UVM Theses and Dissertations
Format:
Print
Author:
Simonova, Ekaterina
Dept./Program:
Computer Science
Year:
2007
Degree:
MS
Abstract:
This thesis considers a key predistribution problem for wireless sensor networks. The problem refers to distributing secret keys among sensors prior to deployment to maximize performance metrics using limited resources available to the sensors. Three important metrics used to evaluate a Key Predistribution Scheme are (1) connectivity that indicates the likelihood that a pair of sensors is able to establish a secure communication path, (2) resilience that shows the capability to maintain secure communication when a number of sensors are compromised, and (3) scalability that characterizes the capability of the scheme to accommodate network growth.
Schemes appearing in the literature can be classified into two categories: basic schemes that achieve a fixed probability of sharing a key between any pair of sensors in the network, and location-aware schemes that use a priori knowledge about sensors' communication needs, such as, location information, to provide connectivity only among sensors that need to and can communicate with each other. Location-aware schemes achieve performance enhancement over the basic schemes by using memory efficiently by requiring sensors to share keys only with a small fraction of sensors located in the communication range and not all the sensors from the network.
However, existing location-aware schemes are based on specific schemes and are not compatible with combinatorial methods that use a set of key groups to generate sensors' key rings. Combinatorial methods are appealing as they achieve deterministic performance close to optimal, often outperforming other schemes based on probabilistic approaches. In addition, existing location-aware solutions do not have enough flexibility in terms of the trade-off between connectivity and resilience.
The contribution of the thesis is the design of the high-level key predistribution framework to generate sensors' key pools to be used to distribute keys among sensors. The framework uses location information to construct key pools in a way that only key pools of sensors located close to each other share keys. The framework can be used with any existing key predistribution scheme to construct sensors' key rings from the key pool, including combinatorial schemes, and is not limited to just a special type of schemes. Using the framework on top of a key predistribution scheme significantly improves resilience as the framework achieves localization of a key usage. Special attention is paid to the heterogeneous sensor networks consisting of nodes with different amount of memory and communication ranges. Good performance of the framework is confirmed by providing both experimental and analytical results.
Schemes appearing in the literature can be classified into two categories: basic schemes that achieve a fixed probability of sharing a key between any pair of sensors in the network, and location-aware schemes that use a priori knowledge about sensors' communication needs, such as, location information, to provide connectivity only among sensors that need to and can communicate with each other. Location-aware schemes achieve performance enhancement over the basic schemes by using memory efficiently by requiring sensors to share keys only with a small fraction of sensors located in the communication range and not all the sensors from the network.
However, existing location-aware schemes are based on specific schemes and are not compatible with combinatorial methods that use a set of key groups to generate sensors' key rings. Combinatorial methods are appealing as they achieve deterministic performance close to optimal, often outperforming other schemes based on probabilistic approaches. In addition, existing location-aware solutions do not have enough flexibility in terms of the trade-off between connectivity and resilience.
The contribution of the thesis is the design of the high-level key predistribution framework to generate sensors' key pools to be used to distribute keys among sensors. The framework uses location information to construct key pools in a way that only key pools of sensors located close to each other share keys. The framework can be used with any existing key predistribution scheme to construct sensors' key rings from the key pool, including combinatorial schemes, and is not limited to just a special type of schemes. Using the framework on top of a key predistribution scheme significantly improves resilience as the framework achieves localization of a key usage. Special attention is paid to the heterogeneous sensor networks consisting of nodes with different amount of memory and communication ranges. Good performance of the framework is confirmed by providing both experimental and analytical results.