Hepatocellular carcinoma (HCC) remains a major health problem for delayed diagnosis, inefficient surveillance and poor prognosis. Recent studies have indicated that non-coding RNAs contribute to the development of new strategies for diagnosis and treatment of HCC. In the present study, we employed 18 pairs of HCC and matched non-tumor tissues for the identification of differentially expressed microRNAs (miRNAs) in HCC, among which 7 paired specimens were selected randomly for microarray detection. Totally, twenty-three miRNAs were screened out to have statistically significant differences with the threshold of P < 0.01 and fold-change ≥ 2.0 or ≤ 0.5 using miRNA microarray. In the validation stage, two miRNAs exhibited higher expression levels in the HCC tissues compared with those in the matched non-tumor tissues, whereas the expression levels of ten miRNAs were lower in the HCC tissues than those in the matched non-tumor tissues. In further analysis, eight miRNAs, including miR-4270, miR-125b-5p, miR-199a-3p, miR-10a-5p, miR-424-5p, miR-195-5p, miR-106b-5p and miR-3651, were retained, when another constraint about the signal intensity of microarray probes was established. Among these miRNAs, our study was the first to show the higher expression level of miR-3651 and the lower expression level of miR-4270 in HCC. The areas under the receiver-operating-characteristic curve values of miR-3651 and miR-4270 were 0.730 and 0.967, respectively, indicating their potential diagnostic values. Our results may help provide the context for expanded interpretations of miRNA studies involved in the progression of liver disease, potentially serving as a diagnostic tool of HCC.