Mobility in mobile sensor networks: A study of sensing performance and privacy

Yu Tak Ma, Purdue University

Abstract

Recent advances in sensor technologies have made sensors more economical, power efficient, and portable to be mounted onto handheld devices for monitoring different environmental factors, and made mobile sensor networks possible. When it is financially infeasible to deploy enough or an excessive number of sensors, while the environmental factors they monitor are critical for public health and safety, such as chemical or radiation monitoring, we deploy mobile sensors that move under our control. We also decide the best mobility strategy to achieve the desired goals. We propose and analyse mobility strategies that give a well-balanced performance for various goals which may be antagonistic. We notice that for a stochastic mobility algorithm, pausing at a location is well-justified to achieve better quality in event monitoring and a closer match with the expected monitoring time of a location by the sensor. We also notice that the quality of event monitoring at a location may not be proportional to the time the sensors spend at the location. In other cases when it is economical to deploy an excessive number of sensors to monitor the environment by attaching them to electronic devices owned by the public, traces of mobile nodes are collected to help design and analyse of such systems and evaluate the expected performance before deployment. We are interested in studying privacy leakage through trace publication. Although published traces have their identity being replaced consistently with random IDs, movements of mobile nodes can be openly observed by others, or they may be learned through web blogs, status in social networks, and causal conversations, etc. It is then possible for an attacker to learn the whole movement history of the participants, breaching the privacy protection. We study comprehensively attack strategies both analytically and experimentally using real and synthetic traces. We observe that with high probability an adversary can identify participants in the trace set with the current scale of trace collection and publication.

Degree

Ph.D.

Advisors

Yau, Purdue University.

Subject Area

Computer science

COinS