Skip to main content

RAIS: Real-time Analytics for the Internet of Sports

Periodic Reporting for period 1 - RAIS (RAIS: Real-time Analytics for the Internet of Sports)

Reporting period: 2019-01-01 to 2020-12-31

New emerging sensing technologies of the Internet of Things (IoT) in our global interconnected world present us with unprecedented amounts of data, available at highly heterogeneous and distributed data sources. IoT is a greenfield market. New players, with new business models, approaches, and solutions, appear and overtake incumbents. Existing IoT solutions are expensive because of the high infrastructure and maintenance costs associated with centralized clouds, large server farms and networking equipment. The sheer amount of communication that will have to be handled when IoT devices grow to the tens of billions will increase those costs substantially. Even if the unprecedented economical and engineering challenges are overcome, cloud servers will remain a bottleneck and point of failure that can disrupt the entire network. This makes Privacy Preserving Decentralized Big Data Analytics a main challenge for the success of future IoT initiatives that capitalize on mining data coming from personal devices by uncovering the collective behaviour and latent phenomenon arising in such systems. Within this project we will focus on developing new technologies for Big Data Analytics on the Edge, Data Stream Processing, Distributed and Decentralized Machine Learning as well as Security/Privacy under one RAIS collective awareness platform with particular focus on sports and recreational activities. RAIS’s main objective is to train a new generation of creative, entrepreneurial and innovative young researchers in Real-time Analytics for the Internet of Sports.
RAIS consortium conducts research through four technical workpackages: WP1 Distributed Sensing Infrastructure & Networking, WP2 Security, Privacy and Trust, WP3 Data Mining & Edge Analytics and WP4 Predictive Analytics for IoS.
WP1 involves 3 ESRs and aims at designing the overlay network infrastructure comprising peers that provide the platform and the communication functionality required for basic collective sensing services and for deploying new applications in the context of RAIS. Within the reporting period and in the first year of research activities (2020) the work within WP1 was focused on studying literature and related work, as well as introducing a novel architecture for user-centered decentralized marketplace (PDS2) for privacy preserving data processing. The architecture employs blockchain technology, privacy-preserving computation and decentralized aggregation techniques. PDS2 ensures that the providers do not lose control of who, how and when can use their data, do not need to sacrifice the privacy of their data and receive adequate compensation for the value they provide to consumers.
WP2 involves 4 ESRs and aims at developing new security and privacy-preserving tools and mechanisms for the IoS domain, following a user-centric view. Within the reporting period the work within WP2 was focused on the literature study on privacy preference specification and enforcement, with a particular focus on IoT and IoS ecosystems. The study also targeted the main learning approaches that can be used for access control policy/privacy preference learning as well as on malware detection in the IoT domain, with a particular focus on distributed solutions. ESRs within WP2 performed the analysis of various privacy leakages and security vulnerabilities that occur in wearable devices and their mobile applications as well as on the cyber attacks for fitness data, privacy policies and preferences of wearable fitness trackers, and privacy risks of the aggregated disclosure of fitness data.
WP3 involves 3 ESRs. During the reporting period the efforts within WP3 were focused on studying literature and related work, as well as developing graph and time series mining algorithms for centralized and semi-centralized settings. The focus was on studying role discovery in networks, with particular aim to recognize structurally similar nodes, that is, nodes that show a similar connectivity pattern in the way they are connected to their neighbors. Roles are especially relevant to the Internet of Sports because they are used for node classification, link prediction, anomaly detection, visualization, and for influence maximization, to better train distributed learning algorithms. A novel graph mining algorithm that can deal with structurally similar nodes was developed. Furthermore, ESRs within WP3 studied the literature on graph representation learning (GRL) techniques and in particular context-aware representation learning was investigated to account for multiple-facet nature of nodes in the graphs/networks. Furthermore, it was explored uncertainty estimation in the context of time-series classification with deep learning techniques. An ensemble based approach called “Deep time-ensembles” was proposed for Human Activity Recognition (HAR) with wearables.
WP4 involves 4 ESRs and aims at detecting exercise habits and Social Contagion and also building a real-time anomaly detection engine for IoS data streams. The fellows within WP4 studied the psychology of physical activity behavior and habits, performed a review of the different habit and intention measurements used in the discipline of psychology, and studied security and privacy aspects of data sharing and communication. Initial data investigation and explanatory data analysis was performed on a very large geolocation data collected before and during the COVID-19 pandemic. Furthermore, the ESRs studied the literature following Kitchenham’s widely-recognized guidelines and a well-defined protocol, and has conducted a systematic review. Finally, the ESRs studied the loss of privacy from data sharing as well as from communication between smartbands and the cloud. Initial results have already been published in relevant conferences.
The RAIS European Training Network (ETN) provides for 14 Early Stage Researchers (PhD students) a world-class training within a broad spectrum of subjects establishing a fertile inter-disciplinary research and innovation community that will advance: 1) wearable sports-sensing and quantified-self devices and accompanying middleware; 2) decentralized block-chain powered IoT platforms for Data mining and Machine Learning, 3) real-time edge analytics and predictive modelling to capture a broad range of sports-related data and trends, critical in a variety of application settings. Within RAIS, the intersectoral collaboration between academia, entrepreneurs, business developers, and industry will strengthen novelty and impact by ensuring that relevant needs are addressed, and that research results are both economically and technically feasible.