Abstract: Due to the efficiency requirements of traffic anomaly detection in electric power data network, an improved anomaly detection algorithm named k-d tree based Lof anomaly detection(KTLAD) based on LOF was proposed. Based on density detection, the algorithm calculated the separating level of each traffic package with nearby ones without pre-set specific abnormal state of traffic. Comparing to the traditional algorithms, the proposed algorithm was more flexible. Simulation results showed that the KTLAD was feasible in traffic anomaly detection in electric power data network and reduced time cost effectively.

Key words: electric power data network, traffic anomaly detection, k-d tree based lof anomaly detection