Computer Science Building, Room 142
Individuals are increasingly encouraged to share private information with
service providers. Privacy is relaxed to increase the utility of the data for
the provider. This dissertation offers an alternative approach in which raw
data stay with individuals and only coarse aggregates are sent to analysts. A
challenge is the reliance on constrained devices for data collection. This
dissertation demonstrates the practicality of this approach by designing and
implementing privacy-aware systems that collect information using low-cost or
ultra-low-power microcontrollers. Smart meters can generate certified readings
suitable for use in a privacy-preserving system every 10 s using a Texas
Instruments MSP430 microcontroller. CRFIDs---batteryless devices that operate
on harvested energy from RF---can generate encrypted sub-aggregates in 17 s to
contribute to a privacy-preserving aggregation system that does not rely on a
trusted aggregator. A secure communication channel for CRFID tags via untrusted
relays achieves a throughput of 18 Kbps.
Advisor: Kevin Fu