Elevated peripheral levels of different cytokines and chemokines in subjects with Alzheimer's disease (AD), as compared with healthy controls (HC), have emphasized the role of inflammation in such a disease. Considering the cross-talking between the central nervous system and the periphery, the inflammatory analytes may provide utility as biomarkers to identify AD at earlier stages. Using an advanced statistical approach, we can discriminate the interactive network of cytokines/chemokines and propose a useful tool to follow the progression and evolution of AD, also in light of sex differences. A cohort of 289 old-age subjects was screened for cytokine and chemokine profiling, measured in plasma, after a thorough clinical and neuropsychological evaluation. A custom algorithm based on Fisher linear discriminant analysis was applied to ascertain a classification signature able to discriminate HC from mild cognitive impairment (MCI) and AD. We observed that a joint expression of three proteins (a "signature" composed by IFN-α2, IL-1α, TNFα) can discriminate HC from AD with an accuracy of 65.24%. Using this signature on MCI samples, 84.93% of them were classified as "non-HC". Stratifying MCI samples by sex, we observed that 87.23% of women were classified as "non-HC", and only 57.69% of males. Indeed, in a scatter plot of IFN-α2 and IL-1α, the HC group was better separated from MCI and AD in women as compared with men. These findings suggest that AD is accompanied by a peripheral inflammatory response that can already be present in MCI subjects, thus providing a mean for detecting this at-risk status and allow an anticipated intervention.