Neural Network Based Approach for Malicious Node Detection in Wireless Sensor Networks
Faculty of Computing, Health and Science
School of Engineering
Even if their resources in terms of energy, memory, computational power and bandwidth are strictly limited, sensor networks have proved their huge viability in the real world, being just a matter of time until this kind of networks will be standardized and used broadly in the field. One of the important problems that are related to the use of wireless sensor networks in harsh environments is the gap in their security. This paper provides a solution to discover malicious nodes in wireless sensor networks using an on-line neural network predictor based on past and present values obtained from neighboring nodes. This solution can be also a way to discover the malfunctioning nodes that were not a subject of an attack. Being localized on the base station level, our algorithm is suitable even for large-scale sensor networks.