Parkinson's disease (PD) with cognitive impairment (PDCI) is essentially diagnosed through clinical and neuropsychological examinations. There is a need to identify biomarkers to foresee cognitive decline in them. We performed label-free unbiased nontargeted proteomics (Q-TOF LC/MS−MS) on the CSF of non-neurological control; PDCI; PD; and normal pressure hydrocephalus (NPH) patients, followed by targeted ELISA for validation. Of the 281 proteins identified, 42 were differentially altered in PD, PDCI, and NPH. With a certain overlap, 28 proteins were altered in PDCI and 25 proteins were altered in NPH. Five significantly upregulated proteins in PDCI were fibrinogen, gelsolin, complement factor-H, and apolipoproteins A-I and A-IV, whereas carnosine dipeptidase-1, carboxypeptidase-E, dickkopf-3, and secretogranin-3 precursor proteins were downregulated. Those uniquely altered in NPH were the insulin-like growth factor-binding protein, ceruloplasmin, α-1 antitrypsin, VGF nerve growth factor, and neural cell adhesion molecule L1-like protein. The ELISA-derived protein concentrations correlated with neuropsychological scores of certain cognitive domains. In PDCI, the Wisconsin card sorting percentile correlated negatively with fibrinogen. Intraperitoneal injection of native fibrinogen caused motor deficits in C57BL/6J mice as assessed by the pole test. Thus, a battery of proteins such as fibrinogen-α-chain, CFAH, and APOA-I/APOA-IV alongside neuropsychological assessment could be reliable biomarkers to distinguish PDCI and NPH.