Robust Device Authentication Using Non-Standard Classification Features

This paper investigates the use of novel hardware features derived from the physical and behavioral characteristics of electronic devices to identify such devices uniquely. Importantly, the features examined exhibit non-standard and multimodal distributions which present a significant challenge to model and characterize. Specifically, the potency of four data classification methods is compared whilst employing such characteristics, proposed model Multivariate Gaussian Distribution (MVGD -address multimodality), Logistic Regression (LogR), Linear Discriminant Analysis (LDA), Support Vector Machine (SVM). Performance is measured based on its accuracy, precision, recall and f measure. The experimental results reveal that by addressing multimodal features with proposed model Multivariate Gaussian Distribution classifier, the overall performance is better than the other classifiers.