|
|
||||
Data
Estimation for A SEcure Networked Sensor Environment (SENSE)
Networks of thousands of sensors present a feasible and economic solution to some of our most challenging problems, such as real-time traffic modeling, military sensing and tracking, and monitoring for bio-terrorist attacks. These small, fragile sensors are limited in energy, computational and storage resources. Many applications are dependent on the secure and reliable operation of the sensor network. The survivability of the network is threatened by resource limitations and security attacks. Applications also require and expect the sensor network to deliver data with little or no corruption and with minimal loss so that optimal knowledge may be obtained by intelligent fusion. This project designs and implements a multi-level architecture to address the security and survivability concerns of sensor networks. At the lowest level, to counteract security threats and vulnerabilities, power-aware secure routing mechanisms, a pre-attack pro-active intrusion avoidance mechanism and a reactive hierarchical intrusion detection system are presented. Given the unreliable nature of sensor networks, a data estimation scheme is developed to compensate for data corruption and data loss. A distributed sensor network testbed is constructed for performance evaluation. This project is funded by Department of Defense and is a collaborative effort involving four universities: Oklahoma State University, The University of Oklahoma, University of Tulsa, and Langston University. Data Estimation, which is briefly described below, is performed by OUDB - the DataBase research group at The University of Oklahoma. Due to loss and corruption of data caused by intrusion attacks, node failures, or battery depletion in wireless sensor networks, data may not be available when needed. Simply requesting sensors to resend the missing data is not a feasible solution since it will both increase the query response time and consume more computer resources and power from the server and sensors. A more favorable solution is to estimate the missing data by developing efficient data estimation algorithms, so that we can use the estimated results to replace the missing sensor data and respond to user queries in real time.
In this research we propose data estimation algorithms using association rule mining on stream data to compensate for missing and corrupted data. Although originally derived for transactions of sale items, mining for association rules can be extended to other fields as well. In our case we can view a set of sale items as a set of all sensors, a sale transaction database as a set of sensor readings, and any two related itemsets as two particular sensors in our network. It is reasonable to expect that some sensors may be related to each other. Our initial goal is to find the relationships between sensors, and later use these relationships (or association rules) to estimate the values of missing sensor readings.
TEST APPLICATIONS PROJECT TEAM PUBLICATIONS
|
||||
For problems or questions regarding this web contact database@cs.ou.edu. |