Small extracellular vesicles (sEVs) have emerged as noninvasive biomarkers in liquid biopsy due to their significant function in pathology and physiology. However, the phenotypic heterogeneity of sEVs presents a significant challenge to their study and has significant implications for their applications in liquid biopsies. In this study, anodic aluminum oxide films with different pore sizes (AAO nanoarray) were introduced to enable size-based isolation and downstream proteomics profiling of sEV subpopulations. The adjustable pore size and abundant Al 3+ on the framework of AAOs allowed size-dependent isolation of sEV subpopulations through nanoconfined effects and Lewis acid−base interaction between AAOs and sEVs. Benefiting from the strong concerted effect, the simple AAO nanoarray enabled specific isolation of three sEV subpopulations, termed "50", "90", and "150 nm" groups, from 10 μL of complex biological samples within 10 min with high capture efficiencies and purities. Moreover, the nanopores of AAOs also acted as nanoreactors for comprehensive proteomic profiling of the captured sEV subpopulations to reveal their heterogeneity. The AAO nanoarray was first investigated on sEVs from a cell culture medium, where sEV subpopulations could be clearly distinguished, and three traditional sEV-specific proteins (CD81, CD9, and FLOT1) could be identified by proteomic analysis. A total of 3946, 3951, and 3940 proteins were identified from 50, 90, and 150 nm sEV subpopulations, respectively, which is almost twice the number compared to those obtained from the conventional approach. The concept was further applied to complex real-case sample analysis from prostate cancer patients. Machine learning and gene ontology (GO) information analysis of the identified proteins indicate that different-sized sEV subpopulations contain unique protein cargos and have distinct cellular components and molecular functions. Further receiver operating characteristic curve (ROC) analysis of the top five differential proteins from the three sEV subpopulations demonstrated the high accuracy of the proposed approach toward prostate cancer diagnosis (AUC > 0.99). More importantly, several proteins involved in focal adhesion and antigen processing and presentation pathways were found to be upregulated in prostate cancer patients, which may serve as potential biomarkers of prostate cancer. These results suggest that the sEV subpopulation-based AAO nanoarray is of great value in facilitating the early diagnosis and prognosis of cancer and opens a new avenue for sEVs in liquid biopsy.