On the feasibility of classification-based product package authentication

Abstract

Depending on the product category the authenticity of a consumer good concerns economic, social and/or environmental issues. Counterfeited drugs are a threat to patient safety and cause significant economic losses. Different from physical-marking based approaches this work investigates authentication of drugs based on intrinsic texture features of the packaging material. Therefore, it is assumed that the packaging material of a certain drug shows constant but discriminative textural features which enable authentication, i.e. to prove if the packaging material is genuine or not. This objective requires considering a binary classification problem with an open set of negative classes, i.e. unknown and unseen counterfeits. In order to investigate the feasibility a novel drug packaging texture databases was acquired. The experimental evaluation of two basic requirements in texture classification serves as an evidence on the basic feasibility.