IEEE Transactions on Cloud Computing最新文献_第2页

A Multi-Strategy Cooperative Dung Beetle Optimization Algorithm for Joint Resource Scheduling in MEC-Enabled Wireless Blockchain Networks mec无线区块链网络中联合资源调度的多策略协同屎壳虫优化算法

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-06-22 DOI: 10.1109/TCC.2026.3706547

Ruiyou Li;Rongqiang Liu;Liang Zhu;Chen Hu;Xinyi Huang

{"title":"A Multi-Strategy Cooperative Dung Beetle Optimization Algorithm for Joint Resource Scheduling in MEC-Enabled Wireless Blockchain Networks","authors":"Ruiyou Li;Rongqiang Liu;Liang Zhu;Chen Hu;Xinyi Huang","doi":"10.1109/TCC.2026.3706547","DOIUrl":"https://doi.org/10.1109/TCC.2026.3706547","url":null,"abstract":"In mining on wireless blockchain networks that support mobile edge computing (MEC), resource scheduling methods are used to maximize mining profits. However, traditional scheduling approaches often fail to allocate resources efficiently, leading to low mining efficiency within these networks. To address this challenge, we propose a multi-strategy cooperative dung beetle optimization algorithm (MSCDBO) designed to adaptively optimize joint resource scheduling for mining decision and resource allocation in MEC-enabled wireless blockchain networks. In joint resource scheduling, mining decision is achieved by adjusting the population size (the number of miners), while resource allocation is performed by updating the population individuals (the participating miners). In the proposed algorithm, a chaos mapping reverse learning strategy is designed to improve the quality of the initial population, while a local mutation strategy based on adaptive T-distribution and Lévy flight is developed to enhance the mining performance. Furthermore, a Tabu search strategy is incorporated to achieve optimal mining profits. A synthetic example result shows that compared with five competitive algorithms (CMA-ES, CTCM, HHO, DE, and DBO), the MSCDBO algorithm improve the overall profit by 5.56%, which verifies the effectiveness of the proposed approach for resource scheduling in MEC-enabled wireless blockchain networks.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1295-1312"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902968","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}

引用次数: 0

Utility Estimation and Quality Evaluation Incentive Mechanism for Federated Learning 联邦学习的效用估计与质量评价激励机制

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-07-23 DOI: 10.1109/TCC.2026.3716632

Fengsen Tian;Xinglin Zhang;Yun Li

{"title":"Utility Estimation and Quality Evaluation Incentive Mechanism for Federated Learning","authors":"Fengsen Tian;Xinglin Zhang;Yun Li","doi":"10.1109/TCC.2026.3716632","DOIUrl":"https://doi.org/10.1109/TCC.2026.3716632","url":null,"abstract":"Data across private clients in edge networks through federated learning (FL) provide crucial resources for the advancement of artificial intelligence (AI), but are often in a non-independent identically distributed (non-IID) manner. As FL clients only provide local models to the task publisher, their information asymmetry, such as training data sizes and non-IID degrees, significantly impacts on the global model accuracy. Therefore, this paper address the problems of how to make use of the clients’ training data attributes, select which clients for participating in FL, and devise reward strategies to meet economic conditions and boost engagement of high-quality clients. In particular, we propose a utility-estimation and quality-evaluation (Uteque) incentive mechanism for FL. Uteque offers FL efficacy through three components: 1) data classification and publication for revealing true training data information of clients, 2) utility estimation and client selection for determining the set of clients to participate in FL, and 3) reward determination and learning quality evaluation for paying clients with truthfulness and individual rationality. This way, it evaluates the learning quality of clients solely based on their local models and formulates reward strategies based on auction theory. Extensive experiments on real-world datasets verify the effectiveness of Uteque for FL.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"2251-2265"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148903113","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}

引用次数: 0

A Lyapunov Optimization Framework for Green Satellite Communications 绿色卫星通信的Lyapunov优化框架

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-07-17 DOI: 10.1109/TCC.2026.3714472

Qilu Wu;Haihan Duan;Nengfang Xiao;Ying Gao;Xiping Hu

{"title":"A Lyapunov Optimization Framework for Green Satellite Communications","authors":"Qilu Wu;Haihan Duan;Nengfang Xiao;Ying Gao;Xiping Hu","doi":"10.1109/TCC.2026.3714472","DOIUrl":"https://doi.org/10.1109/TCC.2026.3714472","url":null,"abstract":"Renewable-powered multibeam satellite systems are increasingly expected to support edge-like service provisioning in remote and infrastructure-limited environments. In such systems, the satellite payload buffers stochastic service requests, accounts for onboard processing constraints, and delivers queued workloads through adaptive multibeam downlink transmission subject to renewable-energy availability. The coexistence of fast-varying wireless channels, stochastic service queues, and slowly evolving energy supply creates a cross-timescale online control challenge. This paper develops a Lyapunov-based online control framework for queue-aware and energy-adaptive multibeam satellite service provisioning. An energy-deficit virtual queue is incorporated into the Lyapunov drift-plus-penalty procedure to track the accumulated gap between payload energy consumption and harvested renewable energy, enabling transmission decisions to respond to both service backlog and long-term energy availability. The resulting policy is implemented through LONE, a low-dimensional scalar-search algorithm with SDR-based beamforming evaluation. Theoretical analysis establishes service-queue stability, energy-deficit virtual-queue stability, and the corresponding delay–energy tradeoff. Simulations show that LONE provides tunable delay–energy performance under stochastic traffic, time-varying channels, and renewable-energy dynamics.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1277-1294"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148903119","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}

引用次数: 0

Optimal Scaling of Cloud Services 云服务的最佳扩展

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-07-07 DOI: 10.1109/TCC.2026.3710834

Andre Bento;Filipe Araujo;Luís Paquete;Raul Barbosa

{"title":"Optimal Scaling of Cloud Services","authors":"Andre Bento;Filipe Araujo;Luís Paquete;Raul Barbosa","doi":"10.1109/TCC.2026.3710834","DOIUrl":"https://doi.org/10.1109/TCC.2026.3710834","url":null,"abstract":"The rapid growth and increasing complexity of cloud-based systems create a need for efficient resource allocation and optimal scaling strategies.However, achieving a balance between system availability and cost is still a significant challenge, as current solutions provide local optimization of specific infrastructure metrics, prioritizing one aspect at the expense of the other.This challenge is further amplified by the dynamic and unpredictable nature of cloud workloads, making the trade-off between cost efficiency and service availability a moving target. This paper shows that this is an NP-hard problem, which we address by optimizing computational resources based on a model that estimates cost and availability. The paper proposes a Greedy optimization algorithm and compares it against other techniques, such as Brute Force, Genetic Algorithm, and Nemhauser-Ullman. To evaluate the proposed approach, we conducted experiments using two microservice applications as benchmarks, and compare the aforementioned optimization algorithms with results on availability and cost, success rate, optimization execution time, and memory usage. While the Greedy approach does not always find the optimal solution, it proved to be a fast and resource-efficient solution that delivered a robust balance between availability and cost, with a success rate of 93.3% and an average cost increase of 3.7% compared to the optimal solution. These findings highlight the importance of selecting the right optimization strategy and offer practical insights for improving the scalability and efficiency of cloud-based systems.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"2005-2016"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148903157","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}

引用次数: 0

BGSG: A Bi-Level Genetic Algorithm to Stackelberg Diagnosis Game for Edge Service Performance Monitoring 基于双层遗传算法的Stackelberg诊断博弈边缘服务性能监控

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-07-06 DOI: 10.1109/TCC.2026.3710336

Shengli Pan;Kouichi Sakurai;Ying-Chang Liang

{"title":"BGSG: A Bi-Level Genetic Algorithm to Stackelberg Diagnosis Game for Edge Service Performance Monitoring","authors":"Shengli Pan;Kouichi Sakurai;Ying-Chang Liang","doi":"10.1109/TCC.2026.3710336","DOIUrl":"https://doi.org/10.1109/TCC.2026.3710336","url":null,"abstract":"Edge services, such as edge AI, are increasingly delivered by a federation of edge devices and edge/cloud resources that collaboratively provide edge MoE models, compute, storage, and connectivity. We abstract these collaborating units as functional links, and each service instance as a path, focusing on adversarial performance monitoring from a budgeted perspective of intelligent hybrid diagnosis choice. We formulate this challenge as a Stackelberg diagnosis game, where the operator selects a budget-feasible mixture of diagnosis schemes, while an attacker strategically allocates a constrained attack across links. To support strategy search, we propose <inline-formula><tex-math>${sf BGSG}$</tex-math></inline-formula>, a GNN-empowered bi-level genetic algorithm framework. The GNN component acts as a topology-aware estimator built on a link–path bipartite graph, mapping diagnosis statistics and enabling amortized performance evaluation. Our theoretical discussion shows that mixed diagnosis strategies can strictly outperform pure strategies in adversarial instances and that the optimistic Stackelberg formulation is well posed under compactness and continuity. Controlled numerical evaluations on a small-scale proof-of-concept topology provide feasibility evidence for the learned diagnosis strategy and its advantage over tested baselines within this limited empirical scope; large-scale deployment validation remains future work.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1491-1503"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148903233","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}

引用次数: 0

Boosting the Performance of Sketch-Based Traffic Measurement for Programmable Data Center Networks: A Machine Learning Approach 提高基于草图的可编程数据中心网络流量测量性能：一种机器学习方法

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-07-27 DOI: 10.1109/TCC.2026.3717206

Zhiyu Yang;Xiong Wang;Jing Ren;Rongping Lin;Sheng Wang

{"title":"Boosting the Performance of Sketch-Based Traffic Measurement for Programmable Data Center Networks: A Machine Learning Approach","authors":"Zhiyu Yang;Xiong Wang;Jing Ren;Rongping Lin;Sheng Wang","doi":"10.1109/TCC.2026.3717206","DOIUrl":"https://doi.org/10.1109/TCC.2026.3717206","url":null,"abstract":"In modern data center networks, fine-grained traffic information is crucial for ensuring cloud service quality and conducting routine network management. Sketch algorithms are widely applied in programmable data center networks for fine-grained traffic measurement due to their ability to provide guaranteed error bounds with low computational and memory overhead. However, the severely limited resources of off-the-shelf programmable network devices mean that even simple sketches can exhibit significant measurement errors. Furthermore, sophisticated sketch-based schemes are challenging to implement and deploy on such hardware. To address this dilemma, this paper introduces a fine-grained, accurate, and generic traffic measurement framework leveraging machine learning techniques. The framework utilizes an ML-based encoder-decoder architecture to infer precise per-flow statistics from raw sketch data captured in the data plane, thereby enhancing traffic status monitoring. To effectively discern flow characteristics and their collision dynamics within the sketch, we have developed two distinct encoder architectures: the Single-flow Single-layer Collision (SSC) encoder, based on Deep Neural Networks (DNN), and the Multi-flow Multi-layer Collision (MMC) encoder, built upon Graph Attention Networks (GAT). Extensive experiments with real and synthesized traffic traces demonstrate that our framework significantly outperforms direct sketch queries, reducing measurement errors by <inline-formula><tex-math>$1!sim ! 2$</tex-math></inline-formula> orders of magnitude while maintaining efficiency.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1504-1518"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148903239","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}

引用次数: 0

A Resilient Edge Data Integrity Verification Approach via Robust Merkle Trees and Steganographic Transmission 基于鲁棒默克尔树和隐写传输的弹性边缘数据完整性验证方法

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-03-26 DOI: 10.1109/TCC.2026.3697006

Md Rashedul Islam;Yong Xiang;Md Palash Uddin;Yushu Zhang;Dezhong Peng;Jonathan Kua;Abdorasoul Ghasemi

{"title":"A Resilient Edge Data Integrity Verification Approach via Robust Merkle Trees and Steganographic Transmission","authors":"Md Rashedul Islam;Yong Xiang;Md Palash Uddin;Yushu Zhang;Dezhong Peng;Jonathan Kua;Abdorasoul Ghasemi","doi":"10.1109/TCC.2026.3697006","DOIUrl":"https://doi.org/10.1109/TCC.2026.3697006","url":null,"abstract":"Edge Data Integrity Verification (EDIV) plays a crucial role in maintaining the authenticity of cached data replicas in decentralized and resource-constrained edge computing environments. However, existing EDIV methods prioritize generating cryptographic proofs while overlooking the need to ensure robustness against localized data tampering and neglecting the inherent security risks involved in transmitting these proofs over untrusted backhaul communication networks. These gaps leave the EDIV process vulnerable to both the manipulation of integrity proofs and their interception during transmission, ultimately undermining the reliability of EDIV. To address these gaps, we propose RAMTStego-EDIV, a novel framework that ensures both the robustness of integrity proofs and their secure transmission. In particular, RAMTStego-EDIV first introduces the Robust Aggregated Merkle Tree (RAMT)—a ternary hash tree that combines real and auxiliary nodes to produce tamper-evident cryptographic proofs, ensuring that even partial manipulation of the data can be reliably detected. Then, it develops a dual-parity transmission control protocol (TCP) steganography technique to protect the transmission of the RAMT-driven cryptographic proofs that covertly embed the proof bits into two TCP header fields: the parity of the payload length and the least significant bit of the TCP window size. This embedding allows proofs to travel invisibly within normal traffic, avoiding detection and interference by adversaries monitoring the network. Both theoretical analysis and experimental results demonstrate that RAMTStego-EDIV prevents proof tampering and interception, confirming its resilience and suitability for deployment in adversarial mobile edge computing environments.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1329-1342"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902639","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}

引用次数: 0

Cooling Matters: Benchmarking Large Language Models and Vision-Language Models on Liquid-Cooled Versus Air-Cooled H100 GPU Systems 冷却问题：在液冷与风冷H100 GPU系统上对大型语言模型和视觉语言模型进行基准测试

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-06-08 DOI: 10.1109/TCC.2026.3700971

Imran Latif;Muhammad Ali Shafique;Hayat Ullah;Alex C. Newkirk;Xi Yu;Arslan Munir

{"title":"Cooling Matters: Benchmarking Large Language Models and Vision-Language Models on Liquid-Cooled Versus Air-Cooled H100 GPU Systems","authors":"Imran Latif;Muhammad Ali Shafique;Hayat Ullah;Alex C. Newkirk;Xi Yu;Arslan Munir","doi":"10.1109/TCC.2026.3700971","DOIUrl":"https://doi.org/10.1109/TCC.2026.3700971","url":null,"abstract":"The unprecedented growth in artificial intelligence (AI) workloads, recently dominated by large language models (LLMs) and vision-language models (VLMs), has intensified power and cooling demands in data centers. This study benchmarks LLMs and VLMs on two HGX nodes, each with 8× NVIDIA H100 graphics processing units (GPUs), using liquid and air cooling. Leveraging GPU Burn, Weights & Biases, and IPMItool, we collect detailed thermal, power, and computation data. Results show that the liquid-cooled systems maintain GPU temperatures between 41–50 <inline-formula><tex-math>$^{circ }$</tex-math></inline-formula>C, while the air-cooled counterparts fluctuate between 54–72 <inline-formula><tex-math>$^{circ }$</tex-math></inline-formula>C under load. This thermal stability of liquid-cooled systems yields 17% higher performance (54 TFLOPs/ GPU vs. 46 TFLOPs/GPU), performance-per-watt, reduced energy overhead, and greater system efficiency than the air-cooled counterparts. These findings underscore the energy and sustainability benefits of liquid cooling, offering a compelling path forward for hyperscale data centers seeking to optimize AI infrastructure.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1645-1656"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902651","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}

引用次数: 0

A Group-Centric Framework for Task Offloading Scheduling in Industrial Edge Computing Networks With Incomplete Information 不完全信息工业边缘计算网络中以组为中心的任务卸载调度框架

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-03-19 DOI: 10.1109/TCC.2026.3695047

Tongxin Zhu;Ouming Zou;Xiaolin Fang;Junzhou Luo;Yingshu Li;Zhipeng Cai

{"title":"A Group-Centric Framework for Task Offloading Scheduling in Industrial Edge Computing Networks With Incomplete Information","authors":"Tongxin Zhu;Ouming Zou;Xiaolin Fang;Junzhou Luo;Yingshu Li;Zhipeng Cai","doi":"10.1109/TCC.2026.3695047","DOIUrl":"https://doi.org/10.1109/TCC.2026.3695047","url":null,"abstract":"Industrial Edge Computing (IEC) networks have attracted growing attention, where industrial devices offload computation-intensive and delay-sensitive tasks to edge servers. Task offloading scheduling is fundamental to ensuring quality of service in IEC networks. However, most existing studies assume complete network information, which is difficult or even infeasible to obtain in practical IEC deployments due to dynamic workloads and partial observability. As a result, their performance degrades in IEC networks with incomplete information. To address this issue, this paper proposes a group-centric task offloading framework tailored to IEC networks with incomplete information and formulates a delay minimization scheduling problem. For the framework design, we develop the GBG algorithm to obtain the optimal grouping and bandwidth allocation strategy.For online scheduling, the scheduling subproblem is modeled as a Partially Observable Markov Decision Process, and the SGOS algorithm is proposed by integrating Long Short-Term Memory with Soft Actor–Critic to handle incomplete information. Extensive simulations and testbed experiments show that the proposed approach consistently outperforms baseline schemes in convergence speed, delay, and workload balance.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"1262-1276"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902967","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}

引用次数: 0

Secure LLM Annotation: Hierarchical Mental Health Risk Intervention With Safety Alignment 安全LLM注释：具有安全一致性的分层心理健康风险干预

IF 5.3 2区计算机科学

IEEE Transactions on Cloud Computing Pub Date : 2026-07-01 Epub Date: 2026-07-23 DOI: 10.1109/TCC.2026.3716348

Yongfeng Tao;Shuojiayin Yang;Chang Liu;Mingwen Zhang;Minqiang Yang;Jiaxin Chen;Wenbo Chen

{"title":"Secure LLM Annotation: Hierarchical Mental Health Risk Intervention With Safety Alignment","authors":"Yongfeng Tao;Shuojiayin Yang;Chang Liu;Mingwen Zhang;Minqiang Yang;Jiaxin Chen;Wenbo Chen","doi":"10.1109/TCC.2026.3716348","DOIUrl":"https://doi.org/10.1109/TCC.2026.3716348","url":null,"abstract":"Large language models (LLMs) for mental health support face a critical tension: avoiding harmful outputs while maintaining empathetic engagement. Existing safety guardrails either over-reject benign queries or generate emotionally inconsistent responses. To address this, we propose the Multi-Gate Crisis Dialogue Framework (MGCDF), an architecture enabling deep fusion of safety anchoring and dialogue quality optimization for crisis response scenarios. Unlike static filtering approaches, MGCDF employs a dynamic risk controller that activates specialized, low-rank adaptation modules based on real-time crisis severity assessment. For training, we introduce a four-tier crisis dialogue corpus (MGCDChat) constructed via Strategic Collaborative Evolution Method (SCEM), enabling fine-grained expert specialization. Experimental evaluations on 96,072 dialogue instances show that the proposed MGCDF attains 89.8% recall, 100% crisis hotline call hit rate and 97.7% safety score in extreme risk scenarios; benchmark tests confirm a throughput of 5.2 tokens/second, 6.1 milliseconds switching overhead, and memory utilization below 41% (38.88 GB). MGCDF outperforms monolithic baseline frameworks in terms of security and semantic consistency, providing a scalable and auditable solution for professional mental health artificial intelligence.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":"14 3","pages":"2100-2110"},"PeriodicalIF":5.3,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148902970","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}

引用次数: 0