In this work, rapid and non-destructive methodology is proposed for screening of biodiesel/diesel blends with respect to the base oil, using near infrared spectroscopy and multivariate data analysis, since for both pure biodiesel and blends, the biodiesel/diesel are targets for tampering. Blends of diesel with cotton, sunflower and soybean oils were employed in this study. Two approaches were evaluated in the building of the classification model, using full-spectrum Soft Independent Modelling of Class Analogy (SIMCA), and Principal Component Analysis-Linear Discriminant Analysis (PCA-LDA). The other approaches were the use of variable selection employing Genetic Algorithm (GA), Successive Projection Algorithm (SPA) and Stepwise all coupled with the LDA model. The results showed which preprocessed NIR spectra and chemometrics are a viable alternative the conventional methods that involve the consumption of large volumes of reagents. Multivariate data analysis methods using selected variables showed a better performance than the methods using a full spectrum.