ADS-B is an important part of next-generation air traffic control systems.Since ADS-B adopts one-way broadcastthe credibility and reliability of ADS-B data lack effective verification means.Statistical analysis of ADS-B data of different TYPE values is conducted.By calculating the curvature of air tracks formed by air position data and building a confidence data set under the constraints of time windowthe anomaly detection and classification of ADS-B air position data are performedand the relationship between abnormal data and TYPE value is analyzed.The experimental results show that the proposed method can rapidly and accurately identify abnormal air position data.The statistical results show that there is about 1.59 abnormal data on average in every 100 ADS-B air position data.