Several European local dairy products are currently labeled with the Protected Designation of Origin (PDO) trademark based on their production area. One of the conditions to obtain this designation is that the raw material (I.e. the milk) needs to be produced in the region delimited within the PDO’s specifications. The aim of this study was to discriminate the milk produced in the PDO area ‘Ardenne’ from the milk produced in other parts (non-PDO) of the Walloon Region using mid infrared (MIR) spectroscopy as analytical tool. More than 1,800,000 standardized MIR spectra from the Walloon bulk milk spectral database generated during the process of milk quality analysis for payment between January 2012 and December 2015 were used. In order to simplify the discrimination analysis model on the whole dataset which required taking into account of several effect (e.g. seasons, months and years), 1,450 sub-datasets were generated using a 5 days’ time window approach throughout the 4 years. In each sub-dataset, only one MIR spectrum per farm was selected randomly, the other spectra were discarded of the sub-dataset. Afterwards, a first derivative pretreatment of the spectrum was applied using Savitzky-Golay derivative with a window size of 5 wavelengths. Then, partial least square discrimination analysis was performed on PDO and non-PDO milk samples of each calibration sets. In average, a correct classification rate of 79% (±2%) was obtained. The average sensitivity and specificity were 77% (±2%) and 80% (±2%), respectively. Given the existing variability between farms (due in particular to the breeds and the feed), the obtained results were promising and highlighted the usefulness of MIR spectroscopy to assess the origin of milk produced in a specific PDO area.