The recent advances in communication technologies and data analytics are opening the way for Social Sensing, i.e. large-scale analysis of individual-level data from smart devices. The collection and analysis of personal data from e.g. smartphones, wearables, IoT devices, Domotic devices, open up new opportunities for managing and regulating cities and society, and for offering new services to citizens and customers.
Nevertheless, the same data can potentially enable profiling, discrimination, surveillance and other privacy-intrusive practices. Such concerns, even if unfounded, can undermine the adoption of technology and reduce the benefits of Social Sending.
There are several approaches to privacy-preservation, the majority of which involve changes in the infrastructure, e.g. in the communication protocol, in the way data is collected or processed. Such changes come at a cost for the service provider which conflict with the goal of privacy-preservation.
The question my research tries to answer is: how can users increase their privacy without requiring changes in the infrastructure (bottom-up approach) and without undermining the quality of service?
I develop new communication protocols that increase individual privacy by dynamically changing the contents of communication, while keeping compatibility with the existing infrastructure and analytics algorithms.
My methodology is a combination of agent-based computer simulations, machine learning and data analysis.
In collaboration with: Catholijn Jonker, Evangelos Pournaras
Learning, Evolution and Social Behavior
The theory of evolution states that an organism that is better adapted to its environment will be favored by natural selection over other organisms and eventually replace them. The intuitive conclusion that evolution selects individuals based on their genetic configuration (called genome) is mostly wrong: individuals are selected on their expressed characteristics and behavior (called phenome) which are in part conditioned by the genome.
This is particularly evident for social animals, ranging from ants to humans, whose behavior is conditioned by the interaction between individual genomes, social interactions and learning.
The interactions between these three components are not yet completely understood, hence the question that motivates my work is: How do learning, evolution and social behavior interact, and how these interactions condition the course of evolution?
I develop agent-based computer simulations with hundreds of agents, each of which is equipped with a neural network, and study how individual behaviors and genetic configurations co-evolve.
In collaboration with: Leonel Aguilar, Dirk Helbing, Leonard Wossnig, Johannes Thiele