With the development of deep learning technologies, video face tampering technologies, represented by Deepfake, can easily generate fake face images of videos by modifying the original video with only a small amount of face images. Therefore, the detection of forged facial videos has become critical to internet content regulation. In this paper, a deep forgery face video detection method with fusion of frequency domain features and spatial domain features (FFS) is proposed to address the problem of deepfake face video detection. At first, the proposed method extracts the wavelet features of images with two-dimensional discrete wavelet transform, then extracts the multidimensional wavelet feature vectors of images according to n-level wavelet decomposition. The frequency domain features of the image extracted by the discrete Fourier transform are also cascaded and fused with the wavelet features. Besides, the proposed method can better address the overfitting problem of detection methods in practical internet application scenarios by establishing a shared updatable strategy. Finally, in order to improve the generalization ability of the detection model and address the problem that the model is vulnerable to malicious attacks under the above shared update strategy, blockchain technology is employed to implement an incentive mechanism. It can motivate participants to provide real and health data, and then achieve the establishment and maintenance of a good shared and renewable environment. Moreover, we use DeepfakeDetection and Celeb-DF datasets to conduct the experiments. Samples with different percentages of high-quality images are selected to simulate the complex environment of image quality in the Internet. Experimental results show that the proposed method can improve the performance of the face forgery detection model effectively.