A quantitative analysis method for corrosion products based on terahertz spectroscopy is proposed in this paper. Mixture samples consisting of three major corrosion products (magnetite, hematite, and goethite) were prepared in 51 different concentrations. The refractive index spectra measured by terahertz time-domain spectroscopy were projected to the 2D score diagram by performing principal component analysis. The Euclidean distances between the mixtures and pure analyte on the diagram were used to build a concentration prediction model. The results indicate that the established model can precisely predict the concentration of magnetite, which is essential for a stability evaluation of the corrosion system.