Identifier

etd-05182005-154832

Degree

Doctor of Philosophy (PhD)

Department

Computer Science

Document Type

Dissertation

Abstract

Wireless sensor networks consisting of numerous tiny low power autonomous sensor nodes provide us with the remarkable ability to remotely view and interact with the previously unobservable physical world. However, incorporating computation intensive security measures in sensor networks with limited resources is a challenging research issue. The objective of our thesis is to explore different security aspects of sensor networks and provide novel solutions for significant problems. We classify security mechanisms into two categories - active category and passive category. The problem of providing a secure communication infrastructure among randomly deployed sensor nodes requires active security measurements. Key pre-distribution is a well-known technique in this class. We propose a novel 2-Phase technique for key pre-distribution based on a combination of inherited and random key assignments from the given key pool to individual sensor nodes. We develop an analytical framework for measuring security-performance tradeoffs of different key distribution schemes. Using rigorous mathematical analysis and detailed simulation, we show that the proposed scheme outperforms the existing solution in every performance aspect. Secure data aggregation in wireless sensor networks is another challenging problem requiring active measures. We address the problem of stealthy attack where a compromised node sends wrong/fictitious data as a reply to a query. We propose a novel probabilistic accuracy model which enables an aggregator to compute accuracy of each sensor reading by exploiting spatial correlation among data values. We also propose some novel, energy efficient statistical methods to enable a user accept the correct value with a high probability. Increasing network lifetime is a passive security mechanism which enables many security mechanisms to work more efficiently. We define length-energy-constrained optimality criteria for energy-optimized routes that impose uniform energy distribution across the network, thus preventing expedited network partition. We propose three different distributed, nearly-stateless and energy efficient routing protocols that dynamically find optimal routes and balance energy consumption across the network. We show that global energy information acquired through this process utilized in conjunction with energy depletion control in the sensornet ensures a significant improvement in terms of network lifetime.

Date

2005

Document Availability at the Time of Submission

Release the entire work immediately for access worldwide.

Committee Chair

Rajgopal Kannan

Share

COinS