A new analytical methodology using HS-SPME/GC-MS was optimized in order to attain maximum sensitivity, using multivariate strategies. The proposed method was employed to evaluate the VOC profile exhaled from canine hair samples collected from 8 healthy dogs and from 16 dogs infected by Leishmania infantum. 274 VOCs were detected, which could be identified as aldehydes, ketones and hydrocarbons. After application of the Soft Independent Modeling of Class Analogy (SIMCA) and Principal Component Analysis (PCA) healthy and infected dogs, with similar VOCs profiles, could be separately grouped, based on compounds such as 2-hexanone, benzaldehyde, and 2,4-nonadienal. The proposed method is non-invasive, painless, readily accepted by dog owners and could be useful to identify several biomarkers with applications in the diagnosis of diseases.