An ANN Based Approach for Wireless Device Fingerprinting

Abstract

Wireless Device fingerprinting is a technique used to uniquely identify a wireless device. Device fingerprinting plays an important role in finding counterfeit devices in a network. In this thesis, a technique of wireless devices fingerprinting has been presented using Artificial Neural Networks (ANNs). We have taken Transmission Time (TT) and frame Inter Arrival Time (IAT) as the features for device fingerprinting. The proposed model has been evaluated using GTID and Sigcomm2008 datasets. The performance analysis of two different training function i.e. Scale Conjugate Gradient (SCG) and Bayesian Regularization (BR) has been done. The result shows that BR performs better than SCG in classifying unique devices in both the datasets. The proposed method gives better accuracy and identifies more number of devices than its counterparts.