Information-theoretic modeling of false data filtering schemes in wireless sensor networks

Abstract

False data filtering schemes are designed to filter out false data injected by malicious sensors; they keep the network immune to bogus event reports. Theoretic understanding of false data filtering schemes and guidelines to further improve their designs are still lacking. This article first presents an information-theoretic model of false data filtering schemes. From the information-theoretic view, we define the scheme's filtering capacity C F i as the uncertainty-reduction ratio of the target input variable, given the output. This metric not only performs better than existing metrics but also implies that only by optimizing the false negative rate and false positive rate simultaneously, can we promote a scheme's overall performance. Based on the investigation from the modeling efforts, we propose HiFi , a hybrid authentication-based false data filtering scheme. HiFi leverages the benefits of both symmetric and asymmetric cryptography and achieves a high filtering capacity, as well as low computation and communication overhead. Performance analysis demonstrates that our proposed metric is rational and useful, and that HiFi is effective and energy efficient.

Malicious Node Detection in Mobile Wireless Sensor Networks

“…Moreover, if there are many malicious agents, we have a high risk of DoS attacks to wireless sensor networks [1]. Because all messages are delivered to the sink, the neighbor nodes of the sink should receive these messages.…”

“…Of course, the sink cannot know this value. For example, imagine a situation where logical node n u [5] of node n u mounted legitimate MAC attacks several times and logical node n u [1] became an LVN twice. In this case, b u[1] {5} = 2.…”

“…In this case, b u[1] {5} = 2. Suppose further that logical node n u [6] mounted legitimate MAC attacks several times and logical node n u [1] became an LVN three times. In this case, b u [1] {6} = 3 and b u [1] {5, 6} = 5.…”

“…Although there are many works on detecting such false messages [1], [15], [28], [31], [34], they cannot detect malicious agents that create false messages.…”

“…Light-weight Packet Marking (LPM) [19] can be used in situations where there are many malicious agents. However, LPM can only detect a suspicious node 1 Graduate School of Information Systems, The University of ElectroCommunications, Chofu, Tokyo 182-8585, Japan a) sei@is.uec.ac.jp group, which contains a suspicious node n, nodes that had sent messages to node n, and nodes that had received messages from node n. If nodes can move, the number of nodes in a suspicious node group can be very large. Therefore, the effectiveness of LPM goes away in this case.…”