Document authentication using graphical codes

Abstract

This paper proposes to investigate the impact of the channel model for authentication systems based on codes that are corrupted by a physically unclonable noise such as the one emitted by a printing process. The core of such a system is the comparison for the receiver between an original binary code, an original corrupted code and a copy of the original code. We analyze two strategies, depending on whether or not the receiver use a binary version of its observation to perform its authentication test. By deriving the optimal test within a Neyman-Pearson setup, a theoretical analysis shows that a thresholding of the code induces a loss of performance. This study also highlights the fact that the probability of the type I and type II errors can be better approximated, by several orders of magnitude, computing Chernoff bounds instead of the Gaussian approximation. Finally we evaluate the impact of an uncertainty for the receiver on the opponent channel and show that the authentication is still possible whenever the receiver can observe forged codes and uses them to estimate the parameters of the model.

Sensitivity of different correlation measures to print-and-scan process

“…The most used authentication test for CDP is the comparison of original CDP with binarized version of P&S CDP [12]. But thanks to research interest on this topic, several original copydetection measures were suggested [3].…”

Image model and printed document authentication: A theoretical analysis

“…In previous works [11,6] we modeled the mentioned authentication system as a hypothesis testing problem and we derived tight bounds on its performance. The authentication system works as follow: a binary authentication image is constructed from a randomly chosen binary sequence x N in {0, 1} N by the legitimate source.…”

“…2), are respectively modeled by probability distributions conditioned to the original data P Y /X,Θ and P Z/X,Θ , Θ andΘ are sets of parameters, taken in Ξ, specifying the devices in each case. As pointed in [11], the receiver observes one of the two possible images y N or z N and have to decide whether it comes from the legitimate source or not (see (5) in Fig. 2), supposed that the models P Y /X,Θ and P Z/X,Θ are known.…”

“…In [11,6], we indicate several methods to approximate β while keeping α fixed and show that an Asymptotic Expression (AE) provides a reliable approximation for β. We present a brief summary for the AE method.…”