The swift expansion of integrated renewable energy within the electrical grid underscores the essential role of distributed energy resources (DERs) in shaping the future of power transmission and distribution networks. This increasing incorporation of DERs has prompted power utility providers to enhance their systematic awareness and deploy advanced load control techniques. Nonetheless, a persistent challenge revolves around optimizing energy dispatching amidst substantial fluctuations originating from DERs. These fluctuations primarily stem from long-term data stochasticity and the inherent unpredictability of integrated DERs. Existing methods fall short in addressing the intricate nature of energy distribution from utilities to households, especially when dealing with exogenous loads beyond smart meters. To address this issue, this paper proposes RoGANDER, a Robust Generative Adversarial Network (GAN) for DERs. It overcomes challenges associated with high spatial and temporal data granularity, along with significant time consumption. The RoGANDER approach leverages generative sequential data to mitigate data unpredictability and stochasticity. Our evaluation results confirm a notable reduction in fluctuations and an enhanced ability to anticipate households’ energy needs.

Federated learning (FL) allows the collaborative training of machine learning (ML) models between an aggregation server and different clients without sharing their private data. However, the FL archetype is mostly vulnerable to malicious model updates from various clients because of the privacy feature that makes the server see clients as a black box. When selecting clients, the server has no control on their contributions during training. This opacity of the server toward clients’ data associated with the huge amount of heterogeneous data brings a security risk and poses a deterioration of the model performance in FL. The impact of client selection and data heterogeneity on FL robustness has been overlooked. In this article, we develop an incentive design for heterogeneous client selection (IHCS) to improve the performance while reducing the security risks in FL. The IHCS approach applies a smarter client selection method using cooperative game theory and dynamic clustering of clients based on their heterogeneity level to overcome the challenges of lacking access to clients’ data, unbalanced data, and the lack of applicable data contribution from clients in FL. To do so, we attribute a recognition value to each client using the Shapley value. This recognition index is then used to aggregate the probability of participation level. We also implement, within the IHCS, a heterogeneity-based clustering (HIC) method that helps inhibit the negative influence of data heterogeneity and increase client contributions. Through extensive experiments with empirical results, the proposed approach outperforms the representative works on robustness of FL.

With the proliferation of computationally powerful edge devices, edge computing has been widely adopted for wide-ranging computational tasks. Among these, edge artificial intelligence (AI) has become a new trend, allowing local devices to work cooperatively and build deep learning models. Federated learning is one of the representative frameworks in distributed machine learning paradigms. However, there are several major concerns with existing federated learning paradigms. Existing distributed frameworks rely on a central server to coordinate the computing process, where such a central node may raise security concerns. Federated learning also relies on several assumptions/requirements, e.g., the independent and identically distributed (i.i.d.) data and model homogeneity. Since more and more edge devices are able to train lightweight models with local data, such models are normally heterogeneous. To tackle these challenges, in this article, we develop a blockchain-empowered federated learning framework that enables learning in a fully decentralized manner while taking the model heterogeneity and data heterogeneity into account. In particular, a federated learning framework with a heterogeneous calibration process, i.e., Model and Feature Calibration (FL-MFC), is developed to enable collaboration among heterogeneous models. We further design a two-level mining process using blockchain to enable the secure decentralized learning process. Experimental results show that our proposed system achieves effective learning performance under a fully heterogeneous environment.

Facial recognition (FR) systems are employed to identify and authenticate individuals. There has been a rise in privacy concerns regarding mass surveillance and unauthorized usages. As a result, one viable approach is adding adversarial noise to distort user profile images so that FR technology can be bypassed. Nonetheless, such approaches could be used by adversaries to avoid detection in surveillance footage and therefore evade identification. To combat this threat, a line of research efforts focuses on generative adversarial network (GAN)-based Denoising and Restoration to remove adversarial noise. In this paper, GAN-based methods are investigated experimentally for assessing their effectiveness. Particularly, three GAN-based approaches, i.e., Blind Face Restoration, Blur and Restore, and Image-to-image Translation, are extensively examined with several representative classification approaches. Our evaluation results show that GAN denoising schemes could improve image visual quality, but are ineffective to remove perturbations for privacy protection attached by Fawkes or Lowkey. We further discuss some future research directions on image transformation-based approaches, which can potentially improve the effectiveness.

While facial recognition (FR) has been widely used by businesses and governments for various purposes, it gives rise to privacy concerns once the consent of users is not handled properly. Hence, researchers have proposed methods to evade FR technology by attaching adversarial perturbations to user profile images. Nonetheless, image denoising-based methods have been proposed to increase the model robustness over adversarial examples. This paper investigates the impact of transformations on adversarial facial images. In particular, a simple but effective framework, TaD (Transformations as Denoising), is proposed to remove possible adversarial perturbations from user images generated by popular FR privacy protection frameworks. Extensive evaluations show the reliability of Fawkes and LowKey with various simple transformations. Experimental results indicate that simple transformations can impact the protection performance, and the choice of DNN-based facial feature extractors can enhance the robustness of facial images with adversarial perturbations. The experimental results also demonstrate strengths and weaknesses of FR methods and give suggestions for further improvements of privacy safeguard tools.

The Internet of Things (IoT) encompasses a near-incalculable collection of dispersed and embedded computing devices acting as sensors and actuators, generating data at an incredible scale. However, a lack of coherency and cross-compatibility in IoT deployments has lead to increasing redundancy and waste of resources. To combat this, various concepts have been proposed for an open IoT search engine (IoT-SE) that serves human and machine users. Invariably, the IoT-SE envisions distributed query retrieval to handle massive volumes of devices and data. Incorporating the massively heterogeneous protocols and properties of devices deployed, the search of such a system for timely and pertinent data is massively challenging, to provide useful knowledge and service for IoT systems. Moreover, enabling and maintaining security and privacy in an IoT-SE is likewise a prodigious task, as end users, IoT devices, and the search system itself, have different protocols and requirements. To this end, a study of security issues in IoT search is conducted to outline the challenges ahead, and a case study to resolve practical security vulnerabilities in an IoT-SE system is carried out. The pertinent issues of security in an IoT-SE system are reviewed. Particularly: 1) a taxonomy is detailed for IoT-SE security issues; 2) the vulnerabilities of machine learning (ML) models in the IoT-SE are considered; and 3) defensive mechanisms are presented for securing IoT Search. A case study is carried out to implement basic security features in the IoT search, addressing the risk of false queries through the design of ML-based solutions. Finally, a roadmap for future research is provided, including the security and privacy for IoT systems connected to the IoT-SE, distributed edge computing in IoT-SE, privacy-preserving data markets in IoT-SE, and distributed ML in IoT-SE.

The fast development of the Internet of Things (IoT) and deep learning enables learning useful patterns from the massive amount of collected data with sporadic nodes in IoT systems. Federated learning has received increasing attention in distributed machine learning where only intermediate parameters are exchanged with training samples that resided at local nodes. Nonetheless, most of the existing federated learning schemes assume a homogeneous distribution of data. The assumption, however, does not apply to IoT systems because of the heterogeneity of the IoT architecture. The nonindependent and identical distribution (non-IID) property in data volume and statistical distribution of IoT nodes can impact the performance of an aggregated global model that fits all nodes. Existing federated learning solutions for non-IID data sets either have to train additional models or require extra data exchange to check the node distribution. However, due to resource constraints in IoT systems, these approaches will increase the burden on limited computation capacity and cause network overhead. To address the issue, in this article, a novel weight-similarity-based client clustering (WSCC) approach is proposed, in which clients are split into different groups based on their data set distributions. An affinity-propagation-based method with the cosine distance of the client’s weight parameters is designed to iteratively and automatically determine dynamic clusters. The proposed approach is ideal for IoT systems since there are no auxiliary models and extra data transmissions are needed. Through the theoretical convergence analysis and empirical results, we show that our proposed WSCC scheme outperforms the representative federated learning schemes under different non-IID settings, achieving up to 20% improvements in accuracy.

The federated learning framework builds a deep learning model collaboratively by a group of connected devices via only sharing local parameter updates to the central parameter server. Nonetheless, the lack of transparency in the local data resource makes it prone to adversarial federated attacks, which have shown increasing ability to reduce learning performance. Existing research efforts either focus on the single-party attack with impractical perfect knowledge setting and limited stealthy ability or the random attack that has no control on attack effects. In this paper, we investigate a new multi-party adversarial attack with the imperfect knowledge of the target system. Controlled by an adversary, a number of compromised devices collaboratively launch targeted model poisoning attacks, intending to misclassify the targeted samples while maintaining stealthy under different detection strategies. Specifically, the compromised devices jointly minimize the loss function of model training in different scenarios. To overcome the update scaling problem, we develop a new boosting strategy by introducing two stealthy metrics. Via experimental results, we show that under both perfect knowledge and limited knowledge settings, the multi-party attack is capable of successfully evading detection strategies while guaranteeing the convergence. We also demonstrate that the learned model achieves the high accuracy on the targeted samples, which confirms the significant impact of the multi-party attack on federated learning systems.

Industrial Internet of Things (IIoT) envisions the tight coupling of numerous critical industrial manufacturing subsystems, such as control, networking, and computing through the ubiquitous Internet of Things technologies. Nonetheless, such interconnectivity poses significant challenges to the successful management and operation of massively distributed industrial manufacturing systems. Without carefully integrated system design, the nonoptimal management and operation of highly intertwined subsystems can lead to the loss of productivity and ultimately the value of factories and plants. To address this issue, in this article, we conduct the integrated design that is capable of simultaneously configuring both control and networking subsystems in IIoT with consideration for their inherent interdependencies. We first analyze the performance of the dynamic backoff exponential (BE) in IEEE 802.15.4 carrier-sense multiple access (CSMA) and show the performance improvement of dynamic BE. We then design a model-free reinforcement learning algorithm to configure the control and networking subsystems automatically via systematic trial and error, as it is impractical to build a model for a highly intertwined complex IIoT system. Considering the time-sensitive characteristics of IIoT systems, we design priority-aware policies based on importance among networking traffic (i.e., sensing traffic and actuation traffic) to improve the convergence speed. The experimental results demonstrate that our priority-aware reinforcement-learning-based integrated design can successfully reconfigure the complex and highly intertwined IIoT system at runtime with a minimal convergence time. Besides, our approach reduces convergence time by 37.5%, and energy consumption by 9.2%, compared to the standard reinforcement learning approach.

Community detection is an effective approach to unveil social dynamics among individuals in social networks. In the literature, quite a few algorithms have been proposed to conduct community detection by exploiting the topology of social networks and the attributes of social actors. In practice, community detection is usually conducted by third parties like advertisement companies, hospitals, with access to social networks for different purposes, which can easily lead to privacy breaches. In this paper, we investigate community detection in social networks aiming to protect the privacy of both the network topologies and the users’ attributes. In particular, we propose a new scheme called differentially private community detection (DPCD). DPCD detects communities in social networks via a probabilistic generative model, which can be decomposed into subproblems solved by individual users. The private social relationships and attributes of each user are protected by objective perturbation with differential privacy guarantees. Through both theoretical analysis and experimental validation using synthetic and real world social networks, we demonstrate that the proposed DPCD scheme detects social communities under modest privacy budget.

The ever-growing proliferation of mobile devices equipped with accelerometers has provided new opportunities to capture the semantic meanings of human activities and improve user experience with behavior-based recommendations, which heavily rely on the accuracy of the recognition of daily human activities. Acceleration-based human activity recognition (HAR) is a challenging problem because each accelerometer records multi-dimensional signals in both spatial and temporal domains that have different attributes for representing different activities or even the same activity. Thus we cannot directly compare these signals with each other, because they are embedded in a non-metric space. In this paper, we present a Personalized Recurrent Neural Network (PerRNN) to dynamically segment and recognize the human activities based on accelerometer data. Enlightened by the idea of spatiotemporal predictive learning, the proposed architecture is capable of memorizing different acceleration signals' appearances and temporal variations in a unified memory pool. We evaluate the performance of the proposed framework on a commonly used dataset, WISDM. Experiment results show that compared with state-of-the-art schemes, our proposed PerRNN system recognizes 6 different human activities with the highest overall accuracy of 96.44%.

Deep learning, as an increasingly powerful and popular data analysis tool, has the potential to improve smart grid operation. One critical issue is that the accuracy of deep learning relies heavily on the integrity of the training dataset, and the data collection process is time-consuming and complex, resulting in that the applying deep learning may not satisfy the needs of time-sensitive applications. Moreover, in the smart grid, predictions must be timely, and cannot wait for the initial dataset to be completely collected by the sensors. Also, the traditional centralized data analytics structure requires the entire dataset to be uploaded to the cloud datacenter for analysis, which incurs significant network resource and increases network congestion. To address these problems, in this paper we consider the allocation of deep learning at the network edge and directly in the Internet of Things (IoT) devices and design an online learning approach to enable small data subset training and continuous model updating to ensure accuracy requirements in time-sensitive environments. In our online learning approach, we implement the Just Another Network model, an optimized Long-Short Term Memory neural network model, to reduce the computation overhead for the deep learning training process. We evaluate our approach using real-world smart grid dataset. Our experimental results show that our online learning approach significantly reduces the training time while satisfying the accuracy requirements.

Big data has become a key basis of innovation and intelligence, potentially making our lives more convenient and bringing new opportunities to the modern society. Towards this goal, a critical underlying task is to solve a series of large-scale fundamental problems. Conducting such large-scale data analytics in a timely manner requires a large amount of computing resources, which may not be available for individuals and small companies in practice. By outsourcing their computations to the cloud, clients can solve such problems in a cost-effective way. However, confidential data stored at the cloud is vulnerable to cyber attacks, and thus needs to be protected. Previous works employ cryptographic techniques like homomorphic encryption, which significantly increase the computational complexity of solving a large-scale problem at the cloud and is impractical for big data applications. For the first time in the literature, we present an efficient secure outsourcing scheme for convex separable programming problems (CSPs). In particular, we first develop efficient matrix and vector transformation schemes only based on arithmetic operations that are computationally indistinguishable both in value and in structure under a chosen-plaintext attack (CPA). Then, we design a secure outsourcing scheme in which the client and the cloud collaboratively solve the transformed problems. The client can efficiently verify the correctness of returned results to prevent any malicious behavior of the cloud. Theoretical correctness and privacy analysis together show that the proposed scheme obtains optimal results and that the cloud cannot learn private information from the client's concealed data. We conduct extensive simulations on Amazon Elastic Cloud Computing (EC2) platform and find that our proposed scheme provides significant time savings to the clients.

Unsupervised anomaly detection on multidimensional time series data is a very important problem due to its wide applications in many systems such as cyber-physical systems, the Internet of Things. Some existing works use traditional variational autoencoder (VAE) for anomaly detection. They generally assume a single-modal Gaussian distribution as prior in the data generative procedure. However, because of the intrinsic multimodality in time series data, previous works cannot effectively learn the complex data distribution, and hence cannot make accurate detections. To tackle this challenge, in this paper, we propose a GRU-based Gaussian Mixture VAE system for anomaly detection, called GGM-VAE. In particular, Gated Recurrent Unit (GRU) cells are employed to discover the correlations among time sequences. Then we use Gaussian Mixture priors in the latent space to characterize multimodal data. The proposed detector reports an anomaly when the reconstruction probability is below a certain threshold. We conduct extensive simulations on real world datasets and find that our proposed scheme outperforms the state-of-the-art anomaly detection schemes and achieves up to 5.7% and 7.2% improvements in accuracy and F1 score, respectively, compared with existing methods.

With the onset of the big data era, designing efficient and effective machine learning algorithms to analyze large-scale data is in dire need. In practice, data is typically generated by multiple parties and stored in a geographically distributed manner, which spurs the study of distributed machine learning. Traditional master-worker type of distributed machine learning algorithms assumes a trusted central server and focuses on the privacy issue in linear learning models, while privacy in nonlinear learning models and security issues are not well studied. To address these issues, in this paper, we explore the blockchain technique to propose a decentralized privacy-preserving and secure machine learning system, called LearningChain, by considering a general (linear or nonlinear) learning model and without a trusted central server. Specifically, we design a decentralized Stochastic Gradient Descent (SGD) algorithm to learn a general predictive model over the blockchain. In decentralized SGD, we develop differential privacy based schemes to protect each party’s data privacy, and propose an l-nearest aggregation algorithm to protect the system from potential Byzantine attacks. We also conduct theoretical analysis on the privacy and security of the proposed LearningChain. Finally, we implement LearningChain on Etheurum and demonstrate its efficiency and effectiveness through extensive experiments.

Paradigm-shifting systems such as cyber-physical systems, collect data of high- or ultrahigh- dimensionality tremendously. Detecting outliers in this type of systems provides indicative understanding in wide-ranging domains such as system health monitoring, information security, etc. Previous dimensionality reduction based outlier detection methods suffer from the incapability of well preserving the critical information in the low-dimensional latent space, mainly because they generally assume an isotropic Gaussian distribution as prior and fail to mine the intrinsic multimodality in high dimensional data. Moreover, most of the schemes decouple the model learning process, resulting in suboptimal performance. To tackle these challenges, in this paper, we propose a unified Unsupervised Gaussian Mixture Variational Autoencoder for outlier detection. Specifically, a variational autoencoder firstly trains a generative distribution and extracts reconstruction based features. Then we adopt a deep brief network to estimate the component mixture probabilities by the latent distribution and extracted features, which is further used by the Gaussian mixture model to estimate sample densities with the Expectation-Maximization (EM) algorithm. The inference model is optimized jointly with the variational autoencoder, the deep brief network, and the Gaussian mixture model. Afterwards, the proposed detector identifies outliers when the estimated sample density exceeds a learned threshold. Extensive simulations on six public benchmark datasets show that the proposed framework outperforms state-of-the-art outlier detection schemes and achieves, on average, 27% improvements in F1 score.

Solving large-scale sparse linear systems of equations (SLSEs) is one of the most common and fundamental problems in big data, but it is very challenging for resource-limited users. Cloud computing has been proposed as a timely, efficient, and cost-effective way of solving such expensive computing tasks. Nevertheless, one critical concern in cloud computing is data privacy. Specifically, clients’ SLSEs usually contain private information that should remain hidden from the cloud for ethical, legal, or security reasons. Many previous works on secure outsourcing of linear systems of equations (LSEs) have high computational complexity, and do not exploit the sparsity in the LSEs. More importantly, they share a common serious problem, i.e., a huge number of memory I/O operations. This problem has been largely neglected in the past, but in fact is of particular importance and may eventually render those outsourcing schemes impractical. In this paper, we develop an efficient and practical secure outsourcing algorithm for solving large-scale SLSEs, which has low computational and memory I/O complexities and can protect clients’ privacy well. We implement our algorithm on Amazon Elastic Compute Cloud, and find that the proposed algorithm offers significant time savings for the client (up to 74 percent) compared to previous algorithms.

Device-to-device (D2D) communications can potentially alleviate cellular network congestion by utilizing local available links, and have attracted intensive attention recently. Cognitive radio (CR) allows users to opportunistically access unused licensed spectrums. It thus serves as a great candidate technology for D2D communications, but has not been widely employed in cellular networks due to hardware development limitations. In this paper, we propose a new architecture, called cognitive mesh assisted cellular network (CMCN), in which several secondary service providers (SSPs) deploy CR routers to facilitate D2D communications among wireless users. To address the competition among the SSPs, we further construct a secondary spectrum auction market. Although a few works have studied spectrum auctions, most of them are designed for single-hop communications, and it is usually not clear whom a winning user communicates with. Uncertain spectrum availability is not considered in previous schemes either. In this paper, we propose a transmission opportunity auction scheme, called TOA, which can address these problems. Extensive simulations are conducted to validate the efficiency of the CMCN architecture and that of the TOA scheme.

Electric power systems are critical infrastructure and are vulnerable to contingencies including natural disasters, system errors, malicious attacks, etc. These contingencies can affect the world's economy and cause great inconvenience to our daily lives. Therefore, security of power systems has received enormous attention for decades. Recently, the development of the Internet of Things (IoT) enables power systems to support various network functions throughout the generation, transmission, distribution, and consumption of energy with IoT devices (such as sensors, smart meters, etc.). On the other hand, it also incurs many more security threats. Cascading failures, one of the most serious problems in power systems, can result in catastrophic impacts such as massive blackouts. More importantly, it can be taken advantage by malicious attackers to launch physical or cyber attacks on the power system. In this paper, we propose and investigate cascading failure attacks (CFAs) from a stochastic game perspective. In particular, we formulate a zerosum stochastic attack/defense game for CFAs while considering the attack/defense costs, budget constraints, diverse load shedding costs, and dynamic states in the system. Then, we develop a Q-CFA learning algorithm that works efficiently in power systems without any a priori information. We also formally prove that the convergence of the proposed algorithm achieves a Nash equilibrium. Simulation results validate the efficacy and efficiency of the proposed scheme by comparisons with other state-of-the-art approaches.

The massive amount of data that is being collected by today's society has the potential to advance scientific knowledge and boost innovations. However, people often lack sufficient computing resources to analyze their large-scale data in a cost-effective and timely way. Cloud computing offers access to vast computing resources on an on-demand and pay-per-use basis, which is a practical way for people to analyze their huge data sets. However, since their data contain sensitive information that needs to be kept secret for ethical, security, or legal reasons, many people are reluctant to adopt cloud computing. For the first time in the literature, we propose a secure outsourcing algorithm for large-scale quadratic programs (QPs), which is one of the most fundamental problems in data analysis. Specifically, based on simple linear algebra operations, we design a low-complexity QP transformation that protects the private data in a QP. We show that the transformed QP is computationally indistinguishable under a chosen plaintext attack (CPA), i.e., CPA-secure. We then develop a parallel algorithm to solve the transformed QP at the cloud, and efficiently find the solution to the original QP at the user. We implement the proposed algorithm on the Amazon Elastic Compute Cloud (EC2) and a laptop. We find that our proposed algorithm offers significant time savings for the user and is scalable to the size of the QP.

The security and privacy threats on e-commerce have attracted intensive attention recently. The explosive growth of online social networks (OSNs) has made them potential new great marketplaces for e-commerce, which, however, raise serious security and privacyconcerns. This is mainly due to the centralized system architecture where the service provider knows all users’ private data and becomes the single point of failure. To this end, we propose a secure and private auction framework, called SPA, for decentralized online social networks (DOSNs). SPA consists of three phases: identity initiation, buyer-seller matching, and private auction. It requires no trust among the participants but can provide security, privacy, authenticity, non-repudiation, and correctness for the auctions. We analyze the computation and communication complexities of the proposed private auction scheme, which are$O(n+K)$for each node where$n$is the number of bidders and$K$is the number of pricing points. In contrast, those of previous auction schemes are$O(nK)$at best. The storage complexity is significantly lower than before as well. Security and privacy of SPA are also analyzed. Extensive experiments are conducted to validate the efficiency of SPA.

With the ever increasing user adoption of mobile devices like smart phones and tablets, the cellular service providers' energy consumption and cost are fast-growing and have received tremendous attention. How to effectively reduce the energy cost of cellular networks and achieve green communications while satisfying cellular users' rocketing traffic demands has become an urgent and challenging problem. In this paper, we investigate the minimization of the long-term time-averaged expected energy cost of a cellular service provider while guaranteeing the strong stability of the network. We first formulate an offline optimization problem with a joint consideration of flow routing, link scheduling, and energy (i.e., renewable energy resource, energy storage unit, etc.) constraints. Since the formulated problem is a time-coupling stochastic Mixed-Integer Non-Linear Programming (MINLP) problem, it is prohibitively expensive to solve. Then, we reformulate the problem by employing Lyapunov optimization theory. A decomposition based algorithm is developed to solve the problem, which is proved to guarantee the network strong stability. Both the lower and upper bounds on the optimal result of the original problem are derived and proven. Simulation results demonstrate that the obtained lower and upper bounds are very tight, and that the proposed scheme results in noticeable energy cost savings.