This paper aims to propose an anomaly detection technic for a vast amount of flight data. The proposed method is based on the fact that the numberof normal flight patterns are enormous compared tothat ofabnormal flight patterns. The proposed method consists of data transformation, data sampling, and feature generation to make anappropriate data format for the analysis.The representation of eachflight parameters isset to be a mean value in a sampling interval. Then,representation values are gathered into a vector to form a feature point of each flightdata. An unsupervised machine learning approach called DBSCAN (clustering method) with the feature vector is used for detecting potential anomalies in the flight data. Finally, the proposed algorithm is verified by utilizing NASA open database.