AHPCRC Projects
Project 3-3: Secure Sensor Data Dissemination and Aggregation |
 |
|
|
||
| Phantom routing in a secure network (Hong Huang, NMSU). | ||
| Return to main article. | ||
| Sidebar: Data Aggregation The topic of in-network data aggregation has gained importance in the wireless sensor networks community over the last several years. For nearly all node technologies, data transmission over wireless links costs hundreds to thousands of times more energy than performing local computation on the same data. For untethered, energy-constrained (e.g., battery-operated) sensor nodes monitoring a physical environment, energy conservation and network lifetime considerations become significant, and it is desirable to trade local computation for communication as much as possible. Sensor nodes can be equipped with computational capabilities, so that they can compute and transmit cumulants, moments, or summaries directly, thus avoiding the expensive transmission of all sensor data to a (possibly distant) base station. After an aggregation tree or forest is formed, data aggregation can be conducted by having the data flow up the tree and aggregated in the standard way. Data aggregation can be integrated with the tree formation process so that they both happen simultaneously. In the tree formation procedure, random priorities are used to determine which probe message should proceed. These random priorities can be used to implement data aggregation, by using probabilistic counting techniques. Essentially, the formation of the tree achieves the computation of the minimum of some random numbers chosen by the hot nodes, with which one can derive an approximation to the final aggregation value. Because the aggregation is done by computing minima, the framework is robust with respect to such factors as packet duplication and multiple encounter events between packets from the same two nodes. Reference: Source: AHPCRC Bulletin, Vol. 1, Issue 2 (2008) |
||
