Near infrared spectroscopy (NIRS) coupled with chemometric methods have been widely used in the detection of food fraud. The objective of this study was to use the NIRS and Principal Components Analysis (PCA) and k-Nearest Neighbors (k-NN) to detect bovine serum (BS) in powdered goat milk. Powdered goat milk samples were purchased in the local market of the Campinas (SP), adulterated in quadruplicate with BS at the concentrations of 0 (control), 5, 10, 15, 20, 25, 30, 40, 50 and 60%, totalizing 18 control samples and 54 adulterated samples. Reflectance spectra were obtained in the range of 10.000 to 4.000 cm-1 (32 scans with 4 cm-1 of increment) in a FT-NIR spectrometer (Perkin Elmer-Waltham, USA, model Spectrum 100N) equipped with NIRA reflectance accessory. The chemometric analysis were developed using the PLS-Toolbox (version 7.3.1) for Matlab (R2013a). The PCA and k-NN (3 nearest neighbors) were obtained after the spectra were preprocessed using the 1st derivative of Savitzky-Golay algorithm (second order polynomial and smoothing window of 15 points) and mean centered. A second k-NN analysis was performed by variables selection (6.000 to 4.000 cm-1). In the PCA, two components explain 99.80% of the variance, and the second component was responsible for the description of the variation on the BS content, distinguishing between authentic and adulterated samples. According to the first k-NN, no adulterated sample was individually classified, while the second k-NN discerned the samples belonging to each adulterated fraction authentic. The NIRS correlated to chemometric methods distinguish with 95% of confidence level the detection of adulteration of powdered goat milk by bovine serum, a cheap product that alters milk intrinsic quality. Thus, the use of NIRS technique stands out from physical-chemical (conventional methods) applied on the quality control of powdered goat milk, as well as being practical and green technique.