Mobile edge caching (MEC) is a promising technique to improve the quality of service (QoS) for mobile users (MU) by bringing data to the network edge. However, optimizing the crucial QoS aspects of message freshness and service promptness, measured by age of information (AoI) and service delay, respectively, entails a tradeoff due to their competition for shared edge resources. This article investigates this tradeoff by formulating their weighted sum minimization as a sequential decision-making problem, incorporating high-dimensional, discrete-valued, and linearly constrained design variables. First, to assess the feasibility of the considered problem, we characterize the corresponding achievable region by deriving its superset with the rate stability theorem and its subset with a novel stochastic policy, and develop a sufficient condition for the existence of solutions. Next, to efficiently solve this problem, we propose a mixed-order drift-plus-penalty algorithm by jointly considering the linear and quadratic Lyapunov drifts and then optimizing them with dynamic programming (DP). Finally, by leveraging the Lyapunov optimization technique, we demonstrate that the proposed algorithm achieves an$O(1/V)$versus$O(V)$tradeoff for the average AoI and average service delay.

Federated learning (FL) enables collaborative training of a model while keeping the training data decentralized and private. However, in Internet of Things systems, inherent heterogeneity in processing power, communication bandwidth, and task size can significantly hinder the efficient training of large models. Such heterogeneity would render vast variations in the training time of clients, lengthening overall training and wasting resources of faster clients. To tackle these heterogeneity challenges, we propose dynamic tiering-based FL (DTFL), a novel system that leverages distributed optimization principles to improve the edge learning performance. Based on clients’ resources, DTFL dynamically offloads part of the global model to the server, alleviating resource constraints on slower clients and speeding up training. By leveraging split learning, DTFL offloads different portions of the global model to clients in different tiers and enables each client to update the models in parallel via local-loss-based training. This helps reduce the computation and communication demand on resource-constrained devices, mitigating the straggler problem. DTFL introduces a dynamic tier scheduler that uses tier profiling to estimate the expected training time of each client based on their historical training time, communication speed, and dataset size. The dynamic tier scheduler assigns clients to suitable tiers to minimize the overall training time in each round. We theoretically prove the convergence properties of DTFL and validate its effectiveness by training large models (ResNet-56 and ResNet-110) across varying numbers of clients (from 10 to 200) using the popular image datasets (CIFAR-10, CIFAR-100, CINIC-10, and HAM10000) under both I.I.D and non-I.I.D systems. DTFL seamlessly integrates various privacy measures without sacrificing performance. Extensive experimental results show that compared with state-of-the-art FL methods, DTFL can significantly reduce the training time by up to 80% while maintaining the model accuracy.

Mobile edge caching (MEC) is a promising technique to improve the quality of service (QoS) for mobile users (MU) by bringing data to the network edge. However, optimizing the crucial QoS aspects of message freshness and service promptness, measured by age of information (AoI) and service delay, respectively, entails a tradeoff due to their competition for shared edge resources. This paper investigates this tradeoff by formulating their weighted sum minimization as a sequential decision-making problem, incorporating high-dimensional, discrete-valued, and linearly constrained design variables. First, to assess the feasibility of the considered problem, we characterize the corresponding achievable region by deriving its superset with the rate stability theorem and its subset with a novel stochastic policy, and develop a sufficient condition for the existence of solutions. Next, to efficiently solve this problem, we propose a mixed-order drift-plus-penalty algorithm by jointly considering the linear and quadratic Lyapunov drift and then optimizing them with dynamic programming (DP). Finally, by leveraging the Lyapunov optimization technique, we demonstrate that the proposed algorithm achieves an O(1/V) versus O(V) tradeoff for the average AoI and average service delay.

Decentralized Multi-agent Learning (DML) enables collaborative model training while preserving data privacy. How-ever, inherent heterogeneity in agents' resources (computation, communication, and task size) may lead to substantial variations in training time. This heterogeneity creates a bottleneck, lengthening the overall training time due to straggler effects and potentially wasting spare resources of faster agents. To minimize training time in heterogeneous environments, we present a Communication-Efficient Training Workload Balancing for Decentralized Multi-Agent Learning (ComDML), which balances the workload among agents through a decentralized approach. Leveraging local-loss split training, ComDML enables parallel updates, where slower agents offload part of their workload to faster agents. To minimize the overall training time, ComDML optimizes the workload balancing by jointly considering the communication and computation capacities of agents, which hinges upon integer programming. A dynamic decentralized pairing scheduler is developed to efficiently pair agents and determine optimal offloading amounts. We prove that in ComDML, both slower and faster agents' models converge, for convex and non-convex functions. Furthermore, extensive experimental results on popular datasets (CIFAR-10, CIFAR-100, and CINIC-10) and their non-I.I.D. variants, with large models such as ResNet-56 and ResNet-110, demonstrate that ComDML can significantly reduce the overall training time while maintaining model accuracy, compared to state-of-the-art methods.ComDML demonstrates robustness in heterogeneous environments, and privacy measures can be seamlessly integrated for enhanced data protection.

Zero Redundancy Optimizer (ZeRO) has been used to train a wide range of large language models on massive GPU clusters due to its ease of use, efficiency, and good scalability. However, when training on low-bandwidth clusters, and/or when small batch size per GPU is used, ZeRO’s effective throughput is limited due to communication overheads. To alleviate this limitation, this paper introduces ZeRO++ composing of three communication volume reduction techniques (lowprecision all-gather, data remapping, and low-precision gradient averaging) to significantly reduce the communication volume up to 4x that enables up to 2.16x better throughput at 384 GPU scale. Our results also show ZeRO++ can speedup the RLHF by 3.3x compared to vanilla ZeRO. To verify the convergence of ZeRO++, we test up to 13B model for pretraining with 8/6-bits all gather and up to 30B model for finetuning with 4/2-bits all gather, and demonstrate on-par accuracy as original ZeRO (aka standard training). As a byproduct, the model trained with ZeRO++ is naturally weight-quantized, which can be directly used for inference without post-training quantization or quantization-aware training.

Recent advancements in privacy-preserving artificial intelligence (AI) have paved the way for enhanced privacy in computational processes. A standing challenge, however, is the robust privacy preservation in AI algorithms, especially when integrated into edge devices and Internet-of-Thing (IoT) infrastructures. Most prevailing solutions have adopted traditional encryption methods which, though secure, often introduce significant overhead and potential dips in accuracy. In this study, we put forth an innovative approach, utilizing the CKKS encryption scheme, aiming to harmoniously balance computational efficiency with stringent data privacy. By harnessing the capabilities of Full Homomorphic Encryption (FHE) under the CKKS scheme, we ensure the preservation of privacy, successfully curbing the inherent noise traditionally linked with accuracy reductions in similar encryption-oriented solutions. Through comprehensive experiments, our approach showcased its potential as a strong contender for privacy preservation, demonstrating commendable performance across all tests, affirming that FHE is indeed viable for devices with constrained computational power and energy resources.

First-come first-serve scheduling can result in substantial (up to 10%) of transiently idle nodes on supercomputers. Recognizing that such unfilled nodes are well-suited for deep neural network (DNN) training, due to the flexible nature of DNN training tasks, Liu et al. proposed that the re-scaling DNN training tasks to fit gaps in schedules be formulated as a mixed-integer linear programming (MILP) problem, and demonstrated via simulation the potential benefits of the approach. Here, we introduce MalleTrain, a system that provides the first practical implementation of this approach and that furthermore generalizes it by allowing it to be used even for DNN training applications for which model information is unknown before runtime. Key to this latter innovation is the use of a lightweight online job profiling advisor (JPA) to collect critical scalability information for DNN jobs---information that it then employs to optimize resource allocations dynamically, in real time. We describe the MalleTrain architecture and present the results of a detailed experimental evaluation on a supercomputer GPU cluster and several representative DNN training workloads, including neural architecture search and hyperparameter optimization. Our results not only confirm the practical feasibility of leveraging idle supercomputer nodes for DNN training but improve significantly on prior results, improving training throughput by up to 22.3% without requiring users to provide job scalability information.

While increasingly more applications are tempted to manage their data in decentralized systems, such as blockchains or distributed ledgers, the data exchange across multiple, potentially heterogeneous, decentralized systems remains an open problem: State-of-the-art protocols cannot meet one or more of the core requirements, such as atomicity, liveness, and scalability. Specifically, in the field of scientific computing, although a blockchain service was recently developed for scientific computing environments, the data exchanges and transactions among distinct ledgers are not supported. Observing that many modern scientific applications are collaborated on by multiple teams and the increasingly complicated (in-situ) workflows thereof, we argue that there is a pressing need to realize an efficient and scalable protocol for distinct ledgers to exchange data in scientific computing. This paper proposes a topological approach to enabling atomic, nonblocking, and scalable data exchanges among an arbitrary number of scientific ledgers in the context of collaborative scientific computing. Specifically, we construct a topological space formed by these ledgers—abstracting those nodes in a cross-ledger transaction as topological objects such as abstract simplex and simplicial complex. These topological objects, in turn, serve as the building blocks of a topological protocol, namely TopoCommit, under practical assumptions. We implement TopoCommit and integrate it into SciChain, a recently published distributed ledger for tracking scientific data provenance. The extensive evaluation of up to 1,008 nodes and 144 distinct ledgers on CloudLab shows that TopoCommit outperforms state-of-the-art protocols by up to 70×.

This article studies robust event classification using imperfect real-world phasor measurement unit (PMU) data. By analyzing the real-world PMU data, we find that it is challenging to directly use this data set for event classifiers due to the low data quality observed in PMU measurements and event logs. To address these challenges, we develop a novel machine learning framework for training robust event classifiers, which consists of three main steps: 1) data preprocessing; 2) fine-grained event data extraction; and 3) feature engineering. Specifically, the data preprocessing step addresses the data quality issues of PMU measurements (e.g., bad data and missing data); in the fine-grained event data extraction step, a model-free event detection method is developed to accurately localize the events from the inaccurate event timestamps in the event logs; and the feature engineering step constructs the event features based on the patterns of different event types, in order to improve the performance and the interpretability of the event classifiers. Based on the proposed framework, we develop a workflow for event classification using the real-world PMU data streaming into the system in real time. Using the proposed framework, robust event classifiers can be efficiently trained based on many off-the-shelf lightweight machine learning models. Numerical experiments using the real-world data set from the Western Interconnection of the U.S. power transmission grid show that the event classifiers trained under the proposed framework can achieve high classification accuracy while being robust against low-quality data.

Many applications deployed to public clouds are concerned about the confidentiality of their outsourced data, such as financial services and electronic patient records. A plausible solution to this problem is homomorphic encryption (HE), which supports certain algebraic operations directly over the ciphertexts. The downside of HE schemes is their significant, if not prohibitive, performance overhead for data-intensive workloads that are very common for outsourced databases, or database-as-a-serve in cloud computing. The objective of this work is to mitigate the performance overhead incurred by the HE module in outsourced databases. To that end, this paper proposes a radix-based parallel caching optimization for accelerating the performance of homomorphic encryption (HE) of outsourced databases in cloud computing. The key insight of the proposed optimization is caching selected radix-ciphertexts in parallel without violating existing security guarantees of the primitive/base HE scheme. We design the radix HE algorithm and apply it to both batch- and incremental-HE schemes; we demonstrate the security of those radix-based HE schemes by showing that the problem of breaking them can be reduced to the problem of breaking their base HE schemes that are known IND-CPA (i.e. Indistinguishability under Chosen-Plaintext Attack). We implement the radix-based schemes as middleware of a 10-node Cassandra cluster on CloudLab; experiments on six workloads show that the proposed caching can boost state-of-the-art HE schemes, such as Paillier and Symmetria, by up to five orders of magnitude.

In the wake of the Web 2.0, crowdsourcing has emerged as a promising approach to maintain a flexible workforce for human intelligence tasks. To stimulate worker participation, many reverse auction-based incentive mechanisms have been proposed. Designing auctions that discourage workers from cheating and instead encouraging them to reveal their true cost information has drawn significant attention. However, the existing efforts have been focusing on tackling individual cheating misbehaviors, while the scenarios that workers strategically form collusion coalitions and rig their bids together to manipulate auction outcomes have received little attention. To fill this gap, in this work we develop a$(t,p)$-collusion resistant scheme that ensures no coalition ofweighted cardinality$t$can improve its group utility by coordinating the bids at a probability of$p$. This paper takes into account the unique features of crowdsourcing, such as diverse worker types and reputations, in the design. The proposed scheme can suppress a broad spectrum of collusion strategies. Besides, desirable properties, including$p$-truthfulness and$p$-individual rationality, are also achieved. To provide a comprehensive evaluation, we first analytically prove our scheme's collusion resistance and then experimentally verify our analytical conclusion using a real-world dataset. Our experimental results show that the baseline scheme, where none of the critical properties is guaranteed, costs up to 20.1 times the optimal payment in an ideal case where no collusion exists, while our final scheme is merely 4.9 times the optimal payment.

Microelectrode arrays (MEAs) are physical devices widely used in various science and engineering fields. One common computational challenge when applying a high-density MEA (i.e., a larger number of wires, more accurate locations of abnormal cells) is how to efficiently compute those resistance values provided the nonlinearity of the system of equations with the unknown resistance values per the Kirchhoff law. This paper proposes an algebraic-topological model for MEAs such that we can identify the intrinsic parallelism that cannot be identified by conventional approaches. We implement a system prototype called Parma based on the proposed topological methodology. Experimental results show that Parma outperforms the state-of-the-practice in time, scalability and memory usage: the computation time is two orders of magnitude faster on up to 1,024 cores with almost linear scalability and the memory is much better utilized with proportionally less warm-up time with respect to the number of concurrent threads.

The combination of Internet of Things (IoT) and artificial intelligence (AI) technology plays an important role in many fields, especially in the field of psychology and medical treatment. This work is mainly to study an affective robotics that can serve humans emotionally based on the IoT and AI technology. The design of affective robotics is important to understand the underlying mechanisms of human behaviors in real life. These mechanisms mainly include human nonverbal behaviors and affective states, which are important but difficult to be precisely modeled. To address this challenge, we introduce a human–robot interaction (HRI) architecture, including emotion recognition, affective computing, emotion diagnosis, and emotion control. First, we propose a system model based on HRI between affective robotics and human, in order to enhance the emotional service. Then, we develop a dynamical model with affective computing and control, where we provide a mathematical formulation method based on stochastic differential equations to quantify the emotional state. Furthermore, we perform the dynamic behavior analysis of the existence, boundedness, and stability of the model solution comprehensively. Numerical results are provided to demonstrate the validity and feasibility of the proposed design techniques.

Mobile crowd sensing (MCS) arises as a promising data collection paradigm that leverages the power of ubiquitous mobile devices to acquire rich information regarding their surrounding environment. In many location-based sensing tasks, workers are required to associate their sensing reports with corresponding geographic coordinates. Such information leaves a trail of worker's historical location record which thus poses a severe threat to their location privacy. On the other hand, individual workers may perceive location privacy differently. Instead of following conventional solutions that aim to perfectly hide user privacy, this paper adopts a novel alternative approach. Auser-centriclocationprivacytrading framework, called ULPT, is constructed to facilitate location privacy trading between workers and the platform. Each worker can decide how much location privacy to disclose to the platform in an MCS task based on its own location privacy leakage budget$\xi$. The higher$\xi$is, the more privacy its reported location discloses. Accordingly, it receives higher payment from the platform as compensation. Besides, ULPT enables the platform to select a suitable set of winning workers to achieve desirable MCS service accuracy while taking into account of its budget limit and worker privacy requirements. For this purpose, a heuristic algorithm is devised with a bounded optimality gap. As formally proved in this manuscript, ULPT guarantees a series of nice properties, including$\xi$-privacy,$(\alpha, \beta)$-accuracy,budget feasibility. Moreover, both rigorous theoretical analysis and extensive simulations are conducted to evaluate tradeoffs among these three.

Large scale training requires massive parallelism to finish the training within a reasonable amount of time. To support massive parallelism, large batch training is the key enabler but often at the cost of generalization performance. Existing works explore adaptive batching or hand-tuned static large batching, in order to strike a balance between the computational efficiency and the performance. However, these methods can provide only coarse-grained adaption (e.g., at a epoch level) due to the intrinsic expensive calculation or hand tuning requirements. In this paper, we propose a fully automated and lightweight adaptive batching methodology to enable fine-grained batch size adaption (e.g., at a mini-batch level) that can achieve state-of-the-art performance with record breaking batch sizes. The core component of our method is a lightweight yet efficient representation of the critical gradient noise information. We open-source the proposed methodology by providing a plugin tool that supports mainstream machine learning frameworks. Extensive evaluations on popular benchmarks (e.g., CIFAR10, ImageNet, and BERT-Large) demonstrate that the proposed methodology outperforms state-of-the-art methodologies using adaptive batching approaches or hand-tuned static strategies in both performance and batch size. Particularly, we achieve a new state-of-the-art batch size of 78k in BERT-Large pretraining with SQuAD score 90.69 compared to 90.58 reported in previous state-of-the-art with 59k batch size.

The concept of smart grid has been introduced as a new vision of the conventional power grid to figure out an efficient way of integrating green and renewable energy technologies. In this way, Internet-connected smart grid, also called energy Internet, is also emerging as an innovative approach to ensure the energy from anywhere at any time. The ultimate goal of these developments is to build a sustainable society. However, integrating and coordinating a large number of growing connections can be a challenging issue for the traditional centralized grid system. Consequently, the smart grid is undergoing a transformation to the decentralized topology from its centralized form. On the other hand, blockchain has some excellent features which make it a promising application for the smart grid paradigm. In this article, we aim to provide a comprehensive survey on the application of blockchain in smart grid. As such, we identify the significant security challenges of smart grid scenarios that can be addressed by blockchain. Then, we present a number of blockchain-based recent research works presented in different literature addressing security issues in the area of smart grid. We also summarize several related practical projects, trials, and products that have emerged recently. Finally, we discuss essential research challenges and future directions of applying blockchain to smart grid security issues.

Recent advances in artificial intelligence have driven increasing intelligent applications at the network edge, such as smart home, smart factory, and smart city. To deploy computationally intensive Deep Neural Networks (DNNs) on resource-constrained edge devices, traditional approaches have relied on either offloading workload to the remote cloud or optimizing computation at the end device locally. However, the cloud-assisted approaches suffer from the unreliable and delay-significant wide-area network, and the local computing approaches are limited by the constrained computing capability. Towards high-performance edge intelligence, the cooperative execution mechanism offers a new paradigm, which has attracted growing research interest recently. In this paper, we propose CoEdge, a distributed DNN computing system that orchestrates cooperative DNN inference over heterogeneous edge devices. CoEdge utilizes available computation and communication resources at the edge and dynamically partitions the DNN inference workload adaptive to devices' computing capabilities and network conditions. Experimental evaluations based on a realistic prototype show that CoEdge outperforms status-quo approaches in saving energy with close inference latency, achieving up to 25.5% ~ 66.9% energy reduction for four widely-adopted CNN models.

We develop an auction framework for privacy-preserving data aggregation in mobile crowdsensing, where the platform plays the role as an auctioneer to recruit workers for sensing tasks. The workers are allowed to report noisy versions of their data for privacy protection; and the platform selects workers by taking into account their sensing capabilities to ensure the accuracy level of the aggregated result. Observe that when moving the control of data privacy from the data aggregator to the workers, the data aggregator has limited market power in the sense that it can only partially control the noise by judiciously choosing a subset of workers based on workers' privacy preferences. This introduces externalities because the privacy of each worker depends on the total noise in the aggregated result that in turn relies on which workers are selected. Specifically, we first consider a privacy-passive scenario where workers participate if their privacy loss can be adequately compensated by the rewards. We explicitly characterize the externalities and the hidden monotonicity property of the problem, making it possible to design a truthful, individually rational and computationally efficient incentive mechanism. We then extend the results to a privacy-proactive scenario where workers have individual requirements for their perceivable data privacy levels. Our proposed mechanisms for both scenarios can select a subset of workers to (nearly) minimize the cost of purchasing their private sensing data subject to the accuracy requirement of the aggregated result. We validate the proposed scheme through theoretical analysis as well as extensive simulations.

Recent years have witnessed the proliferation of mobile computing and the Internet of Things (IoT), in which billions of mobile and IoT devices are connected to the Internet, generating zillions bytes of data at the network edge. However, it is challenging and infeasible to transfer and process zillions bytes of data using the current cloud-device architecture, due to bandwidth constraints of networks, potentially uncontrollable latency of cloud services, and privacy concerns while collecting data from IoT devices. To tackle these challenges, edge computing, an emerging computing paradigm, has received a tremendous amount of interest. By pushing data storage, computing, and controls closer to the network edge, edge computing has been widely recognized as a promising solution to meet the requirements of low latency, high scalability, and energy efficiency, as well as to mitigate the network traffic burdens. However, with the emergence of diverse IoT applications (e.g., smart city, industrial automation, and connected car), it becomes challenging for edge computing to deal with these heterogeneous IoT environments.

Machine learning (ML) has been embedded in many Internet of Things (IoT) applications (e.g., smart home and autonomous driving). Yet it is often infeasible to deploy ML models on IoT devices due to resource limitation. Thus, deploying trained ML models in the cloud and providing inference services to IoT devices becomes a plausible solution. To provide low-latency ML serving to massive IoT devices, a natural and promising approach is to use parallelism in computation. However, existing ML systems (e.g., Tensorflow) and cloud ML-serving platforms (e.g., SageMaker) are service-level-objective (SLO) agnostic and rely on users to manually configure the parallelism at both request and operation levels. To address this challenge, we propose a region-based reinforcement learning (RRL)-based scheduling framework for ML serving in IoT applications that can efficiently identify optimal configurations under dynamic workloads. A key observation is that the system performance under similar configurations in a region can be accurately estimated by using the system performance under one of these configurations due to their correlation. We theoretically show that the RRL approach can achieve fast convergence speed at the cost of performance loss. To improve the performance, we propose an adaptive RRL algorithm based on Bayesian optimization to balance the convergence speed and the optimality. The proposed framework is prototyped and evaluated on the Tensorflow Serving system. Extensive experimental results show that the proposed approach can outperform state-of-the-art approaches by finding near-optimal solutions over eight times faster while reducing inference latency up to 88.9% and reducing SLO violation up to 91.6%.

The success of machine learning has prospered Machine-Learning-as-a-Service (MLaaS) - deploying trained machine learning (ML) models in cloud to provide low latency inference services at scale. To meet latency Service-Level-Objective (SLO), judicious parallelization at both request and operation levels is utterly important. However, existing ML systems (e.g., Tensorflow) and cloud ML serving platforms (e.g., SageMaker) are SLO-agnostic and rely on users to manually configure the parallelism. To provide low latency ML serving, this paper proposes a swift machine learning serving scheduling framework with a novel Region-based Reinforcement Learning (RRL) approach. RRL can efficiently identify the optimal parallelism configuration under different workloads by estimating performance of similar configurations with that of the known ones. We both theoretically and experimentally show that the RRL approach can outperform state-of-the-art approaches by finding near optimal solutions over 8 times faster while reducing inference latency up to 79.0% and reducing SLO violation up to 49.9%.

Demand response (DR), as a vital component of smart grid, plays an important role in shaping the load profiles in order to improve system reliability and efficiency. Incentive-based DR has been used in many DR programs by incentivizing customers to adapt their loads to supply availability. Note that users' behavior patterns can be easily identified from fine-grained power consumption when interacting with the load serving entity (LSE), giving rise to serious privacy concerns. One common approach to address the privacy threats is to incorporate perturbations in users' load measurements. Although it can protect the users' privacy, yet the usage data modification would degrade the LSE's performance in achieving an optimal incentive strategy due to unknown characteristics of the augmented perturbations. In this paper, we cast the incentive-based DR problem as a stochastic Stackelberg game. To tackle the challenge induced by users' privacy protection behaviors, we propose a two-timescale reinforcement learning algorithm to learn the optimal incentive strategy under users' perturbed responses. The proposed algorithm computes the expected utility cost to mitigate the impacts of the random characteristics of the augmented perturbations and then updates the incentive strategy based on the perceived expected utility costs. We derive the conditions under which the proposed incentive scheme converges almost surely to an ε-optimal strategy. The efficacy of the proposed algorithm is demonstrated using extensive numerical simulation using real data.

Mobile fog computing (MFC) is an emerging paradigm that extends cloud computing (CC) by adding a new layer between the cloud and its end users.With the cloud-aware MFC, the cloud can pre-push certain important resources to the fog to reduce the networking latency and release the traffic burden over the links.e end user then is able to perform offline computing on the fog layer so that only the important results need to be delivered to and stored in the cloud.Moreover, the dense geographical deployment of fog servers enables the system to be aware of the end user's location.erefore, some location-sensitive applications could be well supported by the fog-aided cloud systems.Note that the cloud-aware MFC is different from the mobile edge computing (MEC), another promising technology for overcoming the shortcomings of CC, since MFC is able to jointly work with the cloud, but MEC is usually defined by the exclusion of CC.Specifically, in MEC, computing applications, data, and services are pushed away from the centralized nodes to the network edge, which enables network edge to run in an isolated environment from the rest of the network and provides access to local resources and data.In contrast, MFC provides not only a systemlevel horizontal architecture but also a new way to distribute, orchestrate, and manage secure resources across the network rather than just performing computing at the network edge.How to design efficient system architectures, transmission strategies, and protocols for MFC and how to efficiently analyze and evaluate the system performance are very important and essential.ese topics have carved out a new area rich in research and innovation potential.is special issue aims to address all these topics and invite contributions from worldwide leading researchers.

Wearable devices that capture users' rich information regarding their daily activities have unmet authentication needs. Today's solutions, which primarily rely on indirect authentication mechanisms via users' smartphones, thus cumbersome and susceptible to adversary intrusions. Even though there have been some efforts trying to fill this gap, they either rely on some superior sensors, such as cameras and electrocardiogram (ECG) pads, or are awkward to use, e.g., users are asked to perform some pre-defined movement/gesture for authentication. Therefore, an authentication mechanism for wearable devices that is accurate, robust, light-weight and convenient is in dire need.

We develop an auction framework for privacy-preserving data aggregation in mobile crowdsensing, where the platform plays the role as an auctioneer to recruit workers for a sensing task. In this framework, the workers are allowed to report privacy-preserving versions of their data to protect their data privacy; and the platform selects workers based on their sensing capabilities, which aims to address the drawbacks of game-theoretic models that cannot ensure the accuracy level of the aggregated result, due to the existence of multiple Nash Equilibria. Observe that in this auction based framework, there exists externalities among workers' data privacy, because the data privacy of each worker depends on both her injected noise and the total noise in the aggregated result that is intimately related to which workers are selected to fulfill the task. To achieve a desirable accuracy level of the data aggregation in a cost-effective manner, we explicitly characterize the externalities, i.e., the impact of the noise added by each worker on both the data privacy and the accuracy of the aggregated result. Further, we explore the problem structure, characterize the hidden monotonicity property of the problem, and determine the critical bid of workers, which makes it possible to design a truthful, individually rational and computationally efficient incentive mechanism. The proposed incentive mechanism can recruit a set of workers to approximately minimize the cost of purchasing private sensing data from workers subject to the accuracy requirement of the aggregated result. We validate the proposed scheme through theoretical analysis as well as extensive simulations.

Fog computing, as a new paradigm, has many characteristics that are different from cloud computing. Due to the resources being limited, fog nodes/MEC hosts are vulnerable to cyberattacks. Lightweight intrusion detection system (IDS) is a key technique to solve the problem. Because extreme learning machine (ELM) has the characteristics of fast training speed and good generalization ability, we present a new lightweight IDS called sample selected extreme learning machine (SS‐ELM). The reason why we propose “sample selected extreme learning machine” is that fog nodes/MEC hosts do not have the ability to store extremely large amounts of training data sets. Accordingly, they are stored, computed, and sampled by the cloud servers. Then, the selected sample is given to the fog nodes/MEC hosts for training. This design can bring down the training time and increase the detection accuracy. Experimental simulation verifies that SS‐ELM performs well in intrusion detection in terms of accuracy, training time, and the receiver operating characteristic (ROC) value.

Distributed spectrum access (DSA) is challenging, since an individual secondary user often has limited sensing capabilities only. One key insight is that channel recommendation among secondary users can help to take advantage of the inherent correlation structure of spectrum availability in both time and space, and enable users to obtain more informed spectrum opportunities. With this insight, we advocate to leverage the wisdom of crowds, and devise social recommendation aided DSA mechanisms to orient secondary users to make more intelligent spectrum access decisions, for both strong and weak network information cases. We start with the strong network information case where secondary users have the statistical information. To mitigate the difficulty due to the curse of dimensionality in the stochastic game approach, we take the one-step Nash approach and cast the social recommendation aided DSA decision making problem at each time slot as a strategic game. We show that it is a potential game, and then devise an algorithm to achieve the Nash equilibrium by exploiting its finite improvement property. For the weak information case where secondary users do not have the statistical information, we develop a distributed reinforcement learning mechanism for social recommendation aided DSA based on the local observations of secondary users only. Appealing to the maximum-norm contraction mapping, we also derive the conditions under which the distributed mechanism converges and characterize the equilibrium therein. Numerical results reveal that the proposed social recommendation aided DSA mechanisms can achieve a superior performance using real social data traces and its performance loss in the weak network information case is insignificant, compared with the strong network information case.

In spite of the pronounced benefit brought by crowdsensing, a user would not participate in sensing without adequate incentive, indicating that effective incentive design plays a critical role in making crowdsensing a reality. In this work, we examine the impact of two conflicting factors on incentives for users' participation: 1) the concern about privacy leakage and 2) the (positive) network effect from many sensing participants. The former factor hinders privacy- aware users from participating, whereas the latter encourages users' participation. Taking into consideration both factors, we devise a privacy-preserving crowdsensing scheme, in which a reverse `privacy' auction is first run by the crowdsensing platform to select users based on their privacy valuations and the network effect. Then the trusted platform carries out differentially private data aggregation over the collected data such that the released sensing result remains useful for the task agent, while all participants' data privacy is guaranteed. A natural objective here is then to maximize the profit of the task agent, i.e., the difference between its utility and the total reward to the participants. To this end, the platform utilizes a random-sampling based mechanism for the 'privacy' auction, followed by a Laplace mechanism for data aggregation. We show that this auction mechanism design is 4-competitive, and further it exhibits desirable properties, including individual rationality, truthfulness, computational efficiency. Simulation results corroborate the theoretical properties of the proposed privacy-preserving crowdsensing scheme.

We develop a social group utility maximization (SGUM) framework for cooperative wireless networking that takes into account both social relationships and physical coupling among users. Specifically, instead of maximizing its individual utility or the overall network utility, each user aims to maximize its social group utility that hinges heavily on its social tie structure with other users. We show that this framework provides rich modeling flexibility and spans the continuum between non-cooperative game and network utility maximization (NUM)-two traditionally disjoint paradigms for network optimization. Based on this framework, we study three important applications of SGUM, in database assisted spectrum access, power control, and random access control, respectively. For the case of database assisted spectrum access, we show that the SGUM game is a potential game and always admits a socially-aware Nash equilibrium (SNE). We also develop a distributed spectrum access algorithm that can converge to the SNE and also quantify the trade-off between the performance and convergence time of the algorithm. For the cases of power control and random access control, we show that there exists a unique SNE and the network performance improves as the strength of social ties increase. Numerical results corroborate that the SGUM solutions can achieve superior performance using real social data trace. Furthermore, we show that the SGUM framework can be generalized to take into account both positive and negative social ties among users, which can be a useful tool for studying network security problems.

In this paper, we study a privacy-preserving spectrum sharing system to protect secondary users' location privacy while enhancing spectrum access. The location privacy of secondary users can be compromised by an external adversary via the received signal strength (RSS)-based localization technique. To mitigate such privacy threat, we employ a random power perturbation approach that allows each secondary user to judiciously obfuscate the RSS captured by the adversary. While it can protect users' location privacy, the power perturbation approach would inevitably degrade the system performance and bring challenges to the design of the spectrum allocation algorithm. In this work, we adopt a socially-aware database assisted spectrum access system and cast the spectrum allocation under users' power perturbation as a stochastic channel selection game played among the users. To tackle the challenge brought by the privacy protection, we develop a two time-scale distributed learning algorithm, which is shown to converge almost surely to a socially-aware ε-Nash equilibrium. The numerical results show that the higher the privacy protection level is, the more significant the degradation of the network throughput would be.

We study deadline-aware scheduling with adaptive network coding (NC) for real-time traffic over a single-hop wireless network. To meet hard deadlines of real-time traffic, the block size for NC is adapted based on the remaining time to the deadline so as to strike a balance between maximizing the throughput and minimizing the risk that the entire block of coded packets may not be decodable by the deadline. This sequential block size adaptation problem is then cast as a finite-horizon Markov decision process. One interesting finding is that the optimal block size and its corresponding action space monotonically decrease as the deadline approaches, and that the optimal block size is bounded by the “greedy” block size. These unique structures make it possible to significantly narrow down the search space of dynamic programming, building on which we develop a monotonicity-based backward induction algorithm (MBIA) that can find the optimal block size in polynomial time. Furthermore, a joint real-time scheduling and channel learning scheme with adaptive NC is developed to adapt to channel dynamics in a mobile network environment. Then, we generalize the analysis to multiple flows with hard deadlines and long-term delivery ratio constraints. We devise a low-complexity online scheduling algorithm integrated with the MBIA, and then establish its asymptotical utility optimality. The analysis and simulation results are corroborated by high-fidelity wireless emulation tests, where actual radio transmissions over emulated channels are performed to demonstrate the feasibility of the MBIA in finding the optimal block size in real time.

In this paper, we develop a social group utility maximization (SGUM) framework for cooperative networking that takes into account both social relationships and physical coupling among users. Specifically, instead of maximizing its individual utility or the overall network utility, each user aims to maximize its social group utility that hinges heavily on its social ties with other users. We show that this framework provides rich modeling flexibility and spans the continuum space between non-cooperative game and network utility maximization (NUM) - two traditionally disjoint paradigms for network optimization. Based on this framework, we study an important application in database assisted spectrum access. We formulate the distributed spectrum access problem among white-space users with social ties as a SGUM game. We show that the game is a potential game and always admits a social-aware Nash equilibrium. We also design a distributed spectrum access algorithm that can achieve the social-aware Nash equilibrium of the game and quantify its performance gap. We evaluate the performance of the SGUM solution using real social data traces. Numerical results demonstrate that the performance gap between the SGUM solution and the NUM (social welfare optimal) solution is at most 15%.

Smart meters, designed for information collection and system monitoring in smart grid, report fine-grained power consumption to utility providers. With these highly accurate profiles of energy usage, however, it is possible to identify consumers' specific activity or behavior patterns, thereby giving rise to serious privacy concerns. In this paper, this concern is addressed by using battery energy storage. Beyond privacy protection, batteries can also be used to cut down the electricity bill. From a holistic perspective, a dynamic optimization framework is designed for consumers to strike a tradeoff between the smart meter data privacy and the electricity bill. In general, a major challenge in solving dynamic optimization problems lies in the need of the knowledge of the future electricity consumption events. By exploring the underlying structure of the original problem, an equivalent problem is derived, which can be solved by using only the current observations. An online control algorithm is then developed to solve the equivalent problem based on the Lyapunov optimization technique. To overcome the difficulty of solving a mixed-integer nonlinear program involved in the online control algorithm, the problem is further decomposed into multiple cases and the closed-form solution to each case is derived accordingly. It is shown that the proposed online control algorithm can optimally control the battery operations to protect the smart meter data privacy and cut down the electricity bill, without the knowledge of the statistics of the time-varying load requirement and the electricity price processes. The efficacy of the proposed algorithm is demonstrated through extensive numerical evaluations using real data.

This paper investigates pricing-based spectrum access control in cognitive radio networks, where primary users (PUs) sell the temporarily unused spectrum and secondary users (SUs) compete via random access for such spectrum opportunities. Compared to existing market-based approaches with centralized scheduling, pricing-based spectrum management with random access provides a platform for SUs contending for spectrum access and is amenable to decentralized implementation due to its low complexity. We focus on two market models, one with a monopoly PU market and the other with a multiple-PU market. For the monopoly PU market model, we devise decentralized pricing-based spectrum access mechanisms that enable SUs to contend for channel usage. Specifically, we first consider SUs contending via slotted Aloha. Since the revenue maximization problem therein is nonconvex, we characterize the corresponding Pareto-optimal region and obtain a Pareto-optimal solution that maximizes the SUs' throughput subject to their budget constraints. To mitigate the spectrum underutilization due to the “price of contention,” we revisit the problem where SUs contend via CSMA, which results in more efficient spectrum utilization and higher revenue. We then study the tradeoff between the PU's utility and its revenue when the PU's salable spectrum is controllable. Next, for the multiple-PU market model, we cast the competition among PUs as a three-stage Stackelberg game, where each SU selects a PU's channel to maximize its throughput. We explore the existence and the uniqueness of Nash equilibrium, in terms of access prices and the spectrum offered to SUs, and develop an iterative algorithm for strategy adaptation to achieve the Nash equilibrium. Our findings reveal that there exists a unique Nash equilibrium when the number of PUs is less than a threshold determined by the budgets and elasticity of SUs.

We study the diffusion behavior of real-time information. Typically, real-time information is valuable only for a limited time duration, and hence needs to be delivered before its “deadline.” Therefore, real-time information is much easier to spread among a group of people with frequent interactions than between isolated individuals. With this insight, we consider a social network which consists of many cliques and information can spread quickly within a clique. Furthermore, information can also be shared through online social networks, such as Facebook, twitter, Youtube, etc. We characterize the diffusion of real-time information by studying the phase transition behaviors. Capitalizing on the theory of inhomogeneous random networks, we show that the social network has a critical threshold above which information epidemics are very likely to happen. We also theoretically quantify the fractional size of individuals that finally receive the message. The numerical results indicate that real-time information could be much easier to propagate in a social network when large size cliques exist.

This paper studies risk-aware day-ahead scheduling and real-time dispatch for plug-in electric vehicles (EVs), aiming to jointly optimize the EV charging cost and the risk of the load mismatch between the forecasted and the actual EV loads, due to the random driving activities of EVs. It turns out that the inclusion of the load mismatch risk in the objective function complicates the risk-aware day-ahead scheduling and indeed the optimization problem is nonconvex. A key step is to utilize the hidden convexity structure to recast it as a two-stage stochastic linear program, which can be solved by using the L-shaped method. Further, we develop a distributed risk-aware real-time dispatch algorithm, where the aggregator only needs to compute the shadow prices for each EV to optimize its own charging strategy in a distributed manner. We show, based on real data, that the proposed risk-aware day-ahead scheduling algorithm can reduce not only the overall charging cost, but also the peak demand of EV charging.

We study adaptive network coding (NC) for scheduling real-time traffic over a single-hop wireless network. To meet the hard deadlines of real-time traffic, it is critical to strike a balance between maximizing the throughput and minimizing the risk that the entire block of coded packets may not be decodable by the deadline. Thus motivated, we explore adaptive NC, where the block size is adapted based on the remaining time to the deadline, by casting this sequential block size adaptation problem as a finite-horizon Markov decision process. One interesting finding is that the optimal block size and its corresponding action space monotonically decrease as the deadline approaches, and the optimal block size is bounded by the "greedy" block size. These unique structures make it possible to narrow down the search space of dynamic programming, building on which we develop a monotonicity-based backward induction algorithm (MBIA) that can solve for the optimal block size in polynomial time. Since channel erasure probabilities would be time-varying in a mobile network, we further develop a joint real-time scheduling and channel learning scheme with adaptive NC that can adapt to channel dynamics. We also generalize the analysis to multiple flows with hard deadlines and long-term delivery ratio constraints, devise a low-complexity online scheduling algorithm integrated with the MBIA, and then establish its asymptotic throughput-optimality. In addition to analysis and simulation results, we perform high fidelity wireless emulation tests with real radio transmissions to demonstrate the feasibility of the MBIA in finding the optimal block size in real time.

In this study, downlink scheduling of multiuser traffic with hard deadlines and packet-level priorities is cast as a partially observable Markov decision process. User channels are modeled as Markovian and the base station can learn only the channel condition of the currently scheduled user. The optimization of joint channel learning and scheduling presents the combined challenges incurred by the strict deadline constraint of real-time traffic and the partial observability of multiuser channels. In particular, we show that idling adds a new dimension to the action space; and that, through a case study of heterogeneous multiuser networks, idling is indeed the optimal action under certain system states. This somewhat surprising result reveals the existence of the fundamental tradeoffs between exploitation and exploration/idling, going beyond the classic `exploitation vs exploration'. We find that, due to hard deadlines and packet priorities, idling is intimately related to the tradeoff between the successful transmission of backlogged packets and that of future arrivals. In contrast, for the special case with a symmetric two-user system, we show that the scheduling problem exhibits unique structures, rendering a non-idling greedy policy optimal.

It is known that distributed power control in wireless ad-hoc networks is challenging, due to the inherent global coupling between concurrent transmissions interfering with each other. Observing that the globally optimal point lies on the boundary of the feasible region, we transform the utility maximization problem into a more structured problem in the form of maximizing the minimum weighted utility. Then, we develop a centralized algorithm for the minimum weighted utility maximization problem as a benchmark. Next, by using extended duality theory, we introduce penalty multipliers and decompose the minimum weighted utility maximization problem into subproblems for individual users. Appealing to the simulated annealing method, we propose a distributed stochastic power control algorithm, where each user stochastically adjusts its target utility to improve the overall system utility. Although the underlying optimization problem is nonconvex, our algorithm can guarantee global optimality although the convergence rate may be slow due to the usage of simulated annealing. We improve the convergence rate further by devising an enhanced algorithm based on the geometric cooling schedule.

Market-based mechanisms offer promising approaches for spectrum access in cognitive radio networks. In this paper, we focus on two market models, one with a monopoly primary user (PU) market and the other with a multiple PU market, where each PU sells its temporarily unused spectrum to secondary users (SUs). We propose a pricing-based spectrum trading mechanism that enables SUs to contend for channel usage by random access, in a distributed manner, which naturally mitigates the complexity and time overhead associated with centralized scheduling. For the monopoly PU market model, we first consider SUs contending via slotted Aloha. The revenue maximization problems here are nonconvex. We first characterize the Pareto optimal region, and then obtain a Pareto optimal solution that maximizes the SUs' throughput subject to the SUs' budget constraints. To mitigate the spectrum underutilization due to the “price of contention,” we revisit the problem where SUs contend via CSMA, and show that spectrum utilization is enhanced, resulting in higher revenue. When the PU's unused spectrum is a control parameter, we study further the tradeoff between the PU's utility and its revenue. For the multiple PU market model, we cast the competition among PUs as a three-stage Stackelberg game, where each SU selects a PU's channel to maximize its throughput. We characterize the Nash equilibria, in terms of access prices and the spectrum offered to SUs. Our findings reveal that the number of equilibria exhibits a phase transition phenomenon, in the sense that when the number of PUs is greater than a threshold, there exist infinitely many equilibria; otherwise, there exists a unique Nash equilibrium, where the access prices and spectrum opportunities are determined by the budgets/elasticity of SUs and the utility level of PUs.

This paper makes use of the direct heuristic dynamic programming design in a nonlinear tracking control setting with filtered tracking error. A Lyapunov stability approach is used for the stability analysis of the tracking system. It is shown that the closed-loop tracking error and the approximating neural network weight estimates retain the property of uniformly ultimate boundedness under the presence of neural network approximation error and bounded unknown disturbances under certain conditions.

Approximate dynamic programming (ADP) has been widely studied from several important perspectives: algorithm development, learning efficiency measured by success or failure statistics, convergence rate, and learning error bounds. Given that many learning benchmarks used in ADP or reinforcement learning studies are control problems, it is important and necessary to examine the learning controllers from a control-theoretic perspective. This paper makes use of direct heuristic dynamic programming (direct HDP) and several benchmark examples to introduce a unique analytical framework that can be extended to other learning control paradigms and other complex control problems. The sensitivity analysis and the linear quadratic regulator (LQR) design are used in the paper for two purposes: to gauge direct HDP performance characteristics and to provide guidance toward designing better learning controllers. This gauge however does not limit the direct HDP to be effective only as a linear controller. Toward this end, applications of the direct HDP for nonlinear control problems beyond sensitivity analysis and the confines of LQR have been developed and compared with LQR design for command following and internal system parameter changes.

Recently, a system theoretic framework for learning and optimization has been developed that shows how many approximate dynamic programming paradigms such as perturbation analysis, Markov decision processes, and reinforcement learning are very closely related. Using this system theoretic framework, a new optimization technique called gradient-based policy iteration (GBPI) has been developed. In this paper, we show how GBPI iteration can be extended to partially observable Markov decision processes (POMDPs). We also develop the value iteration analogue of GBPI and show that this new version of value iteration, extended to POMDPs, not only theoretically acts like value iteration but also does so numerically.