International Journal of Cloud Computing (21 papers in press)

Regular Issues

• Optimization of Semantic Retrieval for Library Digital Resources Using Graph Embedding Technology  
  by Li Zhang, Like Wang, Bo Dai 
  Abstract: The rise of online library resources has expanded user choices but increased demands for search accuracy. Traditional search systems often lack innovation due to their reliance on user-item interaction data. To overcome this, this study combines the relationship-mining power of graph embedding with the preference-positioning ability of the attention mechanism, proposing a more accurate and personalized retrieval model. Experimental results show a retrieval accuracy of 94.7% and a median absolute error of 0.47, outperforming comparison models. The model's response time is as fast as 3 seconds, with a minimum memory usage of 143MB. It also achieves a coverage rate of 92.4% and an accuracy rate of 93.4% even with limited interaction data, highlighting its effectiveness in user preference mining and personalized search.
  Keywords: Semantic retrieval; Graph embedding technology; Attention mechanism; Reinforcement learning algorithm; User preference.
  DOI: 10.1504/IJCC.2026.10074707
   
  
• Empirical Study on Cryptography and Blockchain Enabled Cloud and IoT  
  by Mylapalli Srikanth, V. Rama Krishna 
  Abstract: Scalable services and applications are vulnerable to several attacks and data leakage. For instance, hacking of personal information and an organization's data is a challenge. The present empirical study is an investigation of cryptography and blockchain-based frameworks for IoT security. This study provides a detailed view of symmetric, asymmetric, and hybrid cryptography algorithms and various blockchain algorithms for cloud based IoT security. All the algorithms are compared based on security parameters. Results indicate that most encryption techniques use symmetric key encryption, which ensures secure communication among several users. This study helps to understand the importance of several security factors in IoT platforms and the merits of cryptography and blockchain frameworks. Additionally, this review discusses the issues associated with each algorithm, providing a clear understanding of their limitations and potential areas for improvement.
  Keywords: Internet of Things; Cryptography; Blockchain; Encryption; Symmetric algorithms; asymmetric algorithms.
  DOI: 10.1504/IJCC.2026.10074708
   
  
• Beyond the Cloud: Exploring the Serverless Software Architectures and Function as a Service  
  by Md Mahbub Alam  
  Abstract: Serverless architectures and Function as a Service (FaaS) are transforming software development by enabling developers to focus on coding while minimising infrastructure management. This study, based on a systematic literature review of 60 articles (20132023), explores principles, challenges, and opportunities of serverless computing with emphasis on non-functional requirements (security, availability, reliability), AI/ML integration, cost optimisation, and energy efficiency. Findings show that serverless improves productivity, scalability, and cost efficiency through auto-scaling and pay-as-you-go models but faces challenges such as cold start latency, dependency complexity, security vulnerabilities, and debugging issues. While serverless offers significant benefits, addressing reliability and security concerns is crucial to fully harness its potential.
  Keywords: Cloud Computing; Digital landscape; scalability; security challenges; AI/ML Application; Efficiency; cost-effectiveness; systematic review; foundational concepts; resource utilization.
  DOI: 10.1504/IJCC.2026.10074982
   
  
• Integrating Edge Computing and Quantum Computing for Efficient Data Processing in Disaster Recovery Systems  
  by Hitendra Ashok Chavan, Prashant Nitnaware 
  Abstract: Integrating edge and quantum computing offers a novel approach to enhancing disaster recovery efficiency. Traditional systems lack sufficient processing power and low latency for real-time response. This study aims to improve data processing speed, decision-making efficiency, and scalability in disaster recovery through edge-quantum computing integration. The proposed methodology employs HQEAF for efficient data processing and reduced computational load, QAOA-VQE for optimal resource allocation and recovery planning, and HNRGAN to integrate edge-quantum computing, enhancing real-time data flow, decision-making, and resource management through graph-based optimisation and attention mechanisms. The result shows that the HNRGAN is the fastest loss reduction, starting at 1.5 and quickly dropping to 0.9, with a final value of 0.4, implemented using Python software. Future scope is Integrating edge and quantum computing can enhance real-time data processing, improve decision-making, and optimise resource allocation in disaster recovery.
  Keywords: Edge Computing; Quantum Computing; Hybrid Quantum-Edge Adaptive Filtering; Quantum Approximate Optimization Algorithm with Variational Quantum Eigensolver; Hyper Node Relational Graph Attentional Netw.
  DOI: 10.1504/IJCC.2026.10075036
   
  
• An Adaptive Multi-Agent DDPG Approach to Task Scheduling and VM Migration in Cloud Computing  
  by Amardeep Singh, Gurpreet Singh, Monika Singh 
  Abstract: Core components in cloud computing systems use virtual machine (VM) migration together with task scheduling to enhance both resource utilisation and system performance. The study establishes cloud-based adaptive multi-agent deep deterministic policy gradient (AMS-DDPG) to enhance task scheduling together with VMs. Cloud environment resource distribution becomes more efficient because adaptive multi-agent functions unite with DDPG functionality through AMS-DDPG. Our introduced iterative concept of war and rat swarm (ICWRS) boosts the performance of the AMS-DDPG approach. It achieves optimisation within AMS-DDPG through implementation of RSO and WSO. The effective modification of AMS-DDPG depends on simulation of war tactics through WSO while RSO simulates swarm tactics. The evaluation of the suggested approach includes multiple workload patterns and system configurations through simulated data as well as real time data of Azure. The study achieves better performance through the combined deep reinforcement learning and natural optimisation methods which form an inclusive hybrid optimisation framework.
  Keywords: Rat Swarm Optimizer (RSO);War Strategy Optimization (WSO);Deep Deterministic Policy Gradient (DDPG);Adaptive Multi-Agent System;Multi-Agent Deep Learning;Iterative Concept of War and Rat Swarm.
  DOI: 10.1504/IJCC.2026.10075151
   
  
• YOLO-Based Real-Time Queen Bee Detection in Beekeeping for Optimised Hive Management  
  by Jason Elroy Martis, Sannidhan M. S, Pradeep Nazareth, Narayan Naik, Sadananda L, Philomina Princiya Mascarenhas 
  Abstract: In this study, the authors explore using the YOLOv9 object detection model to identify queen bees in colonies of Apis cerana indica. About 1,000 annotated colony image data were collected from local Karnataka-based beekeepers who captured a variety of environmental and hive conditions. YOLOv9 was trained and tested on a NVIDIA RTX 3090 processor and used the processor capabilities to ensure that it was accurate and able to make quick inferences. The findings showed precision of 94.7, recall of 95.1, mean average precision (mAP) of 94.9, and inference speed of 7 milliseconds per image. These scores are better than any of the earlier versions of YOLO and other state-of-the-art models and demonstrate that YOLOv9 can maintain both accuracy in detection and efficiency in real time. The proposed solution automates queen bee identification, minimising human error and aiding efficient hive management.
  Keywords: Convolutional neural networks; Object recognition; Automated hive monitoring; Precision agriculture technology; Sustainable apiculture.
  DOI: 10.1504/IJCC.2026.10075154
   
  
• A Hybrid Swarm Intelligence Approach for Energy-Efficient and Delay-Aware VM Placement in Cloud Data Centres  
  by Sanjay Gupta, Sarsij Tripathi 
  Abstract: The optimal placement of dynamic virtual machines is a critical NP-hard problem due to competing objectives. We formulated this as an integer nonlinear programming (INLP) model that takes into account constraints related to application availability, communication latency, bandwidth utilisation, energy consumption, and resource usage. We proposed the ACO-TSO-EOL approach as a solution that uses ant colony optimisation (ACO) for effective placement, while tuna swarm optimisation (TSO) is used to fine-tune ACOs parameters. We used elite opposition-based learning (EOL) at the initialisation stage to establish a well-distributed population of TSO agents over the search space. The proposed approach prevents premature convergence and judiciously balances exploration and exploitation. Our proposed approach outperforms existing state-of-the-art methods by achieving an 11.30% reduction in power consumption. Furthermore, our approach achieves better performance in terms of resource waste, bandwidth utilisation, and communication latency by 28.8%, 25.6%, and 16.18%, respectively. These results suggest that efficacy in optimising resource consumption and reducing operating expenses can be achieved while maintaining high application uptime.
  Keywords: Cloud Computing; Virtual Machine Placement; Multi-Objective Optimisation; Swarm Intelligence; Pheromone; Exploration-Exploitation; Synergistic Hybrid Algorithm.
  DOI: 10.1504/IJCC.2026.10075745
   
  
• Cloud Computing Models and Platforms  
  by Artan Mazrekaj, Besmir Sejdiu, Isak Shabani 
  Abstract: Cloud computing has transformed the way how businesses access and manage IT resources, providing flexible, scalable, and cost-effective solutions. The cloud today provides businesses with many solutions for infrastructure, platform and software services and has forced them to adapt to new technology strategies. Furthermore, the demand for cloud services has prompted the development of new market offerings, representing various cloud deployment and delivery models. This article aims to classify and compare cloud models based on performance, cost, reliability and scalability criteria. It analyzes the main features, strengths, and weaknesses of cloud deployment models such as public, private, community, and hybrid, as well as the principal cloud service models such as IaaS, PaaS, and SaaS. By exploring these different cloud models, this article seeks to propose a decision-making framework that assists organizations in selecting the most appropriate model in alignment with their specific objectives and operational requirements.
  Keywords: cloud computing; public cloud; private cloud; hybrid cloud; community cloud; IaaS; PaaS; SaaS; pay-as-you-go.
  DOI: 10.1504/IJCC.2026.10076022
   
  
• Review: Data Integrity Schemes on Smart Technologies: Enhancing Security on Cloud Computing, Blockchain, IOT and Intelligent Systems  
  by Soumia Benkou, Ahmed ASIMI 
  Abstract: Due to the large amounts of data generated by smart technologies, it has become crucial to ensure the reliability and verifiability of storage and processing. Cloud computing allows the use of the same resources, but it also increases integrity, confidentiality, and trust risks. In response, new systems use blockchain for tamper-proof traceability and homomorphic encryption (HE) for ciphertext analysis. This preserves data confidentiality while enabling its use. This survey organizes data integrity verification (DIV) in the cloud, blockchain, IoT, and smart systems. We (i) clarify the security objectives and different roles of hashing/signatures (integrity/authenticity) and encryption (confidentiality), (ii) present a taxonomy including data types (static/dynamic/stream/quantum), metadata strategies, encryption options (lightweight/HE), and audit models (private/TPA/consensus/real-time), (iii) compare representative schemes with their strengths, weaknesses, and performance tradeoffs, and (iv) describe case studies (healthcare, smart city) and future directions (lightweight HE, edge audit, compliance by design). The result is a design-centric map that correlates threat models and operational boundaries with suitable DIV primitives across diverse distributed infrastructures.
  Keywords: Data Integrity Verification; Cloud Computing; Blockchain; IoT; Intelligent Systems; Homomorphic Encryption.
  DOI: 10.1504/IJCC.2026.10076830
   
  
• An Deep Learning Approach Towards Effective Detection And Prevention Of DDos Attack In Cloud Computing  
  by Visnu Dharsini S, Ancy Breen W, Srinarayani K, Sai Smrithi S, Geetha P 
  Abstract: Cloud computing has tremendously helped the service sector to build the services and computing requirements with improved scalability, storage, access, security, and cost. During the DDoS attack, the network traffic is very high and the real service requests are mishandled by the server. Since the advent of cloud services, many security mechanisms have been developed. This work introduces the advanced deep learning model for the effective detection of DDoS attacks in the cloud sector. Here, the developed Cascaded Adaptive Residual Networks (CAResNet) perform the DDOS detection task. The developed CAResNet is constructed by including the Residual Attention Autoencoder and Residual Capsnet. To further enhance the detection performance, a Random Variable Enhanced-Deep Sleep Optimizer (RVE-DSO) is utilized for the parameter tuning. To validate the effectiveness of the proposed technique, the result is compared with the existing model for the given dataset.
  Keywords: Cloud Computing; Attack Detection; Distributed Denial-of-Service Attack; Cascaded Adaptive Residual Networks; Random Variable Enhanced-Deep Sleep Optimizer.
  DOI: 10.1504/IJCC.2027.10077163
   
  
• MltGNN: A Multi-level Temporal Graph Neural Network for Workload Prediction in Cloud  
  by Xingjie Zeng, Leiming Chen 
  Abstract: Depending on user request demand, the workload of each module in a microservice system fluctuates over time. Many workload prediction methods have been proposed to allocate computing resource dynamically. However, many existing works ignore the relationship between modules and can not deal with different dimensional inputs. To address these challenges, this paper proposes a Multi-level Temporal Graph Neural Network (MltGNN) for workload prediction of a complex system. It consists of three blocks, i.e., an individual time-series encoder block for extracting the time-series information and encoding different dimensional features; a multi-level graph embedding block for extracting the relationship information and updating features of each module; and a fusion decoder block for fusing the raw data and graph information to predict the workload. The model is tested with a real-world microservice dataset. The experimental results show that MltGNN achieves 96.68% accuracy, 1.6% more than LSTM.
  Keywords: Pmicroservice system; workload prediction; time-series information; graph neural network; deep learning; data fusion.
  DOI: 10.1504/IJCC.2026.10077284
   
  
• Enhancing Cloud Security using Novel Framework for Real Time Threat Intelligence and Response Adaptive through Artificial Neural Networks  
  by Anshul Kumar 
  Abstract: The dynamic and distributed attributes of cloud computing pose considerable challenges to the security of data storage, memory, processing, and service delivery. Security measures such as signaturebased intrusion detection and access control tend to rely on static rules, making emerging threats such as zeroday exploits, advanced persistent threats, and insider attacks nearly impossible to guard against. Solutions based on the architecture of artificial neural networks (ANNs) offer considerable promise because they can learn complex structures, detect anomalies, and adapt to novel attack forms. A new neural networkbased framework was developed to improve cloud security, focusing on intrusion identification, malware classification, and behaviour authentication. Deep learning structures, including convolutional neural networks, long short-term memory (LSTM) networks, and combined models, were assessed, demonstrating better performance than conventional approaches in actual cloud settings. Experimental results on standard datasets indicated detection accuracies exceeding 98%, with minimal false positive rates.
  Keywords: Cloud Security; Artificial Neural Networks (ANN); Intrusion Detection; Anomaly Detection; Deep Learning; DDoS Mitigation.
  DOI: 10.1504/IJCC.2027.10077951
   
  
• AI-driven ensemble learning for optimised cloud server performance and energy efficiency  
  by Anshul Kumar, Gagan Tiwari, Aadrash Malviya 
  Abstract: The study addresses the critical need for optimising energy efficiency and execution time in cloud computing environments by leveraging machine learning models. With the rapid expansion of cloud-based infrastructures, efficient resource utilisation becomes essential to enhance performance while minimising energy consumption. The research evaluates three powerful machine learning models XGBoost, gradient boosting, and random forest based on key performance metrics such as CPU and memory usage, network traffic, execution time, energy efficiency, and task prioritisation. The findings reveal that XGBoost outperforms other models with an accuracy of 99.80%, the lowest error rate (5.63%), and the highest sensitivity (100%), making it the most effective approach for cloud resource optimisation gradient boosting and random forest also demonstrate strong performance, with accuracies of 97.60% and 94.49%, respectively, but their higher error rates indicate slightly lower efficiency.
  Keywords: cloud computing; machine learning optimisation; XGBoost; gradient boosting; random forest; adaptive resource allocation; quantum computing in cloud; performance metrics.
  DOI: 10.1504/IJCC.2027.10078015
   
  
• Cloud-Edge-End Collaborative Intelligent System Based on VGG16 and ResNet  
  by Dong Pan, Fang Wang, Ping Zhang 
  Abstract: With the rapid growth of multi-source heterogeneous data, traditional image recognition systems in cloudedge-end architectures face unstable accuracy, high inference delay and limited fault tolerance. To address these issues, this study proposes a collaborative intelligent system that integrates shallow texture perception and deep structural discrimination. A dual-parameter fault-tolerant mechanism, combining pruning rate and confidence threshold, enables adaptive path-weight allocation and anomaly correction. Experiments on two public remote sensing datasets demonstrate strong advantages: maximum accuracy of 96.4%, prediction consistency of 97.3%, and efficient edge inference. In multi-class complexity simulations, the system maintained stable outputs, achieving 91.7% classification accuracy and 91.5% prediction consistency. These results confirm that the proposed approach enhances reliability, fault tolerance, and resource scheduling in heterogeneous cloud-edge environments, with scalability for real-time deployment.-
  Keywords: Cloud-Edge-End collaboration; Image classification; Pruning mechanism; VGG16; ResNet.
  DOI: 10.1504/IJCC.2027.10078323
   
  
• Optimizing Energy Efficiency through Dynamic Multi-Objective Scheduling of Scientific Workflows in Cloud Environments  
  by Nadia Dahmani, Hatem Aziza, Hajer Ben Romdhane, Saoussen Krichen 
  Abstract: As cloud applications evolve rapidly, data centers face increasing computational demands that threaten sustainability and environmental balance. Ensuring long-term viability requires reducing energy consumption, optimising resource use, and maintaining service quality. This paper addresses the challenge of scheduling scientific workflows in dynamic, energy-aware cloud environments. Our approach aims to optimise multiple objectives, including makespan, cost, and energy consumption, while accounting for workload variability and changing resource availability. We propose a dynamic scheduling algorithm that adapts to evolving workflow requirements and improves resource utilisation to enhance energy efficiency. Experimental results using benchmark workflows demonstrate the effectiveness of our method in balancing conflicting objectives. Compared to static approaches, our dynamic multi-objective scheduling reduces energy consumption and cost while preserving similar makespan performance. Overall, the proposed algorithm provides an effective solution for improving scientific workflow scheduling efficiency in dynamic cloud environments.
  Keywords: Energy Efficiency; Resource Utilization; Dynamic Cloud Environments; Scientific Workflows.
  DOI: 10.1504/IJCC.2027.10078445
   
  
• Optimizing Text Detection and Recognition in Video with an Improved Deep Residual Network  
  by Laxmikant Eshwarappa  
  Abstract: An improved deep residual network (IDRN) is introduced for identification and elimination of blurred pixels, which integrates a modified residual block and a linear classifier with an improved scaled sigmoid activation (ISSA) function. Subsequently, CNN features are extracted via CNN and then subjected to the LSTM model. Here, the LSTM processes the CNN features to extract the global features by capturing spatial dependencies and related information. Moreover, local features are calculated using the CNN features based on a pseudo-Softmax operation. Moreover, an Improved LinkNet (Imp-LNet) is proposed to classify the text, where the model incorporates an enhanced stem block with proposed ReLU (ESBPR) for the probability reduction in overfitting during training. Lastly, OCR based text recognition process is carried out. According to the experimental output, the FDR of Imp-LNet approach is extremely low, as compared to the other traditional approaches.
  Keywords: Text detection; Text Recognition; improved deep residual network; modified LinkNet.
  DOI: 10.1504/IJCC.2027.10078469
   
  
• ConfidenceQ: a Novel Confidence-Weighted Q-Learning Approach for Optimal Resource Allocation in a Multi-Broker Collaborated Cloud Systems  
  by Chaitra M, Sumana Maradithaya, Koushik S 
  Abstract: Cloud resource allocation is crucial to ensure increased productivity, optimisation and efficiency. However, it faces several challenges, including heterogeneous workloads, security concerns, and cost savings. Effective allocation mechanisms enable overcoming the challenges while maintaining performance objectives. This paper introduces a novel approach, ConfidenceQ, which combines Q-learning with confidence metrics to optimise resource allocation decisions in a cooperative cloud environment with multiple brokers. In the multi-broker environment, collaborative brokers are enabled to learn optimal Virtual Machine selection strategies for dynamic workloads, where collaborative knowledge sharing between brokers is implemented. ConfidenceQ is implemented in a simulation environment and evaluated, showing improved throughput by 70% and resource utilisation by 87% compared to non-collaborative brokers. This work represents a significant improvement in intelligent cloud resource management using Collaborative ConfidenceQ.
  Keywords: Cloud computing; Cloud broker; Collaboration learning; Reinforcement Learning; Cloud Interoperability; Q-learning; Machine Learning.
  DOI: 10.1504/IJCC.2027.10078603
   
  
• Aggregation Analysis of Cattle Infectious Disease Temperature Features Based on Cloud-Edge Computing  
  by Ruiguang Lv, Chao Zhang, Qing Shao 
  Abstract: Cattle infectious disease temperature monitoring suffers from poor real-time performance, limited dynamic warning, and inefficient resource use. Current methods depend on local computing power and isolated farm data, limiting cross-site generalisation. Experiments on body temperature datasets and the UK cattle herd mobility database demonstrate that, compared to existing methods, this model achieves over 55% improvement in training efficiency, while reducing bandwidth consumption and communication costs by 35% and 69%, respectively. The body temperature prediction error is reduced to 0.14
  Keywords: Cattle; Infectious disease; Temperature feature aggregation analysis; Cloud algorithm; Edge computing.
  DOI: 10.1504/IJCC.2027.10078653
   
  
• Optimizing Cloud Resources Using Hidden Markov Models  
  by Islem Teboub, Amar Aissani 
  Abstract: In response to the growing need for adaptive and efficient resource management within cloud computing environments, this paper introduces a predictive resource allocation model grounded in Hidden Markov Models (HMMs). The proposed framework leverages a multivariate observation vector composed of 15 critical cloud performance metrics and employs the Expectation-Maximization (EM) algorithm for model training. Using the Multi-Cloud Service Composition Dataset for simulation and validation, the results reveal that our model not only provides competitive prediction accuracy but also surpasses conventional methods, including KNN, Decision Tree, Random Forest, Ridge Regression, in terms of CPU and memory efficiency. Furthermore, a composite performance index, integrating both accuracy and resource consumption, is proposed to objectively assess the overall effectiveness of each approach. The obtained results confirm that the proposed model achieves the best trade-off between precision and efficiency, making it well-suited for real-time resource prediction and decision-making in cloud infrastructures.
  Keywords: cloud computing; hidden markov model (hmm); resource allocation; performance prediction; machine learning; multivariate observations; temporal modeling.
  DOI: 10.1504/IJCC.2027.10079208
   
  
• Blockchain-Enabled Cloud Security and Data Management: A Systematic Review  
  by Dina Zoughbi, Kavitha V 
  Abstract: Cloud computing offers scalable data storage and processing, yet centralised security models remain vulnerable to single-point failures, unauthorised access, integrity violations and limited auditability. This systematic review examines how blockchain-based mechanisms enhance cloud security and data management across text, image and audio contexts. Following the PRISMA framework, approximately 400 records were identified, 250 screened, 175 assessed for eligibility and 70 studies included. The review shows that blockchain is mainly applied to improve integrity, traceability, authentication and decentralised access control. Hybrid cryptographic architectures, especially those combining blockchain with AES, ECC, ChaCha20-Poly1305, smart contracts and AI-assisted models, provide stronger performance than standalone blockchain designs. Evidence from healthcare, finance and supply-chain applications indicates improved confidentiality, accountability and secure data exchange. However, scalability, computational overhead, energy use, privacy compliance and implementation complexity remain key barriers. The study concludes that blockchain is most effective when embedded within layered, domain-specific cloud security architectures.
  Keywords: Cloud computing; cloud security; blockchain technology; algorithms; PRISMA.
  DOI: 10.1504/IJCC.2027.10079550
   
  
• Internet of Vehicles-Based Intelligent Trajectory Planning for Autonomous Vehicle Lane Changing  
  by Ting Tan, Shulong Wu, Yulin Ma 
  Abstract: Autonomous vehicles face complex challenges such as asynchronous perception of multi-source information. Therefore, an intelligent lane changing trajectory planning method for autonomous vehicles based on internet of vehicles is proposed. Firstly, the vehicle status and road boundary information are obtained through the collaborative perception layer of the internet of vehicles, and an environmental model is constructed by combining laser radar and multi-sensor fusion. Secondly, establish a longitudinal safety distance and lateral collision avoidance model considering V2X delay, and introduce road boundary repulsion to achieve lane keeping. Finally, the improved PER-APF algorithm intelligently switches repulsion modes through dynamic obstacle data and combines virtual repulsion to solve local optimal problems. The experimental results show that the deflection angle of the method proposed in this paper remains stable at 0.01
  Keywords: Internet of Vehicles; Autonomous vehicles; Change track trajectory; intelligent planning.
  DOI: 10.1504/IJCC.2027.10079570