Nowadays, mobile devices and networks allow us to be constantly connected anywhere. Often, these interactions leave a digital footprint that we are not aware of. Data, such as location information linked to our posts or searches in our browser may be processed by ML techniques that might allow a third party to infer sensitive information about our lives, such as location footprint. Consequently, there is a growing concern about the loss of control over personal data and the potential negative impact in our lives. Technology is now at a crossroads: trying to balance privacy and utility in scenarios that combine massive exchange communications, big databases and distributed and collaborative ML techniques towards the network edge. Precisely, the COMPROMISE project combines our knowledge in the areas of security, privacy, communication protocols, quality of service and ML, to face privacy improvements in network protocols and prevent privacy attacks, protecting communications with mechanisms that balance the trade-off between utility and privacy.
Our proposals seek a suitable compromise between utility and privacy, i.e., between the utility obtained by the user from the use of the service and the utility of gathered data after privacy enhancement techniques (PET) have been applied. The proposed advances in these topics are at the intersection of the following technologies: ML, fog computing, security protocols, vehicular networks, sensing networks, massive and distributed storage systems and dynamic databases. Consequently, the combination of these technologies constitutes the skeleton of current and future applications: distributed data and computation over wireless networks and protocols. The different research teams participating in this proposal have extensive experience in some of those complementary technologies and in the provision of privacy:
UPC group. (i) Anonymization of dynamic stream databases with high utility and high protection guarantees for both syntactic and differential privacy models; (ii) Development of theory and methods of collaborative anonymization; (iii) Design mechanisms and algorithms to protect the privacy of sensitive data exchanged on mobile wireless networks (MWNs) during the operation of network protocols; (iv) Enhancing privacy-aware services over MWNs (e.g., vehicular networks and mesh networks).
UC3M group. (i) Confidentiality vulnerabilities in HTTPS over TLS 1.3 and QUIC, and develop improvements; (ii) Confidentiality vulnerabilities in DNS and develop improvements; (iii) Study attacks based on machine learning techniques in IoT protocols and develop improvements; (iv) Study the privacy threats related to the location of users and propose improvements.
UVIGO group. (i) Smart data sample and smart exchange for ML training; (ii) Models for collaborative & distributed inference; (iii) Optimal & scalable ML architecture; (iv) Discovering privacy vulnerabilities in a collaborative ML architecture.
Read More... Read Less...COMPROMISE is a coordinated project which consists of three subprojects, involving the following partners and team members:
MobilitApp is a tool to help to analyse the urban mobility in our cities. MobilitApp periodically records 3 sensors (accelerometer, magnetometer and gyroscope) from its users' smartphones. The information obtained is processed using machine learning techniques to predict the transportation mode used by the user, i.e. bicycle, bus, car, e-bicycle, e-scooter, metro, motorbike, run, stationary, train, tram and walk. The user's information is treated completely anonymously.
EVcompare allows users to compare different electric and combustion vehicles based on various criteria such as range, price, pollution, and total cost over the years. EVs help to reduce pollution due to transportation, reduced energy and maintenance costs, although the initial purchase cost is higher and battery manufacturing also pollutes. EVcompare takes into account all this inputs to estimate the anual cost of each selected EV compared to the selected combustion vehicle or to another EV.
DataSet 5RoutingMetrics VANET BCN. Leticia Lemus Cárdenas, Ahmad Mohamad Mezher, Mónica Aguilar Igartua. Universitat Politècnica de Catalunya (UPC), UPCommons, 18th October 2021. Dataset composed by five routing metrics (available bandwidth, distance to destination, vehicles' density, MAC layer losses, and vehicle' trajectory) gathered from vehicular network simulations in urban scenarios in Barcelona. The dataset is used to train and test a machine learning-based forwarding algorithm.
MobilitApp Sensor Dataset. Mónica Aguilar Igartua, Gerard Caravaca Ibáñez, Miquel Gotanegra Estañol. Universitat Politècnica de Catalunya (UPC), CORA repositori, 28th June 2024. Dataset composed by three smartphone sensors (accelerometer, magnetometer, gyroscope) gathered from smartphones of volunteers who labeled their unimodal trips while traveling in Barcelona. The dataset is used to train and test a machine learning-based activity prediction algorithm.
Copyright © 2021 COMPROMISE. Template by W3layouts