Bilateral PPG signals have been used for comparative study of two groups of healthy (free from any cardiovascular risk factors) and diabetic (as cardiovascular disease risk group) subjects in the age-matched range 40-50 years. The peripheral blood pulsations were recorded simultaneously from right and left index fingers for 90 s. Pulses have been modeled with the ARX440 model in the interval of 300 sample points with 100 sample points overlap between segments. Model parameters of three segments based on the highest fitness (higher than 80%) of modeled segments were retained for each subject. Subsequently, principal component analysis (PCA) was applied to the parameters of retained segments to eliminate the existing correlation among parameters and provide uncorrelated variables. The first principal component (contains 78.2% variance of data) was significantly greater in diabetic than in control groups (P < 0.0001, 0.74 +/- 2.01 versus -0.53 +/- 1.66). In addition the seventh principal component, which contains 0.02% of the data variance, was significantly lower in diabetic than in control groups (P < 0.05, -0.007 +/- 0.03 versus 0.005 +/- 0.03). Finally, linear discrimination analysis (LDA) was used to classify the subjects. The classification was done using the robust leaving-one-subject-out method. LDA could classify the subjects with 71.7% sensitivity and 70.2% specificity while the male subjects resulted in a highly acceptable result for the sensitivity (81%). The present study showed that PPG signals can be used for vascular function assessment and may find further application for detection of vascular changes before onset of clinical diseases.