Robust Federated Learning (RoFL) extends traditional federated learning, not only by enabling multiple clients to collaboratively train a shared model under the coordination of an edge server, but also by incorporating client-side defense mechanisms (e.g., adversarial training) to defend against adversarial attacks while preserving data privacy. However, recent studies have shown that RoFL also remains vulnerable to the challenges posed by non-independent and identically distributed (non-IID) data distributions across heterogeneous clients, which can degrade overall model generalization and robustness. To mitigate this challenge, in this paper, we propose a novel RoFL framework, called RoFLCCA, to address non-IID challenges while defending against adversarial attacks. In particular, we first introduce a local classifier calibration mechanism that utilizes feature-level augmentation to mitigate the effects of non-IID data. By incorporating global class-wise feature statistics, each client can adjust its classifier using synthetic features derived from these shared representations. Second, we propose a calibrated classifier-guided global adversarial alignment strategy, which enforces consistency between augmented and adversarial predictions to improve robustness. Simulation results demonstrate the effectiveness of the proposed RoFLCCA, which consistently outperforms existing robust federated baselines across different datasets and settings. On average, it achieves a 7.07% improvement in clean accuracy and a 4.71% gain in adversarial robustness, highlighting its ability to enhance both generalization and defense against adversarial threats.

The proliferation of data-intensive, low-latency applications has driven the adoption of multi-access edge computing (MEC) to meet the demand for high-performance computing at the network edge. However, ensuring reliable communication under non-line-of-sight (NLoS) conditions remains a significant challenge. While reconfigurable intelligent surfaces (RISs) and the more recent simultaneously transmitting and reflecting RISs (STAR-RISs) offer promising solutions, their passive nature and susceptibility to multiplicative fading limit performance gains. To address these challenges, we propose a novel active STAR-RIS-assisted MEC system that enhances signal strength and adaptability by enabling amplification and joint control over signal transmission and reflection. Our objective is to minimize the energy consumption of user devices, considering both local task computation and uplink task offloading, while maintaining task queue stability. We formulate a joint energy minimization problem with system constraints and long-term queue stability requirements. This problem is decomposed into subproblems: (i) sequential fractional programming is applied to optimize user transmit power, (ii) convex optimization is used to determine partial task offloading ratios, and (iii) a modified Lyapunov optimization combined with double deep Q-networks (DDQN) is proposed to iteratively solve the active STAR-RIS parameters (amplitude and phase shift), amplification control, and task admission at the user side. Numerical results indicate that our proposed system outperforms the conventional passive STAR-RIS-assisted system by 18.64% and the conventional passive RIS-assisted system by 30.43%, respectively.

The rapid evolution towards the sixth-generation (6G) networks demands advanced beamforming techniques to address challenges in dynamic, high-mobility scenarios, such as vehicular communications. Vision-based beam prediction utilizing RGB camera images emerges as a promising solution for accurate and responsive beam selection. However, reliance on visual data introduces unique vulnerabilities, particularly susceptibility to adversarial attacks, thus potentially compromising beam accuracy and overall network reliability. In this paper, we conduct the first systematic exploration of adversarial threats specifically targeting vision-based mmWave beam selection systems. Traditional white-box attacks are impractical in this context because ground-truth beam indices are inaccessible and spatial dynamics are complex. To address this, we propose a novel black-box adversarial attack strategy, termed Spatial Proxy Attack (SPA), which leverages spatial correlations between user positions and beam indices to craft effective perturbations without requiring access to model parameters or labels. To counteract these adversarial vulnerabilities, we formulate an optimization framework aimed at simultaneously enhancing beam selection accuracy under clean conditions and robustness against adversarial perturbations. We introduce a hybrid deep learning architecture integrated with a dedicated Feature Refinement Module (FRM), designed to systematically reshaping irrelevant, noisy and adversarially perturbed visual features. Evaluations using standard backbone models such as ResNet-50 and MobileNetV2 demonstrate that our proposed method significantly improves performance, achieving up to an +21.07% gain in Top-K accuracy under clean conditions and up to a +37.32% increase in Top-1 adversarial robustness compared to different baseline models.

With the escalating demands for efficient, heterogeneous, and increasing service deliveries in peer devices, collaborative edges, and edge-cloud computing environments, accurate and frequent workload prediction is critical to optimize resource utilization and reduce costs. Traditional approaches struggle with the variability, heterogeneity, and dynamicity of workloads. Lightweight models, while addressing resource limitations, often compromise accuracy, which can lead to task failure or resource wastage. These concerns lead to two fundamental research questions: 1) How can we design an efficient workload prediction system to reduce costs in cloud servers? and 2) How can a lightweight workload prediction system be achieved for edge devices? To address these challenges, we propose a novel transformer-based Multi-Scale Temporal and Hierarchical Knowledge Distillation (MSTH-Former) and Convergence approach. MSTH-Former transfers representations of multiple networks (i.e., scales) into one single student (single scale). This novel framework uses shorter time series data to effectively capture workload patterns, enabling frequent predictions and reducing resource wastage. MSTH-Former integrates hierarchical knowledge distillation techniques with multi-scale temporal patterns to capture diverse interdependencies across scales. MSTH-Former reduces resource wastage by an average of 85% and improves prediction accuracy by 32% compared to the baseline on the benchmark datasets.

Do diverse perspectives help students learn better? Multi-teacher knowledge distillation, which is a more effective technique than traditional single-teacher methods, supervises the student from different perspectives (i.e., teacher). While effective, multi-teacher, teacher ensemble, or teaching assistant-based approaches are computationally expensive and resource-intensive, as they require training multiple teacher networks. These concerns raise a question: can we supervise the student with diverse perspectives using only a single teacher? We, as the pioneer, demonstrate TeKAP, a novel teacher knowledge augmentation technique that generates multiple synthetic teacher knowledge by perturbing the knowledge of a single pretrained teacher i.e., Teacher Knowledge Augmentation via Perturbation, at both the feature and logit levels. These multiple augmented teachers simulate an ensemble of models together. The student model is trained on both the actual and augmented teacher knowledge, benefiting from the diversity of an ensemble without the need to train multiple teachers. TeKAP significantly reduces training time and computational resources, making it feasible for large-scale applications and easily manageable. Experimental results demonstrate that our proposed method helps existing state-of-the-art knowledge distillation techniques achieve better performance, highlighting its potential as a cost-effective alternative. The source code can be found in the supplementary.

As Industry 5.0 emerges, the convergence of advanced technologies like the Internet of Things (IoT) and blockchain is vital in shaping the future of industrial automation. Industry 5.0 emphasizes the collaborative relationship between humans and machines, requiring robust, decentralized systems to ensure security, accountability, and trust in interconnected ecosystems. Currently, IoT data processing is cloud-centric, which introduces challenges like fragmented data silos, limiting the potential for seamless and secure real-time analytics. Blockchain technology offers a solution by providing a decentralized and transparent ledger that can enhance data integrity and security across IoT applications. This study investigates the integration of blockchain with the IoT in the context of Industry 5.0, highlighting the potential for improved data management, security, and human-machine collaboration. By conducting a comprehensive analysis of IoT application designs and blockchain platforms, we evaluate existing literature to uncover the challenges, benefits, and limitations of this integration. Our research contributes by proposing a framework for selecting optimal blockchain platforms for IoT applications in Industry 5.0, providing actionable recommendations for enhanced data trust and resilience. Future research directions are also outlined to address the evolving demands of this technological convergence, ensuring that IoT ecosystems are secure, scalable, and human-centered in the era of Industry 5.0.

Multi-task learning (MTL) has emerged as the preferred strategy for integrating artificial intelligence (AI) into the Internet of Things (IoT) environments by enabling a single model to perform multiple related tasks through shared representations. Yet, deploying large-scale MTL models such as vision transformers on resource-constrained devices remains challenging due to their high computational and memory demands. While edge-cloud collaboration offers partial relief via task offloading, it introduces latency and cloud reliance, which are incompatible for delay-sensitive applications. Edge model caching offers a promising alternative by storing frequently used models locally. However, traditional full-model caching is suboptimal for MTL scenarios where only task-specific model components are utilized, leading to unnecessary storage and computation overhead. To address this, we propose CacheMoE, a task-aware expert model caching framework for multi-task inference in distributed edge environments. CacheMoE leverages a Mixture-of-Experts (MoE) architecture to decompose MTL models into modular, task-specific expert components and selectively caches the most relevant experts across edge nodes (ENs). A Dynamic Weighted Federated Learning (DWFL) algorithm is employed to train personalized expert popularity models for each user equipment (UE) in a privacy-preserving and communication-efficient manner. Based on the predicted task demands, we further design a Multi-Agent Deep Reinforcement Learning (MADRL) algorithm to collaboratively determine expert caching decisions across ENs under storage and latency constraints. Experimental results demonstrate that CacheMoE significantly improves the cache hit ratio, reduces cost, latency, and supports scalable, low-latency multi-task inference across heterogeneous edge environments.

Recent state-of-the-art (SOTA) techniques have demonstrated substantial efficacy in 3D Human Pose Estimation (HPE) from videos. Despite strong progress, no prior work has used soft supervision to handle Hard-to-Estimate (H2E) samples in 3D pose estimation, which is inherently a regression task. Existing H2E-example mining solutions, based on logit distillation, progressive target refinement, and label smoothing, are confined to classification problems. Traditional regression-based problems are deprived of the benefits of these regularization techniques. A soft supervision-based H2E example mining technique is crucial for regression problems.To the best of the author’s knowledge, there are no soft supervision-based regularization techniques exist for regression problems. This paper proposes a novel training strategy, referred to as Progressive Soft-Supervision Training for Regression Problems (PSTR). PSTR introduces the concept of progressivesoft targetsto the 3D pose estimator, a regression-based task. Highly inaccurate predictions, representing the H2E poses, are focused more while preserving the representations for Easy-to-Estimate (E2E) poses. PSTR forces the network to learn an alternate optimum inductive bias in the solution spacevia dynamically shifting gradients. The proposed PSTR improves SOTA performances on the large-scale Human3.6m dataset by a large margin with an average MPJPE, and P-MPJPE of 1.2mmand 1.09mmfor Protocol 1 and 2, respectively, where improvements on PCK, AUC, and MPJPE for MPI_INF_3DHP dataset are, 1.53%, 1.75% and 1.40mmfor 2D-3D pose uplifting and 1.60%, 1.95% and 1.60mmfor RGB to 3D pose estimation tasks, respectively. PSTR can be effortlessly deployed on any regression-based task.

Non-orthogonal multiple access (NOMA) enables multiple users to share the same frequency band, and simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) provides 360-degree full-space coverage, optimizing both transmission and reflection for improved network performance and dynamic control of the indoor environment. However, deploying STAR-RIS indoors presents challenges in interference mitigation, power consumption, and real-time configuration. In this work, a novel network architecture utilizing multiple access points (APs), STAR-RISs, and NOMA is proposed for indoor communication. To address these, we formulate an optimization problem involving user assignment, access point (AP) beamforming, and STAR-RIS phase control. A decomposition approach is used to solve the complex problem efficiently, employing a many-to-one matching algorithm for user-AP assignment and K-means clustering for resource management. Additionally, multi-agent deep reinforcement learning (MADRL) is leveraged to optimize the control of the STAR-RIS. Within the proposed MADRL framework, a novel approach is introduced in which each decision variable acts as an independent agent, enabling collaborative learning and decision making. The MADRL framework is enhanced by incorporating convex approximation (CA), which accelerates policy learning through suboptimal solutions from successive convex approximation (SCA), leading to faster adaptation and convergence. Simulations demonstrate significant improvements in network utility compared to baseline approaches.

The recent emphasis has been on achieving rapid and precise detection of salient objects, which presents a challenge for resource-constrained edge devices because the current models are too computationally demanding for deployment. Some recent research has prioritized inference speed over accuracy to address this issue. In response to the inherent trade-off between accuracy and efficiency, we introduce an innovative framework called LightSOD, with the primary objective of achieving a balance between precision and computational efficiency. LightSOD comprises several vital components, including the spatial-frequency boundary refinement module (SFBR), which utilizes wavelet transform to restore spatial loss information and capture edge features from the spatial-frequency domain. Additionally, we introduce a cross-pyramid enhancement module (CPE), which utilizes adaptive kernels to capture multi-scale group-wise features in deep layers. Besides, we introduce a group-wise semantic enhancement module (GSRM) to boost global semantic features in the topmost layer. Finally, we introduce a cross-aggregation module (CAM) to incorporate channel-wise features across layers, followed by a triple features fusion (TFF) that aggregates features from coarse to fine levels. By conducting experiments on five datasets and utilizing various backbones , we have demonstrated that LSOD achieves competitive performance compared with heavyweight cutting-edge models while significantly reducing computational complexity .

Abstract In the era of industry 4.0 and the widespread use of digital devices, the number of cyber attacks poses an escalating and diverse threat, jeopardizing users' online activities. Intrusion detection systems (IDS) emerge as pivotal solutions, playing a crucial role in detecting anomalous signals within network systems. To counter novel attack patterns, IDS systems require periodic rule updates for effective identification of unusual signals. Typically, these policies are updated based on rule‐based or deep learning algorithms to enhance detection performance. However, the insufficient number of labeled samples remains a challenge for real‐world deployment. In this article, an automated labeling method is presented that has shown high effectiveness, requiring minimal hardware resources, and applicable to IDS systems. Additionally, the approach utilizes transfer learning combined with attention mechanisms to boost the efficiency of abnormal signal detection. The results from the approach are compared with those of a reference model, illustrating an overall improvement of nearly 10% in our model's performance compared to the reference model. This underscores the effectiveness of automating rule adjustments for IDS, contributing significantly to reducing associated financial costs. The research addresses the challenges in deploying IDS in real‐world scenarios and provides a valuable contribution to enhancing cyber threat detection capabilities. A preprint has previously been published [11].

Abstract The traditional client-based HTTP adaptation strategies do not explicitly coordinate between the clients, servers, and cellular networks. A lack of coordination leads to suboptimal user experience. In addition to optimizing Quality of Experience (QoE), other challenges in adapting HTTP adaptive streaming (HAS) to the cellular environment are overcoming unfair allocation of the video rate and inefficient utilization of the bandwidth under the high-dynamics cellular links. Furthermore, the majority of the adaptive strategies ignore important video content characteristics and HAS client information, such as segment duration, buffer size, and video duration, in the video quality selection process. In this paper, we present a content-aware hybrid multi-access edge computing (MEC)-assisted quality adaptation algorithm by taking advantage of the capabilities of edge cloud computing. The proposed algorithm exploits video content characteristics, HAS client settings, and application-layer information to jointly adapt the bitrates of multiple clients. We design separate strategies to optimize the performance of short and long duration videos. We then demonstrate the efficiency of our algorithm against client-based solutions as well as MEC-assisted algorithms. The proposed algorithm guarantees high QoE, equitably selects video rates for clients, and efficiently utilizes the bandwidth for both short and long duration videos. The results from our extensive experiments reveal that the proposed long video adaptation algorithm outperforms state-of-the-art algorithms, with improvements in average video rate, QoE, fairness, and bandwidth utilization of 0.4%–12.3%, 8%–65%, 3.3%–5.7%, and 60%–130%, respectively. Furthermore, when high bandwidth is available to competing clients, the proposed short video adaptation algorithm improves QoE by 11.1% compared to the long video adaptation algorithm.

Emerging cross-device artificial intelligence (AI) applications require a transition from conventional centralized learning systems toward large-scale distributed AI systems that can collaboratively perform complex learning tasks. In this regard, democratized learning (Dem-AI) lays out a holistic philosophy with underlying principles for building large-scale distributed and democratized machine learning systems. The outlined principles are meant to study a generalization in distributed learning systems that go beyond existing mechanisms such as federated learning (FL). Moreover, such learning systems rely on hierarchical self-organization of well-connected distributed learning agents who have limited and highly personalized data and can evolve and regulate themselves based on the underlying duality of specialized and generalized processes. Inspired by Dem-AI philosophy, a novel distributed learning approach is proposed in this article. The approach consists of a self-organizing hierarchical structuring mechanism based on agglomerative clustering, hierarchical generalization, and corresponding learning mechanism. Subsequently, hierarchical generalized learning problems in recursive forms are formulated and shown to be approximately solved using the solutions of distributed personalized learning problems and hierarchical update mechanisms. To that end, a distributed learning algorithm, namely DemLearn, is proposed. Extensive experiments on benchmark MNIST, Fashion-MNIST, FE-MNIST, and CIFAR-10 datasets show that the proposed algorithm demonstrates better results in the generalization performance of learning models in agents compared to the conventional FL algorithms. The detailed analysis provides useful observations to further handle both the generalization and specialization performance of the learning models in Dem-AI systems.

Abstract Real-time moving object detection is an emerging method in Industry 5.0, that is applied in video surveillance, video coding, human-computer interaction, IoT, robotics, smart home, smart environment, edge and fog computing, cloud computing, and so on. One of the main issues is accurate moving object detection in real-time in a video with challenging background scenes. Numerous existing approaches used multiple features simultaneously to address the problem but did not consider any adaptive/dynamic weight factor to combine these feature spaces. Being inspired by these observations, we propose a background subtraction-based real-time moving object detection method, called DFC-D. This proposal determines an adaptive/dynamic weight factor to provide a weighted fusion of non-smoothing color/gray intensity and non-smoothing gradient magnitude. Moreover, the color-gradient background difference and segmentation noise are employed to modify thresholds and background samples. Our proposed solution achieves the best trade-off between detection accuracy and algorithmic complexity on the benchmark datasets while comparing with the state-of-the-art approaches.

Abstract Background subtraction used in object detection, tracking and action recognition is a typical method that separates foreground objects from the background. These applications require accuracy and a complexity reduction technique. Some approaches have been proposed to either increase accuracy or decrease complexity. However, the trade‐off between increasing accuracy and reducing the complexity of background subtraction is a big challenge. To address this issue, a background subtraction‐based real‐time moving object‐detection approach is proposed. The key contribution in authors' proposal is to use a colour image and a novel colour‐gradient blending fused image to achieve accurate background/foreground segmentation. The fused image is a combination of a gradient image and a colour image to correct illumination variations and preserve the edge information. Also, thresholds are adaptively selected based on the dynamic background behaviour to attain a more robust classification system. The proposed model based on real‐time and complex videos from the CD‐2012 and CD‐2014 change detection data sets, and the CMD data set is evaluated. Experimental results indicate that authors' method processes around 43 frames per second and requires six bytes of memory per pixel, which is noticeably more efficient and less complex than other background subtraction methods.

Multiaccess edge computing (MEC) enables autonomous vehicles to handle time-critical and data-intensive computational tasks for emerging Internet-of-Vehicles (IoV) applications via computation offloading. However, a massive amount of data generated by colocated vehicles is typically redundant, introducing a critical issue due to limited network bandwidth. Moreover, on the edge server side, these computation-intensive tasks further impose severe pressure on the resource-finite MEC server, resulting in low-performance efficiency of applications. To solve these challenges, we model the data redundancy and collaborative task computing scheme to efficiently reduce the redundant data and utilize the idle resources in nearby MEC servers. First, the data redundancy problem is formulated as a set-covering problem according to the spatiotemporal coverage of captured images. Next, we exploit the submodular optimization technique to design an efficient algorithm to minimize the number of images transferred to the MEC servers without degrading the quality of IoV applications. To facilitate the task execution in the MEC server, we then propose a collaborative task computing scheme, where an MEC server intentionally encourages nearby resource-rich MEC servers to participate in a collaborative computing group. Accordingly, a cost model is formulated as an optimization problem, the objective of which is to prompt the MEC server to judiciously allocate computing tasks to nearby MEC servers with the goal of achieving the minimal cost while the latency of tasks is guaranteed. Experimental results show that the proposed scheme can efficiently mitigate data redundancy, conserve network bandwidth consumption, and achieve the lowest cost for processing tasks.

In recent years, multi-access edge computing (MEC) is a key enabler for handling the massive expansion of Internet of Things (IoT) applications and services. However, energy consumption of a MEC network depends on volatile tasks that induces risk for energy demand estimations. As an energy supplier, a microgrid can facilitate seamless energy supply. However, the risk associated with energy supply is also increased due to unpredictable energy generation from renewable and non-renewable sources. Especially, the risk of energy shortfall is involved with uncertainties in both energy consumption and generation. In this article, we study a risk-aware energy scheduling problem for a microgrid-powered MEC network. First, we formulate an optimization problem considering the conditional value-at-risk (CVaR) measurement for both energy consumption and generation, where the objective is to minimize the expected residual of scheduled energy for the MEC networks and we show this problem is an NP-hard problem. Second, we analyze our formulated problem using a multi-agent stochastic game that ensures the joint policy Nash equilibrium, and show the convergence of the proposed model. Third, we derive the solution by applying a multi-agent deep reinforcement learning (MADRL)-based asynchronous advantage actor-critic (A3C) algorithm with shared neural networks. This method mitigates the curse of dimensionality of the state space and chooses the best policy among the agents for the proposed problem. Finally, the experimental results establish a significant performance gain by considering CVaR for high accuracy energy scheduling of the proposed model than both the single and random agent models.

Online workload balancing guarantees that the incoming workloads are processed to the appropriate servers in real time without any knowledge of future resource requests. Currently, by matching the characteristics of incoming Internet of Things (IoT) applications to the current state of computing and networking resources, a mobile edge orchestrator (MEO) provides high‐quality service while temporally and spatially changing the incoming workload. Moreover, a fuzzy‐based MEO is used to handle the multicriteria decision‐making process by considering multiple parameters within the same framework in order to make an offloading decision for an incoming task of an IoT application. In a fuzzy‐based MEO, the fuzzy‐based offloading strategy leads to unbalanced loads among edge servers. Therefore, the fuzzy‐based MEO needs to scale its capacity when it comes to a large number of devices in order to avoid task failures and to reduce service times. In this paper, we investigate and propose an online workload balancing algorithm, which we call the fuzzy‐based (FuB) algorithm, for a fuzzy‐based MEO. By considering user configuration requirements, server geographic locations, and available resource capacities for achieving an online algorithm, our proposal allocates the proximate server for each incoming task in real time at the MEO. A simulation was conducted in augmented reality, healthcare, compute‐intensive, and infotainment applications. Compared to two benchmark schemes that use the fuzzy logic approach for an MEO in IoT environments, the simulation results (using EdgeCloudSim) show that our proposal outperforms the existing algorithms in terms of service time, the number of failed tasks, and processing times when the system is overloaded.

Human activity recognition (HAR), driven by large deep learning models, has received a lot of attention in recent years due to its high applicability in diverse application domains, manipulate time-series data to speculate on activities. Meanwhile, the cloud term "as-a-service" has essentially revolutionized the information technology industry market over the last ten years. These two trends somehow are incorporating to inspire a new model for the assistive living application: HAR as a service in the cloud. However, with frequently updates deep learning frameworks in open source communities as well as various new hardware features release, which make a significant software management challenge for deep learning model developers. To address this problem, container techniques are widely employed to facilitate the deep learning software development cycle. In addition, models and the available datasets are being larger and more complicated, and so, an expanding amount of computing resources is desired so that these models are trained in a feasible amount of time. This requires an emerging distributed training approach, called data parallelism, to achieve low resource utilization and faster execution in training time. Therefore, in this paper, we apply the data parallelism to build an assistive living HAR application using LSTM model, deploying in containers within a Kubernetes cluster to enable the real-time recognition as well as prediction of changes in human activity patterns. We then systematically measure the influence of this technique on the performance of the HAR application. Firstly, we evaluate our system performance with regard to CPU and GPU when deployed in containers and host environment, then analyze the outcomes to verify the difference in terms of the model learning performance. Through the experiments, we figure out that data parallelism strategy is efficient for improving model learning performance. In addition, this technique helps to increase the scaling efficiency in our system.

In the edge cloud, when two types of service providers provide edge nodes, each service provider has different logging areas, making it difficult to access QoS violations and complying with service level agreements (SLAs). Thus, the third-party broker plays a role of confirming QoS violations and compliance between the two providers, thereby securing reliability between the CSP and NP. In addition, by improving the existing logging technique in the edge cloud with limited resources, it reduces the CPU usage to improve the availability of cloud services. Therefore, we propose a broker-based framework that improves the service availability by reducing the CPU usage and secures QoS compliance and violations.

Mobile edge computing (MEC) is used to offload services (tasks) from cloud computing in order to deliver those services to mobile Internet of Things (IoT) devices near mobile edge nodes. However, even though there are advantages to MEC, we face many significant problems, such as how a service provider (SP) deploys requested services efficiently on a destination MEC node, and how to discover existing services in neighboring MEC nodes to save edge resources. In this paper, we present a decentralized and revised content-centric networking (CCN)-based MEC service deployment/discovery protocol and platform. We organized a gateway in every area according to a three-tiered hierarchical MEC network topology to reduce computing overhead at the centralized controller. We revised CCN to introduce a protocol to help SP deploy their service on MEC node and assist MEC node discover services in neighboring nodes. By using our proposed protocol, MEC nodes can deploy or discover the requested service instances in the proximity of IoT devices to reduce transmission delay. We also present a mathematical model to calculate the round trip time to guarantee quality of service. Numerical experiments measure the performance of our proposed method with various mobile IoT device services. The results show that the proposed service deployment protocol and platform reduce the average service delay by up to 50% compared to legacy cloud. In addition, the proposed method outperforms the legacy protocol of the MEC environment in service discovery time.

Edge Cloud Computing is emerging as an efficient paradigm that aims at providing responsive and real-time services. By delivering the cloud services to the edges of network in the proximity to the users, the latency of transmission time can be reduced and the heavy burden on the backhaul link are avoided. However, how a third party service provider could deploy the cloud services on a specific edge nodes provided by Infrastructure Provider, how to manage cloud services resources as well as to find the existed cloud services over the network are significant issues in Edge Cloud Computing. In this paper, we introduce an autonomic edge cloud platform with CCN-based cloud service routing protocol (C-SRP) to provide a self-control mechanism for cloud services in Edge Cloud Computing. By integrating with tables defined in Content Centric Networking (CCN), C-SRP has capability of self-deploying cloud services on edge nodes, and self-discovering the cloud services over the network. We implement the autonomic edge cloud platform as well as C-SRP and measure the performance by various cloud services. The results show that our C-SRP is able to reduce the average service delay up to 50% compared with legacy cloud. In addition, it also performs service routing better than the centralized controller protocol in Edge Cloud Computing environment.

Phishing detection is recognized as a criminal issue of Internet security. By deploying a gateway anti-phishing in the networks, these current hardware-based approaches provide an additional layer of defense against phishing attacks. However, such hardware devices are expensive and inefficient in operation due to the diversity of phishing attacks. With promising technologies of virtualization in fog networks, an anti-phishing gateway can be implemented as software at the edge of the network and embedded robust machine learning techniques for phishing detection. In this paper, we use uniform resource locator features and Web traffic features to detect phishing websites based on a designed neuro-fuzzy framework (dubbed Fi-NFN). Based on the new approach, fog computing as encouraged by Cisco, we design an anti-phishing model to transparently monitor and protect fog users from phishing attacks. The experiment results of our proposed approach, based on a large-scale dataset collected from real phishing cases, have shown that our system can effectively prevent phishing attacks and improve the security of the network.

Videos can portray our thinking in a better way than anything can. As a result, online as well as offline videos in entertainment, health, storytelling, education and in surveillance are increasing day by day which further initiates video coding research to decrease the size while keeping satisfactory quality. Due to the ubiquitous necessity of video encoding, a fully functional, high performance and with easy licensing codec have been sought from industry and academia for a long time. In this paper, we investigate the performance of the software encoders of H.264/AVC, H.265/HEVC, VP9 and the new royalty free encoder AV1 from AOMedia for relatively small video files. To ensure fair analysis, we performed extensive experiments by varying different parameters. With default parameters, for the same video bitrates, AV1 obtains better quality as compared to other encodings. However, when we use the slowest time preset `placebo' for H.264 and H.265, AV1 goes slightly behind H.265. Among the other two, VP9 always outperforms H.264. For all schemes, as quality increases, the encoding time also increase.

In recent years, with the advent of the Internet of Things (IoT), fog computing is introduced as a powerful complement to the cloud to handle the IoT's data and communications needs. The interplay and cooperation between the edge (fog) and the core (cloud) has recently received considerable attention. In this paper, we consider task scheduling in a cloud-fog computing system, where a fog provider can exploit the collaboration between its own fog nodes and the rented cloud nodes for efficiently executing users' large-scale offloading applications. We first formulate the task scheduling problem in such cloud-fog environment and then propose a heuristic-based algorithm, whose major objective is achieving the balance between the makespan and the monetary cost of cloud resources. The numerical results show that our proposed algorithm achieves better tradeoff value than other existing algorithms.

Desktop as a service (DaaS) enables people to connect their virtual remote desktops with their on-desk computers, laptops, and other mobile devices. To best serve the users, it is essential for the service providers to know the perceived quality of their system by the customers and users while preserving their business objectives. This paper deals with QoE aware desktop delivery solution for DaaS. Based on our previous works, first we derive one to one exponential relationship models between quality of service (QoS) and different objective qualities in desktop delivery, where screen sizes in pixels unit are taken as the QoS. Then, individual models are combined into an integrated quality of experience (QoE) model, that can quantify total QoE on given QoS values. With the derived QoE model, we propose an adaptive desktop delivery scheme for DaaS. For a given desktop size, the scheme first measures the QoE scores for its two modules and automatically assigns the most appropriate modules to different desktop regions. If the server lacks resources to accomplish the user requirements with good QoE then the scheme suggests the users to reduce their requirements. Simulation results on some selected scenarios explain the uses and effectiveness of the proposed scheme.

Even though demand response of data centers has attracted many studies, there are very limited attempts on an important segment: colocation datacenters. Unlike large-scale (Google-type) datacenters, the colocation operator lacks control over its tenant servers, which entails a special interest in a design of incentive mechanisms, such that the operator can coordinate tenants to reduce the power usage for demand response. However, most previous studies ignore the role of the demand response provider (DRP), who uses pricing signals as a guide for customer response and as a compensation for their cutting electricity usage. To address this oversight, we propose an incentive mechanism Reward-to-Reduce for colocation's economic demand response, which shows an interaction between the DRP compensation to the colocation operator, and the colocation operator reward to tenants. Observing that this interaction contains strategic behaviors, we first formulate a two-stage Stackelberg game, where we show a unique competitive equilibrium of the operator strategy in the second stage, and a nonconvex problem of finding the optimal DRP compensation price in the first stage. We next analyze the second-stage equilibrium using an exact analysis and design an algorithm that can efficiently search the first-stage optimal DRP price with a reduced search space. Since the exact analysis can be impractical due to required tenants' private information, we also propose an approximate approach with limited tenant information. Extensive case studies show that the approximate approach can have the same performance as the exact analysis in a wide array of case studies and the optimal DRP price can be determined effectively, with which the corresponding DRP individual cost is compared with the social cost.

The number of applications based on Apache Hadoop is dramatically increasing due to the robustness and dynamic features of this system. At the heart of Apache Hadoop, the Hadoop Distributed File System (HDFS) provides the reliability and high availability for computation by applying a static replication by default. However, because of the characteristics of parallel operations on the application layer, the access rate for each data file in HDFS is completely different. Consequently, maintaining the same replication mechanism for every data file leads to detrimental effects on the performance. By rigorously considering the drawbacks of the HDFS replication, this paper proposes an approach to dynamically replicate the data file based on the predictive analysis. With the help of probability theory, the utilization of each data file can be predicted to create a corresponding replication strategy. Eventually, the popular files can be subsequently replicated according to their own access potentials. For the remaining low potential files, an erasure code is applied to maintain the reliability. Hence, our approach simultaneously improves the availability while keeping the reliability in comparison to the default scheme. Furthermore, the complexity reduction is applied to enhance the effectiveness of the prediction when dealing with Big Data.

Media content in its digital form has been rapidly scaling up, resulting in popularity gain of cloud computing. Cloud computing makes it easy to manage the vastly increasing digital content. Moreover, additional features like, omnipresent access, further service creation, discovery of services, and resource management also play an important role in this regard. The forthcoming era is interoperability of multiple clouds, known as cloud federation or inter-cloud computing. With cloud federation, services would be provided through two or more clouds. Once matured and standardized, inter-cloud computing is supposed to provide services which would be more scalable, better managed, and efficient. Such tasks are provided through a middleware entity called cloud broker. A broker is responsible for reserving resources, managing them, discovering services according to customer's demands, Service Level Agreement (SLA) negotiation, and match-making between the involved service provider and the customer. So far existing studies discuss brokerage in a narrow focused way. In the research outcome presented in this paper, we provide a holistic brokerage model to manage on-demand and advance service reservation, pricing, and reimbursement. A unique feature of this study is that we have considered dynamic management of customer's characteristics and historical record in evaluating the economics related factors. Additionally, a mechanism of incentive and penalties is provided, which helps in trust build-up for the customers and service providers, prevention of resource underutilization, and profit gain for the involved entities. For practical implications, the framework is modeled on Amazon Elastic Compute Cloud (EC2) On-Demand and Reserved Instances service pricing. For certain features required in the model, data was gathered from Google Cluster trace.

Pervasive and ubiquitous computing services have recently been under focus of not only the research community, but developers as well. Prevailing wireless sensor networks (WSNs), Internet of Things (IoT), and healthcare related services have made it difficult to handle all the data in an efficient and effective way and create more useful services. Different devices generate different types of data with different frequencies. Therefore, amalgamation of cloud computing with IoTs, termed as Cloud of Things (CoT) has recently been under discussion in research arena. CoT provides ease of management for the growing media content and other data. Besides this, features like: ubiquitous access, service creation, service discovery, and resource provisioning play a significant role, which comes with CoT. Emergency, healthcare, and latency sensitive services require real-time response. Also, it is necessary to decide what type of data is to be uploaded in the cloud, without burdening the core network and the cloud. For this purpose, Fog computing plays an important role. Fog resides between underlying IoTs and the cloud. Its purpose is to manage resources, perform data filtration, preprocessing, and security measures. For this purpose, Fog requires an effective and efficient resource management framework for IoTs, which we provide in this paper. Our model covers the issues of resource prediction, customer type based resource estimation and reservation, advance reservation, and pricing for new and existing IoT customers, on the basis of their characteristics. The implementation was done using Java, while the model was evaluated using CloudSim toolkit. The results and discussion show the validity and performance of our system.

Still lacking a standard architecture, cloud computing requires sophisticated ways to estimate resources for the requesting cloud service customers (CSCs). CSC show random behavior in utilizing various services. In this regard, if all the CSCs are treated in the same way, not only cloud service providers (CSPs) suffer because of uctuating utilization behavior of CSCs, but also CSCs suffer, since they do not get any incentive for their loyalty. We propose a dynamic resource estimation method, taking into account CSCs historical record of service utilization or relinquish. With the intent of showing practical implications of our method, we implemented it using Amazon EC2 pricing. Based on various services, differentiated through Amazon's price plans, and historical record of CSCs, the model determines resources to be allocated. More loyal CSC gets better service, while for the contrary case, CSP reserves resources cautiously.

This research discusses a system architecture that comes up with potentially better resiliency and faster recovery from failures based on the renowned genetic algorithm. Additionally, we aim to achieve a globally optimized performance as well as a service solution that can remain financially and operationally balanced according to customer preferences. The proposed methodology has undergone numerous and severe evaluations to be proclaimed of their effectiveness and efficiency, even when put under tight comparison with other existing work.

In the area of cloud infrastructure environment, the management tool to monitor and control the cloud resources is the important factor that can drive the cost benefit of the cloud vendors. But most these tools are bundled within the high cost commercial platforms and are optimized to run on desktop computers. With the vision that Mobile Cloud Computing will be the future technology paradigm that dominates the IT industry, we want to create a cloud management tool that is open source, fast, lightweight and mobile friendly. We take the initial steps by implementing our framework using several popular technologies such as RESTful, Java Message Service, JSON, and we call it “High performance and Lightweight Mobile Cloud Infrastructure Monitor and Benchmark Service” or HiLiCloud. The initial testings show competitive evaluation results.

In remote display protocols, screen image compression plays an important role to improve quality of experience (QoE) of users and reduce the bandwidth consumption. Not all image elements on the display have the same type, so it is wise to apply a screen image classification for compression decision. In this paper, we propose a novel efficient approach for screen image classification that separate the captured screen into 2 types of blocks: text and non-text block. Our method is a 2-stage process that is different from other works because of the appearance of text localization in the screen image as the first stage. This text localization is mainly based on edge feature and morphological operation in which we experiment with many kinds of edge detection methods. Then block-based classification categorizes the screen image blocks based on the positions of the detected text regions. The experimental results of high accuracy rate and low time consumption state our method is efficient in remote display protocol.

Abstract An underlying cloud computing feature, outsourcing of resources, makes the service‐level agreement (SLA) a critical factor for quality of service (QoS), and many researchers have addressed the question of how an SLA can be evaluated. Lately, security SLAs have also received much attention to guarantee security in a user perspective and provide optimal and efficient security service in the security paradigm shifting by cloud computing, such as security as a service. The quantitative measurement of security metrics is a considerably difficult problem and might be considered one of the multi‐dimensional aspects of security threats. To address these issues, we provide a novel cooperative security‐SLA evaluation model for the personal cloud service environment including a multi‐dimensional approach to analyze security threats depending on services type as well as a cooperative model to reach a general consensus of priorities, that is, indicators depending on services type and security metrics based on cloud brokers. Copyright © 2014 John Wiley & Sons, Ltd.

The rapid increase in digital content, specially multimedia, calls now for standardization of Media Cloud and Inter-Cloud computing, for better provisioning of services. Inter-Cloud computing faces some key challenges in terms of handling multimedia, which are discussed in this paper along with our research status towards their solutions. We also present Inter-Cloud basic architecture and Media Cloud storage design considerations. Some key findings on storage heterogeneity are also part of this paper.

In-memory Computing (IMC) is the new trend for enabling high-performance computation and fast data processing. It is currently being used for large enterprises, e-commerce shops who need real-time interactions, low latency responses and instant results. Given the enhancement of the IMC, we apply this new paradigm to the Virtual Desktop Environment (VDE), and look into the performance differences in comparison with traditional VDE. The end results shows positive feedback, but there are trade-offs we need to concern for the In-memory Virtual Desktop Environment.

Rapidly increasing digital media has triggered the importance of cloud computing. Cloud computing provides ease of management and ubiquitous access facility to the growing digital content. The ratio with which digital content has been increasing, it now requires multiple clouds to interoperate for the purpose of scalability, efficient service provisioning, and better management. This scenario is known as inter-cloud computing or cloud federation. One of the key entities in inter-cloud computing is cloud broker. Cloud broker is a match-maker between the service provider and the customer. Broker plays its role to reserve resources and perform pricing and billing. Service providers experience customers having different traits and characteristics. Some tend to utilize all of the resources, while others may quit in between, due to various reasons. Based on the characteristics of customers, their resources are predicted and pricing is performed. We have proposed a model which addresses this issue by determining resources and prices, on the basis of historical record of each customer. We have implemented our model using Java / Net Beans 8.0 and tested on Cloud Sim 3.0.3 toolkit. The results presented here justify and endorse our model.

Virtual machine (VM) migration has recently emerged as a potential tool to improve the performance, cost and fault tolerance of data centers. However, migrating VMs requires more hardware resources to reserve target slots during the migration. A large number of simultaneous migrations can exhaust the physical resources of a data center. Consequently, the system may deny new service requests. In this paper, we introduce an efficient strategy to limit the number of target slots used for the migration process with the aim of maintaining high data center availability. In particular, the number of target slots is determined by the tradeoff between the waiting time of the queued migration processes and the probability that the system rejects new user requests due to the lack of available resources. We also propose a method to reduce the waiting time of the migration process by adding a sufficient number of slots during the migration process. Accordingly, a data center can use fewer resources while still guaranteeing the quality of the service. The efficiency of our proposed method is demonstrated by the experimental results.

As we know, cloud computing has already become a dominant computing paradigm with its significant impact on the distribution of computing resources and business attribution. However, due to the lack of fairness in pricing and SLA violation during the service in cloud computing, customers are dissatisfied with the existing pricing model of current cloud computing environment, and there have an immense chance to lose those discontented customers. Moreover, the cloud service provider's total business can be obstructed in the long run. Therefore, we wish to provide a solution that can preserve customers' appreciation by refundable service. In this paper, we propose a unique technique that can boost up customer satisfaction and diminish cloud service provider anxiety for continuing their business. Basically, we applied a third party cloud broker that can handle all of the business procedure instead of cloud service provider. Our method offers refunding in case of unutilized resource as well service quality degradation that is service level agreement (SLA) violation. The experiment results demonstrate different refund amount for cloud consumers' considering with several attribute.

Applications that require timeliness such as; video conferencing, Voice over IP (VoIP), Video-on-Demand (VoD) etc., require quantified and well managed Quality of Service (QoS). For this, the flow label field in IPv6 header and the Label field in MPLS header should be used for efficient QoS provisioning. In this paper, flow label specifications were first investigated and then a new structure is proposed for QoS provisioning. The 20-bit field in IPv6 as well as MPLS is meant to provide QoS, other than labeling or identification purpose. But its usage is not standardized and defined that how these 20-bits must effectively be used to provide maximum possible QoS in a better way. This paper discusses it by proposing portions for QoS by explicitly specifying bandwidth, delay, and packet loss. By keeping the QoS parameters open in this way, it would be easy for a flow to decide what to reserve and how much to reserve. After that, mapping of flow label with class fields of IPv6 and MPLS is presented. In virtual networks, specially for Thin client, service quality degradation is a major issue, when IPv4-IPv6 virtual networks co-exist. This is also where this redefinition can come very handy.

Mobile Cloud Computing (MCC) is broadening the market for mobile devices with more and more variety and heavier services by using service images on cloud. With the limitation of battery life time, cpu and memory capacity, etc., mobile device can not adapt with those new services. And remote display solution becomes the key technology to remove the barrier of those limitations in MCC. Due to the scalability, mobility of mobile devices, changing in resource demand or network condition, service images must be managed in the most efficient way. In this paper, we formulate the service image placement problem as an optimization problem by minimizing the cost function that is the combination of composite cost and resource demand. We further propose the algorithms to address the service image placement issue and do simulation to evaluate our proposed algorithms. By the results of experiment, we show that our approach can improve the network traffic and the quality of service than others.

ABSTRACT Because of the resource constraints on sensor networks, many recent key management schemes exploit the location of nodes in order to establish pairwise keys. Despite these efforts, location discovery for sensor networks is very difficult because existing schemes have failed to consider changes in the signal range and the deployment error. As a result, network performance may accidentally decrease, whereas costs increase. In this paper, we used the theory of the signal range and deployment error knowledge to analyze sensor nodes location information. Combining with the probabilistic pre‐distribution method, we proposed a scheme that provides high connectivity and secure communication with better communication overhead by defining an efficient number of adjacent cells and determining the adequate length of the cell. With such optimistic results, novel deployment knowledge is also expected to provide superior performance with different types of key distribution schemes. Copyright © 2011 John Wiley & Sons, Ltd.

Distributed resource allocation is a very important and complex problem in emerging horizontal dynamic cloud federation (HDCF) platform. The HDCF platform differs from the existing vertical supply chain federation (VSCF) models in terms of establishing federation and dynamic pricing. There is a need to develop algorithms that can capture this complexity yet can be easily implemented and used to solve distributed resource allocation problem in HDCF platform. In this paper, we propose a game-theoretic solution to this problem that ensures mutual benefits so that the cloud providers (CPs) are encouraged to form a HDCF platform. We study two resource allocation games - cooperative and non-cooperative games to analyze interaction among CPs in a HDCF environment. Also both centralized and distributed algorithms are presented to find optimal solutions which have low overhead and robust performance. Various simulations were carried out to validate and verify the effectiveness of the proposed resource allocation games.

Ubiquitous technology through sensor networks is being applied to numerous industrial fields specially to increase the quality of human life (QoL). Therefore, Wireless Sensor Networks (WSNs) lossless data is one of the communications challenges to provide accurate data. Although, end-to-end data retransmission has evolved as a reliable transportation in Internet, this method is not applicable to WSNs due to the lack of reliability of wireless link and resource constraints in sensor nodes. In our previous paper, we proposed a reliable data transfer using path-reliability and implicit ACK called RTOD on WSN. However, path reliability calculation components such as RSSI, channel error rate, number of transmission in RTOD method have not been studied thoroughly. In this paper, we analysis path reliability components and simulate by using NS-2. Moreover, we propose limited number of transmission method (LTM) for WSNs. Proposed scheme shows average 4.1% fault tolerance.

Cloud computing is an evolving term these days. It describes the advance of many existing IT technologies and separates application and information resources from the underlying infrastructure. Personal Cloud is the hybrid deployment model that is combined private cloud and public cloud. By and large, cloud orchestration does not exist today. Current cloud service is provided by web browser or host installed application directly. According to the ITU-T draft, we might consider cloud orchestration environment in collaboration with other cloud providers. Previous work proposed security framework that has limitation of scalability for cloud orchestration. In this paper, we analyze security threats and requirements for previous researches and propose service model and security framework which include related technology for implementation and are possible to provide resource mobility.

This paper presents a Secured Wireless Sensor Network-integrated Cloud computing for u-Life Care (SC3). SC3 monitors human health, activities, and shares information among doctors, care-givers, clinics, and pharmacies in the Cloud, so that users can have better care with low cost. SC3 incorporates various technologies with novel ideas including; sensor networks, Cloud computing security, and activities recognition.

To address interoperability and scalability issues for cloud computing, in our previous paper, we presented a novel cloud market model called CACM that enables a dynamic collaboration (DC) platform among different Cloud providers. As the initiator of dynamic collaboration, primary Cloud provider (pCP) needs an efficient local task selection and allocation algorithm to partition the whole tasks and allocate those tasks to be executed locally. Existing task allocation algorithms cannot be directly applicable in a DC environment since they may cause low resource utilization of local resources. So in this paper we propose a general task selection and allocation framework to improve resource utilization for pCP. The framework utilizes an adaptive filter to select tasks and a modified heuristic algorithm to allocate tasks. Moreover, a trade-off metric is developed as the optimization goal of heuristic algorithm, so that it is able to manage and optimize the trade-off between QoS of tasks and utilization of resources.

Abstract In recent years, Wireless Sensor Networks (WSNs) have demonstrated successful applications for both civil and military tasks. However, sensor networks are susceptible to multiple types of attacks because they are randomly deployed in open and unprotected environments. It is necessary to utilize effective mechanisms to protect sensor networks against multiple types of attacks on routing protocols. In this paper, we propose a lightweight intrusion detection framework integrated for clustered sensor networks. Furthermore, we provide algorithms to minimize the triggered intrusion modules in clustered WSNs by using an over‐hearing mechanism to reduce the sending alert packets. Our scheme can prevent most routing attacks on sensor networks. In in‐depth simulation, the proposed scheme shows less energy consumption in intrusion detection than other schemes. Copyright © 2009 John Wiley & Sons, Ltd.

In this paper, we propose a promising multi-objective (MO) optimization model for partner selection in a market-oriented dynamic collaboration (DC) platform of cloud providers (CPs) to minimize the conflicts among providers that may happen when negotiating among providers. The model not only uses their individual information (INI) but also past collaborative relationship information (PRI) for partner selection which is seldom considered in existing approaches. A multi-objective genetic algorithm (MOGA) called MOGA-IC is also proposed to solve the model as the model is NP-hard. The algorithm is developed using two popular MOGAs- NSGAII and SPEA2. The experimental results show that MOGAIC with NSGA-II outperformed the MOGA-IC with SPEA2 in finding useful Pareto optimal solution sets. In addition, other simulation experiments are conducted to verify the effectiveness of the MOGA-IC in terms of satisfactory partner selection and conflicts minimization.

Privacy issue is receiving a great deal of attention since the need of privacy is increasing and new threats are emerging. The growing concern of users for their personal information has made it critical to implant effective technologies for privacy and data management. A common way for privacy preservation is restricting access to data like the classic Role-based Access Control (RBAC) Model. But the RBAC is limited as it does not provide users enough flexibilities and functionalities. In order to minimize the disclosure of data and support higher flexibilities for users to manage their privacy information, this paper provides a privacy data graph based on the traditional RBAC model to illustrate the linkage between data elements. Moreover, the notion of purpose is added to specify the intended usage of data and allow users to set personal privacy preferences through purpose. A case study in the healthcare domain is provided. As our model is generic, it can be also adapted to other fields. A detailed view of our proposed privacy system with experimental result is provided.

Wireless sensor networks have many potential applications for both civil and military tasks. However, WSNs are susceptible to many types of attacks because they are deployed in open and unprotected environment. Selective forwarding attack is one of the easiest implement and damaged attacks in multi-hop routing protocols. In this paper, we proposed a lightweight detection algorithm based only on the neighborhood information. Our detection algorithm can detect selective forwarding attack with high accuracy and little overhead imposed on detection modules than previous works.

Some classes of real-time systems operate in environments that cannot be modeled with static approaches. In such an environment, we neither have a priori knowledge about the system workload, nor is it possible to have a priori knowledge about worst-case execution time (WCET). There is no guarantee that all the deadlines of the periodic tasks will be met by using the rate monotonic analysis (RMA) approach. This paper presents an empirical and space-efficient way to predict the execution time. Experimental results have shown that for the dynamic real-time system, the best approach is a combination of static system profiling and dynamic prediction.