Lung cancer is the leading cause of cancer-related deaths among men and women in the world, accounting for the 25% of cancer mortality. Early diagnosis is an unmet clinical issue. In this work, we focused to develop a novel approach to identify highly sensitive and specific biomarkers by investigating the use of extracellular vesicles (EVs) isolated from the pleural lavage, a proximal fluid in lung cancer patients, as a source of potential biomarkers. We isolated EVs by ultracentrifuge method from 25 control pleural fluids and 21 pleural lavages from lung cancer patients. Analysis of the expression of EV-associated miRNAs was performed using Taqman OpenArray technology through which we could detect 288 out of the 754 miRNAs that were contained in the OpenArray. The differential expression analysis yielded a list of 14 miRNAs that were significantly dysregulated (adj. p-value < 0.05 and logFC lower or higher than 3). Using Machine Learning approach we discovered the lung cancer diagnostic biomarkers; miRNA-1-3p, miRNA-144-5p and miRNA-150-5p were found to be the best by accuracy. Accordance with our finding, these miRNAs have been related to cancer processes in previous studies. This results opens the avenue to the use of EV-associated miRNA of pleural fluids and lavages as an untapped source of biomarkers, and specifically, identifies miRNA-1-3p, miRNA-144-5p and miRNA 150-5p as promising biomarkers of lung cancer diagnosis.