Novel biomarkers and non-invasive diagnostic methods are urgently needed for the screening of gastric cancer to reduce its high mortality. We employed quantitative proteomics approach to develop discriminatory biomarker signatures from human saliva for the detection of gastric cancer. Salivary proteins were analyzed and compared between gastric cancer patients and matched control subjects by using tandem mass tags (TMT) technology. More than 500 proteins were identified with quantification, and 48 of them showed significant difference expression (p < 0.05) between normal controls and gastric cancer patients, including 7 up-regulated proteins and 41 down-regulated proteins. Five proteins were selected for initial verification by ELISA and three were successfully verified, namely cystatin B (CSTB), triosephosphate isomerase (TPI1), and deleted in malignant brain tumors 1 protein (DMBT1). All three proteins could differentiate gastric cancer patients from normal control subjects, dramatically (p < 0.05). The combination of these three biomarkers could reach 85% sensitivity and 80% specificity for the detection of gastric cancer with accuracy of 0.93. This study provides the proof of concept of salivary biomarkers for the non-invasive detection of gastric cancer. It is highly encouraging to turn these biomarkers into an applicable clinical test after large scale validation.