Personal data are constantly collected and shared via web cites, mobile applications like social networking and navigation apps, smart home devices like smart TVs and voice assistants, and IoT devices. Personal data are then monetized to support targeted advertising, personalized services, differential pricing, risk assessment and influencing public opinion. This happens at the expense of privacy and fairness for individuals and the society. To address this, governments around the world are enacting privacy laws, e.g. GDPR (European Union) and CCPA (California). Unfortunately, since profits can be at odds with privacy considerations, industry players have an incentive to circumvent the law. What is more, the technical concepts and associated tools developed so far and used by the laws are neither strong enough nor wide enough in scope. Last, users themselves are conflicted: they enjoy the plethora of personalized services but are alarmed by the loss of their privacy. In this proposal we advocate for a user-centered approach to privacy where each user may dictate how much privacy is willing to trade in exchange for services. We will systematically investigate the efficiency of state of the art privacy mechanisms, both formal, e.g. differential and information theoretic privacy, and data-driven, e.g. generative adversarial privacy, in terms of how well they protect data privacy while maintaining some utility of the obfuscated data and the services that depend upon them. We will do so not only via analysis but also via real world experiments in the context of applications at the forefront of personal data privacy leaks. We will also introduce novel privacy tools for real world use cases which allow users to select the desired level of data privacy and utility of service. Use cases of interest include mobile smartphone data leaks, online tracking via web browsing and apps usage, and user profiling within popular apps like video sharing.

The European Union policy for climate and energy imposes significant targets for a high integration of renewable energy sources in the period from 2020 to 2030. System operators have to deal with operational flexibility to respond to variability and to uncertainty of the renewable generation, ensuring the network reliability and security. While significant efforts have been made into the developing accurate forecasts, much work remains to integrate the forecasting in the electric system operations. The successful incorporation of forecasts into grid operation emerges as an important challenge. Accurate photovoltaic (PV) generation forecasts are major themes of the research roadmap of many international task forces, as Smart Grids SRA 2035 to support the flexibility increasing of the power systems. In this context, the project aims to support large scale integration of PV systems in countries with a high solar resource and a significant potential of small capacity PV systems such as Greece. The Institute of Communication and Computer Systems (ICCS) is the most important Hellenic research institute, committed to support Hellenic Electricity Distribution Network Operator S.A. (HENDO) that is dealing with a radical modernization of the existing network. The THINKPV project encourages the ICCS and its industrial partners to facilitate PV grid integration by the development of a probabilistic forecasting system based on machine learning, taking advantage of data that can be measured in the distribution network, in order to improve forecast accuracy compared to the state of art. The model will be assembled into a solar power forecasting system that will be operational at the Electric Energy Systems Laboratory (EESL) of the ICCS to operate directly with tools for simulating power system operations. A prototype of operational solar forecasting systems will be demonstrated for HENDO, providing also a training program for its efficiency and correct application.

The FET project “EXTRA” (Exploiting eXascale Technology with Reconfigurable Architectures) aimed at devising efficient ways to deploy ultra-efficient heterogeneous compute nodes for future exascale High Performance Computing (HPC) applications. One major outcome was the development of a framework for mapping applications to reconfigurable hardware, relying on the concept of Decoupled Access – Execute (DAE) approach. This project focuses on the commercialization of the EXTRA framework for cloud HPC platforms. More specifically, it targets the deployment of Virtual Machines (VMs) that integrate the EXTRA DAE Reconfigurable Architecture (EDRA) on custom hardware within the cloud infrastructure of one of the largest cloud service providers, the Amazon Web Services marketplace. End-users will be able to automatically map their applications to “EDRA-enhanced” VMs, and directly deploy them onto Amazon’s cloud infrastructure for optimal performance and minimal cost. Towards a successful exploitation outcome, the project will put effort on (a) applying the required software- and hardware-level modifications on the current EXTRA framework to comply with Amazon’s infrastructure, (b) devising a business strategy to effectively address various user groups, and (c) disseminating the benefits of this solution to large public events and summits.