Using mobile data more intelligently

If the data from all smartphones worldwide is networked, we could all soon be carrying a global supercomputer in our pockets. The large computing capacity would improve the real-time collection and analysis of data. But there are still technical hurdles and data protection concerns to overcome. Scientists working on the SwissSenseSynergy project, which is funded by the Swiss National Science Foundation (SNSF), have addressed these issues and proposed new ways of collecting and using such data.The project focuses mainly on crowd sensing, where information about a specific location area can be collected by accessing smartphone sensors. A typical example of this is mapping apps that can detect traffic congestion using the smartphone's acceleration sensors. As the networked devices collect information about many aspects of our environment (e.g. movements, noises, people and air quality), crowd sensing can help us decide where to eat, what clothes to wear or how to travel. "This information can be used for a wide variety of applications: for marketing forecasts or to predict the behavior of groups of people," explains the project coordinator, Torsten Braun from the University of Bern. However, there are still a few hurdles to overcome for such crowd-sensing apps. There are conflicts between data collection, data protection and the impact on the user-friendliness of the smartphone. For example, hardware resources are impaired by data transmission and inadequate security measures can encourage identity theft.four teams have developed new concepts for improving crowd-sensing technology and practical recommendations for its application. Their work focuses on four key areas: more accurate location data, improved data protection, industrial applications, more efficient data collection.Tracking function exceeds GPS standardTorsten Braun's team in Bern improved the location accuracy in buildings and under the ground to 1.1 meters in 90 percent of cases. This is roughly equivalent to the performance of GPS systems. Only the sensor data of the devices and radio signals are used, which, unlike GPS, also penetrate walls and concrete. The researchers collect the data transmitted by the smartphone sensors and information on WiFi signal strength. This information is then processed by several machine learning algorithms. "The next step is then to determine the location that the users are heading for," says Braun. "This could be interesting for shopping centers or train stations, for example. "Scientists from the Universities of Bern and Geneva have jointly developed a mobile app that combines crowd sensing with indoor positioning and smart spaces. This new mobile app integrates sophisticated location algorithms and sensor measurements with location data, which are then stored in a cloud. There, the data is available for the Internet of Things and can be used in personalized and location-based automation applications for numerous "smart" objects and products.A team from the University of Applied Sciences of the Italian-speaking part of Switzerland (SUPSI) in Lugano has developed models that use predictive location data for data transmission in social media. The experiments have shown that fast data transmission works not only in social networks such as Facebook and Twitter, but also in the physical ad hoc networks formed by cell phones. The messages could respond to local behavior, evaluate feedback in real time and circulate faster between selected users. The research provides a deeper understanding of the social influences on human behavior and shows connections between physical locations, shared preferences and event-based social communities.Artificial noise as data protection"One of the biggest difficulties for researchers is the balancing act between data use and data protection," says Torsten Braun. "The accuracy of the data can come at the expense of privacy." If user data is also recorded during data collection, the willingness to participate decreases. To ensure the security of the data, the team at Chalmers University of Technology in Sweden has developed machine learning methods for data analysis and automated decision-making that enable "differential data protection". The personal data is protected by a carefully tuned noise (random data) that is introduced into the data collected by the devices, and researchers at the University of Geneva have addressed another conflict: the desire to collect the largest possible amount of data with the lowest possible hardware load through crowd sensing. If users fear a burden on their smartphone, they may reject apps that access otherwise unused sensors. This project investigates game-theoretic models to find out how the load can be distributed across multiple phones and users. As part of a field experiment in San Francisco, volunteer participants downloaded an app to map noise levels in the city. While collecting useful data for the city administration, they simultaneously tested different methods for distributing the load across multiple devices.SwissSenseSynergy's interdisciplinary approach has developed new techniques that could be useful for research and applications. Under the name Vivo, the project is developing a new type of test arrangement that relies on voluntary participants in the pilot phase of development.