Crowdsensing is a paradigm for sensor networks that leverages existing sensors in the environment that are not owned or operated by the people using the sensor data. The advantage of crowdsensing over traditional sensor networks is that, with the pervasive deployment of sensor-heavy smart phones, there is a widespread sensor network available in most places at most times. This drastically reduces the overhead cost of acquiring information for many types of applications and experiments. Crowdsensing, however, introduces many open research challenges in areas such as energy efficiency, networking, security, privacy, and incentives.
One of our focuses is in the security and privacy concerns of crowdsensing systems. In crowdsensing systems, data is produced by people and devices that the system doesn't trust. The people provide data that may include sensitive information (or may be combined with other information to produce something the user considers sensitive). Additionally, research has shown that for people to continue participating in crowdsensing systems, they need to receive some benefit, so incentive systems have been developed, but ensuring the security of those systems from users that want to acquire incentives without contributing useful data to the system is still an open problem.
We have been funded by the National Science Foundation to conduct research in these areas of security and privacy in crowdsensing. Our recent work has resulted a system called CAPP which uses a neural network trained on context information that detects if a user collected data in the place and time they claim to have collected data. Preliminary results from this work were published at the IEEE Workshop on Context and Activity Modeling and Recognition (COMOREA) in 2020.