Secret Key Generation for Body-Wearable Wireless Sensor Devices

The last decade has witnessed a rapid surge in application of body sensor net- works, especially in the domains of military, healthcare and fitness. In such networks, wearable wireless sensor devices are used to measure and communi- cate a subject’s vital signs (such as heart rate, body temperature, blood glucose level, etc.) Security, in this context, is a critical issue as these devices deal with personal health data, requiring strict privacy and confidentiality. However, tra- ditional secret-key generation mechanisms (such as Diffie-Hellman) are typically computation and power intensive and not suitable for resource-constrained sen- sor devices. This thesis aims at realizing a practical and low-cost secret key generation mechanism for wearable sensor devices. First, we investigate a secret key generation mechanism that extracts shared secret keys using properties of the near-body wireless channel between two com- municating parties. For a fully body-worn scenario, our experimental results, using off-the-shelf IEEE 802.15.4 devices, indicate that this approach is feasible for dynamic scenarios where communicating devices are placed in non-line-of- sight positions on the body. We also suggest an enhancement for existing key generation mechanisms using a filtering mechanism which considerably reduces bit mismatches. Second, we employ a channel hopping technique to decorrelate secret bit ex- traction. Due to fast sampling rates, successive samples of channel properties are correlated in time, yielding weak keys with reduced entropy. To overcome this, we use channel hopping to increase channel diversity. We conduct extensive experi- ments to show that channel hopping increases frequency diversity and effectively decorrelates successive channel samples and thereby dramatically improving the strength of the secret key. Furthermore, we identify key parameters affecting performance, namely channel spacing, the number of available channels, and user activity. We show that it is possible to devise an optimal hopping strategy to maximize the benefit by balancing between channel spacing and the number of channels for high frequency and temporal diversity. In this thesis, we describe, implement and validate our solutions in real body- worn scenarios. It is hoped that our effort contributes to the research and devel- opment of usable body sensor networks.