Early detection of noise-induced hearing loss (NIHL) in patients with long-term noise exposure is vital for improving public health and reducing social burden. However, at present, the diagnosis of NIHL mainly depends on audiometric testing, and the primary test is pure-tone audiometry. Moreover, testing requires professional operators and complex equipment; thus, NIHL is often diagnosed at a later disease stage. Using a liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based proteomic approach, we identified 9 differentially expressed proteins (DEPs), namely, 6 upregulated serum proteins and 3 downregulated serum proteins, in samples from 50 ground crew personnel working at an air force station. Then, according to the results, we predicted that caldesmon (CALD1), myocilin (MYOC), zyxin (ZYX), creatine kinase M-type (CKM), insulin-like growth factor-binding protein 2 (IGFBP2), complement factor H-related protein 4 (CFHR4), prenylcysteine oxidase 1 (PCYOX1), heat shock cognate 71 kDa protein (HSPA8), and immunoglobulin lambda variable 3-21 (IGLV3-21) were associated with NIHL. We selected these DEPs as variables to perform logistic regression. Finally, a logistic regression model was constructed based on IGFBP2, ZYX, CKM, and CFHR4. The area under the curve was 0.894 (95% CI = 0.812 to 0.977). These findings suggested that IGFBP2, ZYX, CKM, and CFHR4 in serum are differentially expressed in NIHL patients and have the potential to be biomarkers for predicting the risk for NIHL. Further experiments in mice showed that ZYX and IGFBP2 in the cochlear were increased after noise exposure. ZYX and IGFBP2 may be involved in the occurrence and development of NIHL.