IEEE Transactions on Cloud Computing最新文献

{"title":"Enabling Collusion-Resistant Hierarchical Attribute-Based Multi-Keyword Search for Personal Health Records Systems in Cloud","authors":"Qianqian Su;Rui Zhang","doi":"10.1109/TCC.2026.3709757","DOIUrl":"https://doi.org/10.1109/TCC.2026.3709757","url":null,"abstract":"The proliferation of cloud-based healthcare services has created a growing demand for secure and fine-grained sharing of Personal Health Records (PHRs). To support layered access control, Hierarchical Attribute-Based Keyword Search (HABKS) schemes have been developed. However, existing designs suffer from inherent collusion vulnerabilities that is users with access to different subtrees can jointly recover higher-level secrets, breaking access boundaries. To address this problem, we propose a collusion-resistant hierarchical multi-keyword searchable encryption scheme for privacy-preserving PHR sharing. Different from the previous HABKS, our scheme effectively prevents the propagation dependence of data encryption keys at nodes and user collusion attacks. This benefits from the root node key decoupling mechanism we designed, which enables each policy subtree to generate keys with embedded randomness. Additionally, we design an identity-bound trapdoor and a dual verification mechanism that tightly bind user attributes and keyword tokens, ensuring verifiable and fine-grained access control. Our scheme supports expressive multi-keyword queries with constant-size trapdoors, and offloads most pairing operations to the cloud, requiring only one pairing locally. Security proofs and experimental results confirm that our scheme achieves strong collusion resistance, efficient keyword retrieval, and practical suitability for secure hierarchical PHR systems.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1772-1784"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Two-Round Threshold Key Update for Updatable Oblivious Key Management in Cloud Storage","authors":"Zhiwei Wang;Xiangchen Zhu;Yang Liu","doi":"10.1109/TCC.2026.3717979","DOIUrl":"https://doi.org/10.1109/TCC.2026.3717979","url":null,"abstract":"Cloud key management services (KMS) must periodically rotate master keys to limit the exposure window of compromised key material, yet full re-encryption of large-scale cloud storage objects is operationally prohibitive. Updatable Oblivious Key Management Systems (UOKMS) address this by issuing a compact update token that re-randomises ciphertexts without client interaction, while distributing the master key across <inline-formula><tex-math>$n$</tex-math></inline-formula> cloud KMS nodes so that no single KMS node is a point of failure or trust. Existing threshold UOKMS constructions leave the key-rotation sub-protocol and its security proofs largely implicit. We present a provably secure two-round threshold key update protocol that uses Pedersen distributed key generation (DKG) with Feldman commitments in Round 1 and Beaver-triple multiplication with public verification in Round 2 to rotate the master key <inline-formula><tex-math>$k$</tex-math></inline-formula> to <inline-formula><tex-math>$k^{prime }=rho k$</tex-math></inline-formula> entirely within the cloud KMS tier, threshold-computing the compact update token <inline-formula><tex-math>$Delta =rho ^{-1}$</tex-math></inline-formula> and delivering it to the cloud storage tier without any KMS server or external combiner reconstructing <inline-formula><tex-math>$rho$</tex-math></inline-formula> or <inline-formula><tex-math>$Delta$</tex-math></inline-formula>. All broadcast values are publicly verifiable through Feldman commitments, which enables detection and exclusion of malicious cloud nodes. We give formal definitions and complete proofs of correctness, <inline-formula><tex-math>$(n,t)$</tex-math></inline-formula>-threshold secrecy, client obliviousness, forward security, post-compromise security, proactive security, and robustness against malicious servers, reducing all properties to the discrete logarithm assumption and the information-theoretic security of Shamir sharing. A prototype deployed on commodity hardware achieves an online update latency of 8.9 ms for <inline-formula><tex-math>$n=9$</tex-math></inline-formula> servers, a <inline-formula><tex-math>$2times$</tex-math></inline-formula> improvement over our implementation of the CCS’19 threshold extension and a <inline-formula><tex-math>$4.3times$</tex-math></inline-formula> improvement over an actively-secure MPC baseline, together with a ciphertext re-tokenisation throughput of approximately <inline-formula><tex-math>$37{,}000$</tex-math></inline-formula> objects/s on a 12-core machine. We further quantify the offline preprocessing cost, communication volume, and bandwidth overhead, and show that the additional malicious-secure preprocessing traffic is amortisable across rotation epochs and remains small for typical cloud KMS deployments.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1343-1357"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902640","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"BalanceMoE: An Efficient Dynamic Load Balance Framework to Accelerate Mixture-of-Expert Training","authors":"Yunqi Gao;Bing Hu;Mahdi Boloursaz Mashhadi;Wei Wang;Pei Xiao;Rahim Tafazolli;Mérouane Debbah","doi":"10.1109/TCC.2026.3689276","DOIUrl":"https://doi.org/10.1109/TCC.2026.3689276","url":null,"abstract":"The sparsely-activated Mixture-of-Expert (MoE) techniques support scaling the parameter sizes of pre-trained models to the trillion-level without increasing the computational costs. However, in large-scale cloud computing environments, dynamic load imbalance caused by the random expert selection of samples draws a huge challenge for distributed training efficiency. To address these challenges, we propose a lighter-weight and lower communication overhead dynamic load balancing framework, called BalanceMoE, to accelerate MoE model training. BalanceMoE is based on two key novel ideas. Firstly, we model a worker-pair-based expert transfer mechanism that considers the tradeoff between the expert parameter communication and the time reduction obtained. We perform a theoretical analysis and design a highly lightweight algorithm to obtain a near-optimal load balancing solution for per-iteration time reduction. Then, we present our proposed scheme for parallelization of expert computing and transfer, which overlaps the parameter communication of transferred experts and the computing of non-transferred experts to reduce per-iteration training time. We implement BalanceMoE architecture on the PyTorch framework. Extensive experiments on two clusters demonstrate that at training speed, BalanceMoE can achieve up to 1.26x, 1.79x and 2.62x speedup compared to the state-of-the-art SmartMoE, FasterMoE and FastMoE, respectively. At memory usage, BalanceMoE saves up to 71% and 36% of memory compared to FasterMoE and FastMoE, respectively. At energy consumption, BalanceMoE saves up to 13% of the energy consumed within each training iteration compared to SmartMoE.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1462-1478"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902772","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"BAAGS: Blockchain Assisted Aggregate Group Signature for Medical Data Sharing","authors":"Junze Lu;Chunqiang Hu;Bin Cai;Tao Xiang","doi":"10.1109/TCC.2026.3708845","DOIUrl":"https://doi.org/10.1109/TCC.2026.3708845","url":null,"abstract":"With the widespread adoption of electronic health records (EHRs), different institutions are now able to collaborate more efficiently, potentially unlocking the value embedded in medical data. However, most current studies focus solely on the protection of EHR data, overlooking the need for mutual privacy protection in data exchange processes. Moreover, privacy protection should exhibit a degree of flexibility to allow identity decryption when necessary. Blockchain technology offers tamper-proof security, while group signature mechanisms can achieve an appropriate level of anonymity. To meet these requirements and practical application scenarios, we propose a blockchain-assisted cross-institutional group signature scheme, called BAAGS, for medical data sharing. First, to meet the need for validity periods for members within healthcare institutions, we design a time-bound key group signature mechanism that supports dynamic updates. Second, to address the issue of signature verification efficiency in a blockchain environment, we propose an aggregate verification method. Finally, with blockchain assistance, the legitimacy of data holders can be verified, preventing malicious data leaks. Through rigorous security proofs and performance analysis, our scheme ensures both security and high efficiency.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1448-1461"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902788","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Causal-Inference-Driven Communication-Enhanced Computation Offloading for IRS-Assisted IoV","authors":"Yuanlin Lin;Kai Peng;Kunkun Yue;Bohai Zhao;Xiaolong Xu;Victor C.M. Leung","doi":"10.1109/TCC.2026.3721081","DOIUrl":"https://doi.org/10.1109/TCC.2026.3721081","url":null,"abstract":"In the mobile edge computing (MEC)-enabled Internet of Vehicle (IoV) networks, the communication quality between vehicles and edge servers during task computation offloading is vulnerable to degradation caused by obstacle interference, thereby reducing user quality of service. The Intelligent Reflecting Surface (IRS) is a passive and programmable device, can be installed on the exterior of buildings to reflect and enhance signals through phase shift adjustment. However, a coupling relationship exists between IRS phase control and task offloading decisions. Optimizing them separately by decoupling fails to reflect the highly coupled and dynamic interactions in real systems. This may lead to a mismatch between offloading decisions and channel conditions, making it difficult to achieve optimal system performance in real-world environments. To address this challenge, we propose a joint optimization method for IRS phase control and task computation offloading based on causal inference (OITC). Specifically, we first analyze and model the dynamic channel between vehicles and the IRS. Then, causal inference is used to verify the causal relationship between IRS phase control and task computation offloading. Moreover, we combine causal inference with the Multi-agent Proximal Policy Optimization (MAPPO) algorithm to jointly optimize IRS phase control and task offloading decisions. Ultimately, extensive experiments demonstrate that the proposed algorithm outperforms traditional multi-agent deep reinforcement learning (MADRL) methods in terms of communication quality, system delay, and energy consumption.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1553-1568"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902849","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SLA Compliance in TOSCA-Based Multi-Access Edge Computing Deployments","authors":"Álvaro Santos;Jorge Bernardino;Noélia Correia","doi":"10.1109/TCC.2026.3714735","DOIUrl":"https://doi.org/10.1109/TCC.2026.3714735","url":null,"abstract":"The increasing adoption of network-connected devices is driving computation closer to the network Edge, enabling latency-sensitive services through <i>Multi-access Edge Computing</i> (MEC). While this paradigm improves responsiveness, it makes it significantly difficult to manage the <i>Quality of Service</i> (QoS) in distributed application mashups, particularly when dynamic <i>Service Level Agreements</i> (SLAs) must be enforced across interdependent services. Existing TOSCA-based approaches provide expressive deployment models but offer limited support for runtime SLA management and for handling the cascading effects of corrective actions in complex MEC environments. This article proposes a framework for dynamic SLA management of TOSCA-described mashups deployed over MEC infrastructures. The framework combines a non-intrusive TOSCA extension, for client-specific SLA specification, with an impact-aware optimization model that explicitly captures both direct and indirect effects of remediation actions. The experimental evaluation demonstrates that neglecting indirect impacts frequently leads to suboptimal or even detrimental decisions. The proposed optimization strategy significantly improves SLA compliance by selecting actions that account for propagation effects across services, highlighting the importance of holistic, impact-aware decision-making in dynamic MEC deployments.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"2111-2124"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902964","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Study and Evaluation of Optimal Placement of Applications in Mixed Environments Including Quantum Computers for Environmental Adaptation","authors":"Yoji Yamato","doi":"10.1109/TCC.2026.3698110","DOIUrl":"https://doi.org/10.1109/TCC.2026.3698110","url":null,"abstract":"In recent years, the number of applications using heterogeneous hardware such as FPGAs and GPUs has been increasing. However, to fully utilize them, an understanding of the hardware is necessary, which is a high barrier. Based on this background, we have proposed environment-adaptive software that automatically converts code written by programmers for normal CPUs according to the deployment environment, enabling high performance operation. This paper targets automatic offloading to appropriate hardware in a mixed environment that contains normal CPUs, multi-core CPUs, FPGAs, GPUs, and quantum computers. We confirm how much the performance of normal CPUs can be achieved with each hardware, and based on cost performance, we determine the appropriate hardware for offloading and perform automatic offloading. We confirm that the proposed method can automatically offload by measuring the processing time using actual heterogeneous hardware.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"2125-2135"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902965","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"MAP: GPU Resource Planning for Multi-Agent Systems With Stochastic and Verbose Outputs","authors":"Qi Zhang;Mingyang Li;Qiang Li;Xiaoqi Hua;Fahao Chen;Deze Zeng;Peng Li","doi":"10.1109/TCC.2026.3712315","DOIUrl":"https://doi.org/10.1109/TCC.2026.3712315","url":null,"abstract":"Multi-agent systems (MAS) powered by large language models (LLMs) have shown strong potential in complex reasoning and decision-making tasks. While existing research has made significant efforts in designing agent collaboration strategies and optimizing runtime execution, a foundational systems challenge remains underexplored: GPU resource planning, i.e., how to allocate limited GPU resources across diverse MAS services while meeting throughput-oriented service-level agreements (SLAs). This article introduces MAP, a resource planning framework that minimizes GPU provisioning costs for MAS deployments while ensuring SLA compliance. MAP tackles two key technical challenges. First, it models the GPU allocation problem as a chance-constrained optimization, accounting for the stochastic nature of LLM-generated output lengths and providing probabilistic guarantees on service performance. Second, it proposes a topology-aware output regularization mechanism that selectively compresses agent outputs based on their position and role within the collaboration graph, effectively reducing verbosity and improving computational efficiency while preserving task accuracy within a bounded degradation threshold. Extensive experiments across a range of MAS workloads show that MAP reduces GPU usage by up to 77% compared with the conservative over-provisioning baseline, while maintaining or improving service quality.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1963-1974"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148903102","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AMIMSO: Adaptive Markov Inspired Mutation Swarm Optimization for Multi-Objective Task Scheduling in End-Edge-Cloud Environments","authors":"Xue Zhai;Shanchen Pang;Sibo Qiao;Shihang Yu;Haiyuan Gui","doi":"10.1109/TCC.2026.3687347","DOIUrl":"https://doi.org/10.1109/TCC.2026.3687347","url":null,"abstract":"With the widespread adoption of end-edge-cloud computing architectures, modern distributed systems face the dual challenges of real-time data processing and resource optimization in task offloading and resource allocation. Efficient scheduling of diverse tasks presents a complex multi-objective optimization problem. To address this, this paper introduces a multi-objective optimization model that leverages the Lagrange multiplier method to balance task scheduling time, resource consumption, and task completion rates, achieving effective trade-offs among these objectives. Additionally, we propose an adaptive Markov inspired mutation swarm optimization algorithm(AMIMSO). By incorporating a Markov state transition mechanism, the algorithm dynamically adjusts inertia weights and learning factors based on the system’s state, enabling adaptive switching between global exploration and local exploitation. This approach efficiently solves multi-objective optimization problems. Furthermore, the integration of mutation operations enhances the algorithm’s capability to escape local optima, significantly improving global search efficiency and convergence speed. The experimental results based on simulation environment demonstrate that AMIMSO outperforms existing methods in terms of task completion rates, resource utilization efficiency, and system responsiveness, offering a robust and efficient solution for task scheduling and resource optimization in dynamic environments.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1392-1406"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148903110","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"FedLID: Leveraging Dimensionality to Overcome Noisy Labels in Federated Learning","authors":"Yufei Kang;Chenghao Hu;Baochun Li","doi":"10.1109/TCC.2026.3708715","DOIUrl":"https://doi.org/10.1109/TCC.2026.3708715","url":null,"abstract":"Federated Learning (FL) facilitates multiple clients to train a shared model on their own data in a privacy-preserving manner. In real-world FL applications, clients may contain label noise due to human annotation errors. Existing methods often struggle in the presence of data heterogeneity, as they operate under the assumption that noisy data tend to produce higher losses. This assumption convolutes the distinction between hard samples and mislabeled samples, particularly in data heterogeneous environments. In this study, we comprehend the mislabeled data within the FL process by examining the dimensionality of their deep representation subspace. Our investigations reveal distinct trends in local intrinsic dimensionality between noisy and clean local samples. On the basis of this finding, we propose FedLID as a practical solution that refurbishes mislabeled data and regularizes noisy data simultaneously. Empirically, we demonstrated that FedLID surpasses current state-of-the-art methods in managing noisy labels in federated learning with data heterogeneity. Furthermore, experiments under symmetric label noise and on the large-scale CIFAR-100 dataset confirm that FedLID generalizes robustly across diverse noise types and dataset scales, showcasing the potential of subspace dimensionality in advancing data quality research.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1858-1869"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148903165","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}