Abstract: At present, research and application of medical big data are more and more extensive. But inevitably, medical big data is of some deception, and in many scenarios, it can result in wrong conclusions and influence. In this paper, firstly we analyze the causes of medical big data deception from the data deception per se and pitfalls of machine learning. Then, we introduce how to avoid data pitfalls in statistics and analyze the strategies to tackle attacks on models. The importance and methods achieving model interpretability in the medical area are also mentioned.