APF-PF: The new real-time anomaly detection model of massive log flow
Qing Wu, Guoping Wu · 2016
Computer technology is developing rapidly, the volume of logs in large internet companies witnesses an explosive growth. It is significant for the improvement of customer satisfaction and system stability to analyze and detect logs and timely find customer behaviors and the anomaly of system state. Aiming at massive log flow, TLSCA algorithm is first put forward based on sequence compression algorithm to achieve the lossless compression of scene; Secondly, fractal analysis technology is introduced so as to put forward the log stream piecewise and fractal model; then, based on the piecewise and fractal model, brand new anomaly detection algorithm for parameter-free data stream is put forward to solve the onerous parameter setting. Finally, the high efficiency of APF-PF model for anomaly detection of mass data stream is verified through experiment.