Malicious Node Detection using Route Prediction based on HMM

Abstract

Driving route prediction methods based on Hidden Markov Model accurately predicts a vehicle’s entire route throughout a trip. Trip history of driver alone cannot be used for predicting the route. Routine history of routes can be modelled and learned for predicting purposes. Driver behavior, another factor of route prediction can be considered as another factor of route prediction. Route recommendation mechanism helps to identify the probability of mobility of vehicles over time. This method can be extended to identify malicious nodes within network traffic. First we define a road network model, the driving routes in a hexagonal coordinate system, build HMM models to predict the movement using a method of training set based on K-means++ technique. The route predicted is taken as input and transmitted along with network data using encrypted headers. A method to identify malicious nodes in VANETs using HMM of prediction about routes helps to identify malicious message from a compromised node. One method of identifying suspicious message is the signal strength which is incompatible with its originator's geographical position. We provide encrypted headers in protocols for detecting suspicious transmissions. Identified malicious node information is disseminated in the network. Evaluation of the detection rate and the efficiency of solution is analyzed using cryptographic methods based on cloud computing. This helps to identify the malicious nodes in the network traffic.