparative analysis is provided for the average values of all parameters utilized in the study. Table 4 indicates that the proposed approach consistently demonstrated enhanced performance across all parameters, encompassing precision, recall, f-measure, and accuracy. This can be primarily attributed to the improved privacy preservation with minimal complexity, resulting in the superior performance of our proposed methodology. To enhance clarity in the manuscript, it is essential to explicitly state that, while the primary focus is on the studies referenced as [10] and [19], a comprehensive evaluation of other pertinent research has also been conducted. This evaluation is summarized in Table 4, which provides a broader context and comparison of related works. By doing so, the manuscript not only acknowledges the contributions of these key studies but also situates them within a wider academic framework, underscoring the depth of the analysis undertaken. This approach reinforces the validity and relevance of the proposed framework in light of existing literature.
Table 4 presents a comparison study highlighting the suggested methodology significant benefits concerning accuracy, precision, recall, F-measure, AUC-ROC, Specificity and Execution time. These results show that trust-building, privacy protection, and data-sharing efficiency in cloud environments have all been successfully addressed by the research. Presenting a comprehensive and well-rounded solution, the proposed approach represents a substantial improvement in safe cloud-based data sharing, surpassing the performance of current methods. The methods robustness, essential in real-world circumstances where data correctness and reliability are critical, is further highlighted by its high average accuracy.
The proposed methodology as shown in Figure 6 achieves an impressive average precision of 0.958, demonstrating its robust ability to retrieve relevant instances with high accuracy. This precision plays a crucial role in enhancing both the integrity and security of cloud networks, as the methodology consistently ensures that pertinent data is accurately identified. Compared to other models like those presented in studies [10], [19], which also show strong precision, our approach offers a well-balanced performance that effectively meets the demands of cloud-based environments, ensuring trust and privacy. Additionally, the recall metric for the proposed method is 0.989, significantly higher than other approaches, indicating its superior ability to correctly identify relevant instances without missing crucial data points. This high recall rate is particularly important in cloud environments where the failure to detect relevant information can compromise both trust and privacy. Moreover, the F-measure, representing the harmonic mean of precision and recall, is recorded at 0.976 for the proposed methodology, highlighting its balanced performance and robustness. This is in contrast to the lower F-measure values of competing models, indicating less effective trade-offs between precision and recall. The higher F-measure underscores the proposed methods efficiency in balancing these two metrics, which is critical in scenarios where precision and recall are equally important. In terms of accuracy, the proposed method also leads with a score of 96.416%, reflecting its overall effectiveness in making correct predictions. This high accuracy indicates the methods ability to deliver reliable results, even compared to other approaches consistently. The proposed methodology superior accuracy enhances data-sharing efficiency and ensures higher trust and privacy, which are paramount in cloud computing environments. In addition to precision, recall, and F-measure, incorporating a broader range of performance metrics can significantly enhance the robustness of our evaluation framework. AUC-ROC (Area Under the Receiver Operating Characteristic Curve) serves as a crucial metric that summarizes the models ability to distinguish between positive and negative classes across different threshold settings, providing insights into its overall performance regardless of class distribution. Furthermore, specificity is essential metrics that complement each other. This approach is vital in scenarios where both false negatives and false positives can lead to substantial consequences. Additionally, assessing execution time is crucial for understanding the practicality of the proposed methodology in real-world applications, as it highlights the models speed and resource consumption during execution. By integrating these additional metrics into our analysis, we provide a more comprehensive evaluation of the proposed frameworks performance, ensuring that it meets the demands of various operational environments while maintaining high standards of accuracy and reliability. The time complexity of the proposed framework is analyzed to ensure its efficiency in processing and privacy preservation. The core components include the k-anonymity algorithm, which operates with a complexity of O(n log n) due to the sorting of datasets, and the Firefly algorithm for trust generation, which has a complexity of O(n2) in its basic form. The overall time complexity of the proposed method, therefore, aggregates these components, leading to an efficient execution profile of O(n2) for large datasets. This performance is contrasted with existing state-of-the-art algorithms, such as those found in the literature, where time complexities can range from O(n log n) to O(n3). The proposed method demonstrates a favorable balance between computational efficiency and the robust privacy measures it provides, making it a strong candidate for real-world applications in cloud environments.
D. Iterative Convergence Analysis
The graphical representation (Figure 7 and 8) illustrates a comparative analysis of convergence trends for various performance metrics across different iterations, contrasting the proposed methodology with existing algorithms. Each subplot delineates a specific performance metric, encompassing precision, recall, F-measure, and accuracy, while the x-axis denotes iteration count and the y-axis showcases metric values. Smooth curves portray the convergence dynamics, offering a visually compelling depiction of iterative trends. Examining the curves reveals insightful nuances: for instance, in precision convergence, the proposed methodology outperforms existing algorithms [10], [19], achieving a precision of 0.85 after 50 iterations, compared to 0.78 and 0.82, respectively. Additionally, observing the recall curve for the proposed methodology indicates consistent improvement, rising steadily from 0.75 to 0.88 over 100 iterations. These trends facilitate performance assessment, with higher metric values signifying enhanced algorithmic accuracy and effectiveness. Moreover, analysis of convergence patterns aids in identifying optimal convergence points, guiding further optimization efforts. For instance, the accuracy curve for the proposed methodology stabilizes at 0.88 after 70 iterations, suggesting an optimal convergence point. Overall, the graph offers a comprehensive visualization of convergence dynamics and performance metrics, enabling thorough evaluation and comparison of algorithmic efficacy. Integrating improved privacy protection, trust building, and efficient resource distribution into a single framework presents the suggested methodology as a comprehensive response to the complex problems associated with safe and effective cloud-based data sharing. This thorough contribution is significant given the ever-changing digital landscape, where efficiency and data security are critical factors. This study is innovative because it takes a comprehensive and integrated approach to address the various issues around safe cloud data exchange. The model is not only an advance over current approaches but also a holistic solution that simultaneously addresses crucial challenges due to its superior outcomes in privacy protection, trust generation, and resource allocation, as demonstrated by the comparative analysis.
E. Practical Implementation and Relevance
This section shows several compelling real-world application scenarios that demonstrate how the proposed framework can be practically implemented across various cloud environments, highlighting both its relevance and effectiveness:
Multi-Tenant Cloud Platforms: In dynamic multi-tenant cloud infrastructures, where numerous organizations share computing resources, maintaining data privacy and efficient resource distribution is critical. For example, in large-scale e-commerce platforms managing sensitive customer data, the proposed framework leverages k-anonymity to anonymize user information, ensuring privacy. Meanwhile, the T-MBFD algorithm optimally allocates computing resources for operations like real-time order processing and personalized recommendations. This integration guarantees both secure data handling and optimal system performance, enhancing scalability while safeguarding user trust.
Cloud-Based Healthcare Systems: In healthcare cloud systems, privacy and performance are paramount due to the handling of sensitive patient information. For instance, in a cloud-based system for medical diagnostics or patient records, the framework uses k-anonymity to anonymize patient data, ensuring compliance with privacy regulations such as HIPAA. Simultaneously, the T-MBFD algorithm dynamically allocates computational resources for critical tasks like image analysis or AI-driven diagnostics. This ensures healthcare providers can process large datasets quickly and securely, improving patient care without compromising confidentiality or system performance.
Smart Cities and IoT-Enabled Services: Smart city infrastructure relies on massive data streams from IoT devices, such as traffic sensors and environmental monitors. The proposed framework, applied in real-time traffic management systems, ensures privacy by anonymizing location-based data from vehicles. The T-MBFD algorithm dynamically manages cloud resources to optimize traffic flow, processing data from multiple sensors without overloading the system. This real-time resource management improves urban mobility while maintaining data security, making the framework an essential tool in the smart city paradigm.
Financial Cloud Systems: In financial services, where real-time data processing and stringent data privacy are required, the proposed framework excels. For example, in cloud environments handling vast financial transaction data, the system anonymizes transaction histories using k-anonymity, protecting users from potential data breaches. The T-MBFD algorithm efficiently allocates resources for fraud detection algorithms, market trend analysis, and risk assessments. This ensures high throughput and secure data processing, making the framework ideal for high-frequency trading platforms and financial institutions needing to balance performance with regulatory compliance.
These real-world scenarios demonstrate the adaptability and strength of the proposed framework in addressing both privacy concerns and resource optimization, making it a practical solution for diverse cloud computing environments.
Conclusion
This research introduces a novel hybrid framework that enhances privacy preservation through k-anonymity and optimizes resource allocation using the T-MBFD algorithm. Our framework also integrates an updated Firefly algorithm for improved trust generation, which proved effective in real-world cloud environments. The experimental results, using a 95,000-record dataset, demonstrated that the proposed approach outperformed existing models in terms of accuracy, precision, recall, and F-measure. Specifically, the framework achieved an average precision of 0.958, recall of 0.989, F-measure of 0.976, and accuracy of 96.416%. These results indicate a commendable balance between privacy protection, resource utilization, and trust management, offering a robust solution for cloud-based data sharing. However, some limitations exist. First, while the algorithm performs well in simulations, its behavior in real-time, large-scale cloud environments remains to be tested. Second, the frameworks precision, although high, can still be improved to enhance overall trust generation in highly dynamic cloud environments. Future research can focus on overcoming these limitations. Key directions include testing the framework in diverse and more complex conditions, utilizing larger datasets for performance evaluation, and exploring further optimization of precision and real-world deployment. Enhancing scalability and evaluating its effectiveness in live operational settings will also provide deeper insights into its practical applicability. By addressing these aspects, the proposed solution can be expanded to achieve broader and more robust applicability in cloud computing.