For the long-distance oil and gas pipelines, the third-party damage (TPD) is a main risk, which is randomness and uncertainty, and difficult to prevent. At present, the safety early warning technologies such as line patrol, fiber-optical vibration and Unmanned Aerial Vehicle (UAV) line patrol are mainly methods adopted by the TPD. But that has many problems such as untimely warning, false alarm and missed report. Combined with the easily obtained mobile phone location data with time-space sequence, a third-party damage behavior identification model for pipelines based on mobile phone location data was established here. Firstly, the mobile phone location information was preprocessed to obtain more accurate third-party activity location information near the target pipeline. The trajectory points were clustered and analyzed based on the density of data, and a stop recognition method based on spatiotemporal clustering was proposed. The key features of the staying point were semantically marked, and the abnormality degree of the staying point is calculated based on the TF IDF rule to accurately extract the abnormal staying point within the pipeline monitoring range. Secondly, extracted and segmented the third-party trajectory, completed the neighborhood search of the trajectory where the stay point located in accordance with trajectory location characteristics, and calculated the behavior difference degree of the neighbor trajectory segment according to multiple trajectory movement characteristics such as velocity, acceleration and rotation angle. Finally, established the model of pipeline TPD decision tree based on the pipeline risk characteristics, and in depth analysis the correlation between various characteristics and the types of third-party sabotage activities. In the end, used the behavior characteristics of the third party to judge the type of TPD. Through the training and testing of the collected historical characteristic data set of the third-party, the accuracy of the identification model established in this paper is 90.9%, and the mobile equipment information in the vicinity of a long-distance pipeline section within 30 days was processed, and the abnormal activities of nearby third parties were monitored according to the 53994 valid data obtained. The results show that the model can accurately identify abnormal behaviors based on the trajectory, it is helpful for timely detection of TPD damage activities such as private excavation, engineering damage and oil theft by drilling, which provides an effective basis for intelligently PTD damage to the pipeline and maintaining the integrity of the pipeline.