A Lightweight Fault Detection Method for Sensor Networks Based on Anomaly Execution Statistics

Fault detection of deployed senor networks is difficult due to server constrained resources and limited access to the networks. A lightweight fault detection method is proposed to address the problem of unknown software fault detection in post-deployment phase. By analyzing characteristics of node programs, a statistical model and corresponding test statistic are devised to detect unknown fault at runtime. Evaluation on typical sensor network applications shows that the storage and runtime overhead are acceptable. A case study further demonstrates the effectiveness of the proposed method.