Mobile edge computing (MEC) is an emerging technology to support resource-intensive yet delay-sensitive applications using small cloud-computing platforms deployed at the mobile network edges. However, the existing MEC techniques are not applicable to the situation where the number of mobile users increases explosively or the network facilities are sparely distributed. In view of this insufficiency, unmanned aerial vehicles (UAVs) have been employed to improve the connectivity of ground Internet-of-Things (IoT) devices due to their high altitude. This paper proposes an innovative UAV-enabled MEC system involving the interactions among IoT devices, UAV and edge clouds (ECs). The system deploys and operates a UAV properly to facilitate the MEC service provisioning to a set of IoT devices in regions where existing ECs cannot be accessible to IoT devices due to terrestrial signal blockage or shadowing. The UAV and ECs in the system collaboratively provide MEC services to the IoT devices. For optimal service provisioning in this system, we formulate an optimization problem aiming at minimizing the weighted sum of the service delay of all IoT devices and UAV energy consumption by jointly optimizing UAV position, communication and computing resource allocation, and task splitting decisions. However, the resulting optimization problem is highly non-convex and thus difficult to solve optimally. To tackle this problem, we develop an efficient algorithm based on successive convex approximation to obtain sub-optimal solutions. Numerical experiments demonstrate that our proposed collaborative UAV-EC offloading scheme largely outperforms baseline schemes that solely rely on UAV or edge clouds for MEC in IoT.

To provide intelligent and personalized services on smart devices, machine learning techniques have been widely used to learn from data, identify patterns, and make automated decisions. Machine learning processes typically require a large amount of representative data that are often collected through crowdsourcing from end users. However, user data could be sensitive in nature, and training machine learning models on these data may expose sensitive information of users, violating their privacy. Moreover, to meet the increasing demand of personalized services, these learned models should capture their individual characteristics. This paper proposes a privacy-preserving approach for learning effective personalized models on distributed user data while guaranteeing the differential privacy of user data. Practical issues in a distributed learning system such as user heterogeneity are considered in the proposed approach. In addition, the convergence property and privacy guarantee of the proposed approach are rigorously analyzed. Experimental results on realistic mobile sensing data demonstrate that the proposed approach is robust to user heterogeneity and offers a good trade-off between accuracy and privacy.

Alternating direction method of multipliers (ADMM) is a widely used tool for machine learning in distributed settings where a machine learning model is trained over distributed data sources through an interactive process of local computation and message passing. Such an iterative process could cause privacy concerns of data owners. The goal of this paper is to provide differential privacy for ADMM-based distributed machine learning. Prior approaches on differentially private ADMM exhibit low utility under high privacy guarantee and assume the objective functions of the learning problems to be smooth and strongly convex. To address these concerns, we propose a novel differentially private ADMM-based distributed learning algorithm called DP-ADMM, which combines an approximate augmented Lagrangian function with time-varying Gaussian noise addition in the iterative process to achieve higher utility for general objective functions under the same differential privacy guarantee. We also apply the moments accountant method to analyze the end-to-end privacy loss. The theoretical analysis shows that the DP-ADMM can be applied to a wider class of distributed learning problems, is provably convergent, and offers an explicit utility-privacy tradeoff. To our knowledge, this is the first paper to provide explicit convergence and utility properties for differentially private ADMM-based distributed learning algorithms. The evaluation results demonstrate that our approach can achieve good convergence and model accuracy under high end-to-end differential privacy guarantee.

Recently intensive efforts have been made on the transformation of the world's largest physical system, the power grid, into a “smart grid” by incorporating extensive information and communication infrastructures. Key features in such a “smart grid” include high penetration of renewable and distributed energy sources, large-scale energy storage, market-based online electricity pricing, and widespread demand response programs. From the perspective of residential customers, we can investigate how to minimize the expected electricity cost with real-time electricity pricing, which is the focus of this paper. By jointly considering energy storage, local distributed generation such as photovoltaic (PV) modules or small wind turbines, and inelastic or elastic energy demands, we mathematically formulate this problem as a stochastic optimization problem and approximately solve it by using the Lyapunov optimization approach. From the theoretical analysis, we have also found a good tradeoff between cost saving and storage capacity. A salient feature of our proposed approach is that it can operate without any future knowledge on the related stochastic models (e.g., the distribution) and is easy to implement in real time. We have also evaluated our proposed solution with practical data sets and validated its effectiveness.

The advanced metering infrastructure (AMI) in the smart grid provides real-time information to both grid operators and customers, exploiting the full potential of demand response (DR). However, it introduces new privacy threats to customers. Prior works have proposed privacy-preserving methods in the AMI, such as temporal or spatial aggregation. A main assumption in these works is that fine-grained data do not need to be attributable to individuals. However, this assumption does not hold in incentive-based demand response (IDR) programs where fine-grained metering data are required to analyze individual demand curtailments, and hence, need to be attributable. In this paper, we propose a privacy-preserving scheme for IDR programs in the smart grid, which enables the DR provider to compute individual demand curtailments and DR rewards while preserving customer privacy. Moreover, a customer can reveal his/her identity and prove ownership of his/her power usage profile in certain situations, such as legal disputes. We achieve both privacy and efficiency in our scheme through a combination of several cryptographic primitives, such as identity-committable signatures and partially blind signatures. As far as we know, we are the first to identify and address privacy issues for IDR programs in the smart grid.

In this paper, we investigate the minimization of the total energy cost of multiple residential households in a smart grid neighborhood sharing a load serving entity. Specifically, each household may have renewable generation, energy storage as well as inelastic and elastic energy loads, and the load serving entity attempts to coordinate the energy consumption of these households in order to minimize the total energy cost within this neighborhood. The renewable generation, the energy demand arrival, and the energy cost function are all stochastic processes and evolve according to some, possibly unknown, probabilistic laws. We develop an online control algorithm, called Lyapunov-based cost minimization algorithm (LCMA), which jointly considers the energy management and demand management decisions. LCMA only needs to keep track of the current values of the underlying stochastic processes without requiring any knowledge of their statistics. Moreover, a decentralized algorithm to implement LCMA is also developed, which can preserve the privacy of individual household owners. Numerical results based on real-world trace data show that our control algorithm can effectively reduce the total energy cost in the neighborhood.