Detection of Replica Node Attack Based on Exponential Moving Average Model in Wireless Sensor Networks

Due to the broadcast nature of wireless communication, wireless sensor networks (WSNs) are susceptible to several attacks. Amongst them, replica attack is one of the predominates as it facilitates the attackers to perform some other attacks. So, it is of immense significance to design a competent security method for WSNs. Introducing a trust method is the primary concern for assisting well-organized use of the available energy in each node in the energy restricted environment. In order to tradeoff between energy usage and attack detection, energy-based prediction approach is deemed to be a suitable one. A statistical method, exponential moving average (EMA) model based replica detection is proposed to detect replica node attack based on energy consumption threshold in WSNs. The difference between actual and predicted energy consumption exceeding the threshold level is considered as malicious. In this paper, future energy drop of a sensor node is forecasted using statistical measure instead of probabilistic method. In EMA model, the transition from higher power consuming state (active state) to lower power consuming states (sleep and sense states) is controlled by a fixed schedule. The accumulated average time of the node was in any state in the past is used to estimate the time duration of a node that spends in that state. Unlike Markov Model, the estimations of energy are made periodically. By this, computational overhead on the microcontroller of the sensor is greatly reduced in EMA approach. The simulation results taken using TRM simulator shows that choosing the threshold value which is neither too large nor too small results in optimum level of detection accuracy and lifetime of the network.