Network Protocol Analytics For End User Context Identification

End user context identification on the server side can serve an important role in growing internet communication. Our target is to identify the mobility context of the end user on the server or sender side connected in a wireless network. Lack of previous work and unavailability of specific tools makes this work challenging. Naive approach includes, the trend of variation of throughput with mobility and Pearson correlation coefficient comparison of traced transport layer parameters. However both approach does not mark any strict signature for mobility of the end user, leads to lack of accuracy in estimation of mobility of the end user. Our approach is to model the traced parameters (parameters are traced using NS2) of the transport layer graphically using Bayesian Networks, then learn the joint probability distribution between the parameters using BNT toolbox of Matlab. Finally using graphical model, learned parameters from BNT toolbox and Pearson correlation coefficient we estimate the mobility context of end user with minimum error. Graphically modelling tracks the unknown relation between the parameters which is not carried out by Pearson correlation coefficient and Bayesian toolbox calculate the joint probability distribution between the parameters. Comparing the raw data with graphical model and Pearson coefficient gives the estimate of mobility of end user. Major advantage of this solution is its robustness because graphical model tracks the unknown relation between the parameters and Pearson coefficient tracks the correlation between them which has a fairly good variation with mobility.