Mobile network traffic anomaly detection method and system based on feature dimension reduction

Abstract

The invention discloses a mobile network traffic anomaly detection method and system based on feature dimensionality reduction, and the method comprises the steps: dividing a city region into M * N grid regions according to the distribution of city base stations, and aggregating the cellular traffic value of each grid region by using pandas to obtain a cellular traffic total value with an hour as a unit; dividing a detection time period into K time slots to form a time sequence vector, and taking the time sequence vector as an original cellular flow vector xj; extracting low-dimensional traffic features cj from the original cellular traffic vectors xj of all grid areas by using an LSTM auto-encoder; determining suspicious abnormal low-dimensional traffic features in the low-dimensional traffic features corresponding to all the grid regions; using the K-means clustering for carrying out anomaly confirmation on the suspicious and abnormal low-dimensional traffic features, completing mobile network traffic anomaly detection based on feature dimension reduction. The method and the system can realize anomaly detection of the mobile network traffic and have the advantages of being large in number of processing areas and short in data processing time.

Description

technical field [0001] The invention relates to an anomaly detection method and system, in particular to a method and system for abnormal detection of mobile network traffic based on feature dimensionality reduction. Background technique [0002] Anomaly detection is one of the important tasks in wireless network data analysis and management. Anomalies in wireless networks refer to patterns that deviate from normal / expected behavior, these patterns can be network congestion in wireless networks, DDoS amplification attacks, port / service scans, and spurious traffic caused by network failures. Anomaly detection is extremely valuable to service providers. Detection of user traffic anomalies that have occurred can provide network operators with more information about hotspot areas, examine the rationality of existing resource allocation schemes, guide the dynamic allocation and adjustment of network resources, and propose intelligent fault diagnosis solutions. [0003] In the e...

AI Technical Summary

Problems solved by technology

However, there are still some problems with this method of anomaly detection. Directly using the clustering algorithm to detect anomalies can detect anomalies in high-traffic areas, but ignore anomalies in low-traffic areas.

In addition, there is an anomaly detection method based on K-means clustering based on traffic patterns. In large-scale long-term sequence detection problems, there are defects such as limited number of processing areas and limited processing time.

Images