Camouflaged object detection (COD), segmenting hidden objects that are integrated with the surrounding environment, is a valuable challenging task. Existing deep learning methods often struggle to accurately identify disguised objects with complete and refined object structures. For this purpose, we proposes a new edge-feature interaction network (EFINet) for COD. Our method explores and fully utilizes rich edge semantics to force the model to generate prominent object structure features, and then fuses context information to promote precise localization of camouflage object, thereby improving the effectiveness of COD. Extensive experiments on three challenging benchmark datasets have shown that our EFINet outperforms the existing 14 state-of-the-art methods under four widely used evaluation metrics.