Resumen de Mobile crowd sensing platform for data quality and users privacy protection
Cristian TANAS
Mobile Crowd Sensing (MCS) represents a paradigm shift in sensor networking research, where the task of collecting information regarding a particular phenomena of interest is distributed among a crowd of possibly anonymous volunteers who collectively sense and share data from their surrounding environment through their sensor-enabled mobile devices. Human participation is the fundamental building block of this new kind of sensor networks and the inherent mobility of individuals offers unprecedented opportunities for both sensing coverage and effective data transmission.
The MCS paradigm entails unique research challenges, most of them based on the fact that human participation is in the loop. Therefore, we find user participation, reliability of the sensed data and users concerns on privacy to prevail as the three most essential issues that should be addressed before deploying any MCS network in the wild. Indeed the usefulness of any MCS network heavily depends on the individuals willingness to participate and to sense and share data. Furthermore, in MCS scenarios there is, a priori, no control over the individuals contributing sensed data. Thus, we cannot assume that all individuals will be equally honest. Counterfeit data could be provided by individuals in order to maximize their own benefits. Finally, the sensed data may enclose sensitive information regarding the individual (e.g. location, voice samples, among others), which could lead to a potential threat to the individual's privacy.
In this thesis, we address all the aforementioned issues and provide a proof-of-concept implementation of a typical MCS application. For each of the main research challenges present in MCS scenarios, we propose a mechanism to minimize their impact on the performance of the processes of a MCS network. To that end, we propose a reputation-based system, which combined with a collective knowledge model, allows us to determine a single measure for assessing the reliability of the contributed data. Furthermore, we propose a rewarding model that integrates reputation scores and incentives provided for contributing sensed data into a unique value, in a way that the reputation score determines the reward to be perceived and vice versa. Finally, both models are glued together using the Bitcoin cryptocurrency, which, in turn, provides an anonymity scheme that protects the users' anonymity while maintaining the same level of rewarding and reputation.
