Small extracellular vesicles (sEVs) are reliable biomarkers for early cancer detection; however, conventional detection methods such as immune-based assays and microRNA analyses are not very sensitive and require sample pretreatments and long analysis time. Here, we developed a molecular imprinting-based dynamic molding approach to fabricate antibody-conjugated signaling nanocavities capable of size recognition. This enabled the establishment of an easy-to-use, rapid, sensitive, pretreatment-free, and noninvasive sEV detection platform for efficient sEV detection-based cancer diagnosis. An apparent dissociation constant was estimated to be 2.4 × 10-16 M, which was ∼1000 times higher than that of commercial immunoassays (analysis time, 5 min/sample). We successfully used tears for the first time to detect cancer-related intact sEVs, clearly differentiating between healthy donors and breast cancer patients, as well as between samples collected before and after total mastectomy. Our nanoprocessing strategy can be easily repurposed for the specific detection of other types of cancer by changing the conjugated antibodies, thereby facilitating the establishment of liquid biopsy for early cancer diagnosis.