Enhanced THz tags authentication using multivariate statistical analysis

In this paper, we report on the unitary authentication of identically realized diffraction grating-based tags structures in the THz domain by using multivariate statistical analysis. We proceed to a dimension reduction with Principal Component Analysis (PCA) as a preprocessing step before a Gaussian classification and we evaluate the error rates. We then classify the tags using a Linear Discriminant Analysis (LDA). We demonstrate that PCA gives average error rates lower than 0.5% whereas LDA is able to classify the tags with error rates lower than 6.10-5considering its 3 first axes.