Efficiently running federated learning (FL) on resource-constrained devices is challenging since they are required to train computationally intensive deep neural networks (DNN) independently. DNN partitioning-based FL (DPFL) has been proposed as one mechanism to accelerate training where the layers of a DNN (or computation) are offloaded from the device to the server. However, this creates significant communication overheads since the intermediate activation and gradient need to be transferred between the device and the server during training. While current research reduces the communication introduced by DNN partitioning using local loss-based methods, we demonstrate that these methods are ineffective in improving the overall efficiency (communication overhead and training speed) of a DPFL system. This is because they suffer from accuracy degradation and ignore the communication costs incurred when transferring the activation from the device to the server. This article proposesEcoFed– a communication efficient framework for DPFL systems.EcoFedeliminates the transmission of the gradient by developing pre-trained initialization of the DNN model on the device for the first time. This reduces the accuracy degradation seen in local loss-based methods. In addition,EcoFedproposes a novel replay buffer mechanism and implements a quantization-based compression technique to reduce the transmission of the activation. It is experimentally demonstrated thatEcoFedcan reduce the communication cost by up to 133× and accelerate training by up to 21× when compared to classic FL. Compared to vanilla DPFL,EcoFedachieves a 16× communication reduction and 2.86× training time speed-up.

Information-centric networking (ICN) is a promising paradigm shift that aims to tackle the traditional Internet architectural problems and to fulfill the future Internet requirements. The traditional Internet architecture is a host-oriented architecture (i.e., TCP/Internet protocol (IP) approach) due to which the Internet of Things (IoT) have been facing issues related to data dissemination across the distant locations. Therefore, a quick comprehension to enhance the communication for improving the content transmission services is of upmost importance. To deal with the challenges of traditional IP networks, the ICN paradigm was proposed which is different from traditional IP networking in terms of: 1) naming; 2) routing and forwarding; and 3) caching. One of the most common and important features of ICN architectures is in-network caching, which can significantly reduce content retrieval latency and improve data availability. Furthermore, in an ICN-based IoT environment, content caching at intermediate network nodes reduces the path stretch between end users and caches the content to meet future demands. This article compares and thoroughly investigates ICN-based caching strategies in terms of content retrieval latency, cache hit ratio, stretch, and link load, with a focus on IoT-based environments. Following a thorough simulation study, we discovered that ICN in-network caching is one of the most beneficial features for enhancing IoT-based networks.

Applying federated learning (FL) on Internet of Things (IoT) devices is necessitated by the large volumes of data they produce and growing concerns of data privacy. However, there are three challenges that need to be addressed to make FL efficient: 1) execution on devices with limited computational capabilities; 2) accounting for stragglers due to computational heterogeneity of devices; and 3) adaptation to the changing network bandwidths. This article presentsFedAdapt, an adaptive offloading FL framework to mitigate the aforementioned challenges.FedAdaptaccelerates local training in computationally constrained devices by leveraging layer offloading of deep neural networks (DNNs) to servers. Furthermore,FedAdaptadopts reinforcement learning (RL)-based optimization and clustering to adaptively identify which layers of the DNN should be offloaded for each individual device on to a server to tackle the challenges of computational heterogeneity and changing network bandwidth. The experimental studies are carried out on a lab-based testbed and it is demonstrated that by offloading a DNN from the device to the serverFedAdaptreduces the training time of a typical IoT device by over half compared to classic FL. The training time of extreme stragglers and the overall training time can be reduced by up to 57%. Furthermore, with changing network bandwidth,FedAdaptis demonstrated to reduce the training time by up to 40% when compared to classic FL, without sacrificing accuracy.

The promising vision of information-centric networking (ICN) and of its realization, named data networking (NDN), has attracted extensive attention in recent years in the context of the Internet of Things (IoT) and wireless sensor networks (WSNs). However, a comprehensive NDN/ICN-based architectural design for WSNs, including specially tailored naming schemes and forwarding mechanisms, has yet to be explored. In this article, we present single-channel cluster-based information-centric WSN (CCIC-WSN), an NDN/ICN-based framework to fulfill the requirements of cluster-based WSNs, such as communication between child nodes and cluster heads (CHs), association of new child nodes with CHs, discovery of the namespace of newly associated nodes, and child node mobility. Through an extensive simulation study, we demonstrate that CCIC-WSN achieves 71%-90% lower energy consumption and 74%-96% lower data retrieval delays than recently proposed frameworks for NDN/ICN-based WSNs under various evaluation settings.

Nowadays, online applications are moving to the cloud, and for delay-sensitive ones, the cloud is being extended with edge/fog domains. Emerging cloud platforms that tightly integrate compute and network resources enable novel services, such as versatile IoT (Internet of Things), augmented reality or Tactile Internet applications. Virtual infrastructure managers (VIMs), network controllers and upper-level orchestrators are in charge of managing these distributed resources. A key and challenging task of these orchestrators is to find the proper placement for software components of the services. As the basic variant of the related theoretical problem (Virtual Network Embedding) is known to be NP-hard, heuristic solutions and approximations can be addressed. In this paper, we propose two architecture options together with proof-of-concept prototypes and corresponding embedding algorithms, which enable the provisioning of delay-sensitive IoT applications. On the one hand, we extend the VIM itself with network-awareness, typically not available in today's VIMs. On the other hand, we propose a multi-layer orchestration system where an orchestrator is added on top of VIMs and network controllers to integrate different resource domains. We argue that the large-scale performance and feasibility of the proposals can only be evaluated with complete prototypes, including all relevant components. Therefore, we implemented fully-fledged solutions and conducted large-scale experiments to reveal the scalability characteristics of both approaches. We found that our VIM extension can be a valid option for single-provider setups encompassing even 100 edge domains (Points of Presence equipped with multiple servers) and serving a few hundreds of customers. Whereas, our multi-layer orchestration system showed better scaling characteristics in a wider range of scenarios at the cost of a more complex control plane including additional entities and novel APIs (Application Programming Interfaces).

Blockchain is the most popular security network technology in the 21stcentury and still under active research. The consensus algorithms in Blockchain plays a major role and are crucial for maintaining the integrity and security of a distributed network. In addition, there are various problems arising from the consensus algorithms adopted in the existing Blockchain network. Therefore, it is of upmost importance to develop a consensus algorithm to address these problems. In this paper, therefore, we propose an efficient consensus algorithm in a Blockchain network environment consisting of several distributed users. First, we study the process of consensus among users to add new blocks to the existing Blockchain. Second, we introduce the other conventional consensus algorithms, analyze the problems of each algorithm, and propose Rock-Scissors-Paper (RSP) algorithm to mitigate these problems. The RSP algorithm is an algorithm that achieves consensus among distributed users using three static balanced variables, Rock (R), Scissors (S), and Paper (P) to avoid attacks by malicious participants. Furthermore, the propose RSP algorithm is compared with the existing consensus algorithms to predict the performance and its effects in a distributed mobile network environment.