The integration of expert knowledge in news classification has traditionally been a difficult problem. This paper focus on designing an expert knowledge enhanced fake news detection method. First, we classify expert knowledge into two categories: common sense knowledge and authoritative event information. For common sense knowledge, this paper designed a knowledge-reinforced transformer model that utilizes a knowledge encoding tree to embed entity knowledge into news texts and a masking mechanism to address the redundant knowledge problem brought on by over embedding. For authoritative event information, this paper uses latent semantic analysis methods to select factual information that may be used to classify fake information, and then uses the attention mechanism to fuse the selected factual information features with news text features. We verify our method on public datasets “Weibo21” and “Covid_news”. The results show a 1.3% to 18% improvement in classification accuracy, verifying the effectiveness of the methods proposed in this paper.