Camouflaged object detection (COD) is a pixel-level task that aims to accurately identify and segment camouflaged objects from scenes exhibiting similar visual properties. Currently, most methods utilize the RGB domain to represent and learn the content of the scene. However, these methods struggle to capture subtle camouflage cues. To address this issue, we propose a novel frequency-aware COD method to promote the collaborative learning between frequency domain and RGB domain information. Specifically, we propose the frequency-aware fusion module (FFM) to integrate multi-scale features from the perspective of frequency information, consisting of a frequency learning module (FLM) and a frequency decomposition unit (FDU). Our FFM explores the distinct roles of different frequency components to emphasize global semantics for capturing camouflaged objects and discard irrelevant information for mitigating interferences. Experiments demonstrate that our FINet exhibits superiority and effectiveness compared to existing representative methods.