Distributed Intrusion Detection of Byzantine Attacks in Wireless Networks with Random Linear Network Coding (1303.2553v1)

Abstract: Network coding is an elegant technique where, instead of simply relaying the packets of information they receive, the nodes of a network are allowed to combine \emph{several} packets together for transmission and this technique can be used to achieve the maximum possible information flow in a network and save the needed number of packet transmissions. Moreover, in an energy-constraint wireless network such as Wireless Sensor Network (a typical type of wireless ad hoc network), applying network coding to reduce the number of wireless transmissions can also prolong the life time of sensor nodes. Although applying network coding in a wireless sensor network is obviously beneficial, due to the operation that one transmitting information is actually combination of multiple other information, it is possible that an error propagation may occur in the network. This special characteristic also exposes network coding system to a wide range of error attacks, especially Byzantine attacks. When some adversary nodes generate error data in the network with network coding, those erroneous information will be mixed at intermeidate nodes and thus corrupt all the information reaching a destination. Recent research efforts have shown that network coding can be combined with classical error control codes and cryptography for secure communication or misbehavior detection. Nevertheless, when it comes to Byzantine attacks, these results have limited effect. In fact, unless we find out those adversary nodes and isolate them, network coding may perform much worse than pure routing in the presence of malicious nodes. In this paper, a distributed hierarchical algorithm based on random linear network coding is developed to detect, locate and isolate malicious nodes.