Identifying packet droppers in sensor networks via report analysis

In an unattended sensor network, sensor nodes can be compromised. Based on compromised nodes, the adversary can launch various attacks and packet dropping is one of the easiest attacks. Many schemes have been proposed to mitigate the packet dropping attack, but few can effectively and efficiently identify the droppers. In this thesis, we propose a simple yet effective scheme to identify packet droppers. This scheme only requires sending and forwarding nodes to report their observations to the base station, and the base station can analyze the reports, identify inconsistencies in the reports, and then locate the droppers. All nodes are organized into a tree rooted at the base station and each node is required to report the number of packets it has received as well as the number and the composition of packets forwarded by its parent node on the tree, which it has overheard. Using rules we propose, the base station can analyze the received reports to check if there is inconsistency among the reports; if some inconsistencies are found, the base station can further infer the identities of packet droppers. The scheme can also tolerate erroneous reports, natural packet dropping and so on. A mark-based scheme is also proposed to identify packet dropper or modifier. Extensive simulations have been conducted to demonstrate the effectiveness of the scheme