Abstract: Fog computing enables location dependent resource allocation and low latency services, while fostering novel market and business opportunities in the cloud sector. Aligned to this trend, we refer to Fog-to-cloud (F2C) computing system as a new pool of resources, set into a layered and hierarchical model, intended to ease the entire fog and cloud resources management and coordination. The H2020 project mF2C aims at designing, developing and testing a first attempt for a real F2C architecture. This document outlines the architecture and main functionalities of the management framework designed in the mF2C project to coordinate the execution of services in the envisioned set of heterogeneous and distributed resources.

Abstract: The ability of fine-tuning the performance of Bluetooth Low Energy (BLE) communication is essential to create low-power wireless applications with heavy user interaction, such as smart thermostats or door locks. One of the key challenges when designing such applications is finding the right trade-off between a system's responsiveness and energy-efficiency. Although there exists research works that improve the performance of BLE communication, all these approaches focus on connection-based BLE. Most BLE-based applications, however, spend the majority of their time in connection-less device discovery, waiting for approaching users. The energy-efficiency and timeliness in this state are defined by parameters that are often statically set at compile time. Although supported by the BLE specifications, how to dynamically adapt these parameters to user behavior is still an open question. In this paper, we tackle this challenge and design a strategy to improve the energy-efficiency and responsiveness of BLE device discovery. Towards this goal, we model the device discovery process and identify its key parameters. We further design an adaptive advertising strategy that allows smart objects to adapt their device discovery parameters to the user behavior. We implement this adaptive strategy and measure its performance in a real-world application, the Nuki Smart Door Lock. Our experiments show that a smart lock using our strategy consumes 48% less energy while reducing the device discovery time by up to 63% compared to the use of static parameters. Furthermore, we discuss how nearby BLE devices can be used to inform the lock about approaching user devices and hence to improve its responsiveness in low-power phases even further.

Abstract: Electric vehicles (EVs) are rapidly becoming more common and ownership is set to rise globally in coming years. The potential impacts of increased EVs on the electrical grid have been widely investigated and in its current state, existing grid infrastructure will struggle to meet the high demands at peak charging hours. The limited range of electric cars compounds this issue. We therefore propose CognitiveCharge, a novel approach to predictive and adaptive disconnection aware opportunistic energy discovery and transfer for the smart vehicular charging. CognitiveCharge detects and reacts to individual nodes and network regions which are at risk of getting depleted by using implicit predictive hybrid contact and resources congestion heuristics. CognitiveCharge exploits localised relative utility based approach to adaptively offload the energy from parts of the network with energy surplus to depleting areas with non-uniform depletion rates. We evaluate CognitiveCharge using a multi-day traces for the city of San Francisco, USA and Nottingham, UK to compare against existing infrastructure across a range of metrics. CognitiveCharge successfully eliminates congestion at both ad hoc and infrastructure charging points, reduces the time that a vehicle must wait to charge from the point at which it identifies as being in need of energy, and drastically reduces the total number of nodes in need of energy over the evaluation period.

Abstract: Connected car technology promises to drastically reduce the num- ber of accidents involving vehicles. However, this technology de- pends on the vehicle precise location to work. The adoption of Global Positioning System (GPS) as a navigation device has limi- tations that compromise geolocation information in certain envi- ronments due to line-of-sight technology. This work introduces the Time Series Dead Reckoning System (TedriS ) as a solution for Dead Reckoning navigation in situations where GPS fails. TedriS is a simple, easy-to-implement and efficient dead reckoning naviga- tion system that uses Time Series Regression Models (TSRM) and the vehicle’s rear wheel speed sensor data from Control Area Net- work (CAN) to estimate its absolute position. The estimate position process is carried out in a training phase and in a predicting phase. In the training phase, a novel technique applies TSRM and stores the relationship between GPS and CAN data that will be used in the predicting phase. TedriS evaluation performance was performed using traces collected in the Campus of Federal University of Rio de Janeiro (UFRJ) - Brazil and with experiments with a robot carried out at French Research Institute in Computer Science (INRIA) - France. The results show an accuracy compatible with Dead Reckoning navigation state-of-art systems.

Abstract: Visible light positioning, or VLP, has emerged as a low-cost approach to enabling a variety of indoor location-based services for indoor smart spaces. However, a survey of existing approaches to VLP reveals some challenges to compare one system to the another. Advances in key areas are expected to enable new levels of performance at low cost. These include innovations at the source (LEDs, LDs/LIDAR, and ToF sensors), at the receiver (diversity receivers and AoA sensors), and in the design of the overall end-to-end VLP system. Again, comparing these improvements from one system to the another is difficult due to varying assumptions and operating conditions. In this paper we classify VLP techniques in an attemp to reconcile the wide range of characteristics. We also propose a new concept called an active zone in recognition that best performance is needed primarily in a subset of the volume of an indoor space. Finally, we show the performance of a base VLP system under the new metric and conclude with how our VLP testbed can be used to verify quantify the region we call the {\it active zone}.

Abstract: We explores the feasibility of using only Bluetooth received signal strength of a smart device to infer user context. Devices that use Bluetooth wireless communication constantly transmit their wireless signal. Adversaries or non-malicious applications can leverage that information to extrapolate the activities in which a user is engaged and the location of certain devices. We explore the accuracy of classification of a range of activities, including walking, typing, writing, and using a mouse. This classification is carried out by simply using the RSSI of a user-worn smart watch.

Abstract: LoRa and the LoRaWAN specification is a technology for Low Power Wide Area Networks (LPWAN) designed to allow connectivity for connected objects, such as remote sensors. Several previous works revealed various weaknesses regarding the security of LoRaWAN v1.0 (the official 1st draft) and these led to improvements included in LoRaWAN v1.1, released on Oct 11, 2017. In this work, we provide a first look into the security of LoRaWAN v1.1. We present an overview of the protocol and, importantly, present several threats to this new version of the protocol. Besides, we propose our own ramification strategies for the mentioned threats, to be used in developing next version of LoRaWAN. The threats presented were not previously discussed, they are possible even within the security assumptions of the specification and are relevant for practitioners implementing LoRa-based applications as well researchers and the future evolution of the LoRaWAN specification.

Abstract: Mobility plays an important role when analysing natural phenomenon. With regard to the electronic engineering and computer science, further understanding of mobility can inspire the better development of artificial systems and intelligent algorithms. In order to intelligently handle mobility, appropriate representation of movement behaviour is essential. As a classical mobility model, the random walk establishes a theoretical basis for analytical study. In literature, the random walk model and its variants have been mathematically analysed. However, to the best of our knowledge, no existing research has provided a path model for correlated random walks in two-dimensional Euclidean space. In this paper, the concept of on-path certainty is proposed to describe movement path and the spatial distribution of on-path certainty is experimentally solved and analysed. As a result, functional relationships of the proposed path model are revealed and discussed for further research.

Abstract: The transformation of existing power grids into Smart Grids (SGs) aims to facilitate grid energy automation for a better quality of service by providing fault tolerance and integrating renewable energy resources in the power market. This evolution towards a smarter electricity grid requires the ability to transmit in real time a maximum of data on the network usage. A Wireless Sensor Network (WSN) distributed across the power grid is a promising solution, given the reduced cost and ease of deployment of such networks. These advantages come up against the unstable radio links and limited resources of WSN. In order to reduce the amount of data sent over the network, and thus reduce energy consumption, data prediction is a potent solution of data reduction. It consists on predicting the values sensed by sensor nodes within certain error threshold, and resides both at the sensors and at the sink. The raw data is sent only if the desired accuracy is not satisfied, thereby reducing data transmission. We focus on time series estimation with Least Mean Square (LMS) for data prediction in WSN, in a Smart Grid context, where several applications with different data types and Quality of Service (QoS) requirements will exist on the same network. LMS proved its simplicity and robustness for a wide variety of applications, but the parameters selection (step size and filter length) can directly affect its global performance, choosing the right ones is then crucial. Having no clear and robust method on how to optimize these parameters for a variety of applications, we propose a modification of the original LMS that consists of training the filter for a certain time with the data itself in order to customize the aforementioned parameters. We consider different types of real data traces for the photo voltaic cells monitoring. Our simulation results provide a better data prediction while minimizing the mean square error compared to an existing solution in literature.