Autophagy plays an important role in the occurrence and development of hepatocellular carcinoma (HCC). We aimed to develop an autophagy-related genes signature predicting the prognosis of HCC and to depict a competing endogenous RNA (ceRNA) network. Methods: Differentially expressed autophagy-related genes (DE-ATGs), miRNAs and lncRNAs and clinical data of HCC patients were extracted from TCGA. The GO and KEGG analysis were performed to investigate the gene function. Univariate and multivariate Cox regression analysis were used to identify a prognostic signature with the DE-ATGs. And a nomogram, adapted to the clinical characteristics, was established. Then, we established a ceRNA network related to autophagy genes. Results: We screened out 27 differentially expressed genes which were enriched in GO and KEGG pathways related to autophagy and cancers. In univariate and multivariate Cox regression analysis, BIRC5, HSPB8, and SQSTM1 were screened out to establish a prognostic risk score model (AUC=0.749, p<0.01). Kaplan-Meier survival analysis showed that the overall survival of high-risk patients was significantly worse. Furthermore, the signature was validated in the other two independent databases. The nomogram, including the autophagy-related risk signature, gender, stage and TNM, was constructed and validated (C-index=0.736). Finally, the ceRNA network was established based on DE-ATGs, differentially expressed miRNAs and lncRNAs. Conclusion: We constructed a reliable prognostic model of HCC with autophagy-related genes and depicted a ceRNA network of DE-ATGs in HCC which provides a basis for the study of post-transcriptional modification and regulation of autophagy-related genes in HCC.