Abstract
The subject of sensor network has received a lot of attentions recently. The software a sensor node executes is often statically preinstalled in its solid-state storage. This approach limits the functionality of a sensor node. In addition, this approach results in poor upgradeability of sensor software. In this thesis, we propose an efficient differential algorithm that considerably reduces the difference of two images into a small script file. We use this algorithm to derive the script file at the serve side. The script file is next sent to the sensor node along with the restoring code to dynamically upgrade its software without a large bulk of network transmission. The experimental results show that we successfully reduce the transmission traffic by more than 90% in several cases.