Counterfeit medicines pose a huge threat to public health worldwide. High amounts of counterfeit pharmaceuticals enter the European market and therefore detection of these products is essential. Attenuated Total Reflection Fourier-Transform infrared spectroscopy (ATR-FTIR) might be useful for the screening of counterfeit medicines since it is easy to use and little sample preparation is required. Furthermore, this approach might be helpful to customs to obtain a first evaluation of suspected samples. This study proposes a combination of ATR-FTIR and chemometrics to discriminate and classify counterfeit medicines. A sample set, containing 209 samples in total, was analyzed using ATR-FTIR and the obtained spectra were used as fingerprints in the chemometric data-analysis which included Principal Component Analysis (PCA), k-Nearest Neighbours (k-NN), Classification and Regression Trees (CART) and Soft Independent Modelling of Class Analogy (SIMCA). First it was verified whether the mentioned techniques are capable to distinguish samples containing different active pharmaceutical ingredients (APIs). PCA showed a clear tendency of discrimination based on the API present; k-NN, CART and SIMCA were capable to create suitable prediction models based on the presence of different APIs. However k-NN performs the least while SIMCA performs the best. Secondly, it was tested whether these three models could be expanded to discriminate between genuine and counterfeit samples as well. k-NN was not able to make the desired discrimination and therefore it was not useful. CART performed better but also this model was less suited. SIMCA, on the other hand, resulted in a model with a 100% correct discrimination between genuine and counterfeit drugs. This study shows that chemometric analysis of ATR-FTIR fingerprints is a valuable tool to discriminate genuine from counterfeit samples and to classify counterfeit medicines.