Presentation 2013-11-15
A Robust Principal Component Analysis for Traffic Anomaly Detection
Takahiro MATSUDA, Tatsuya MORITA, Takanori KUDO, Tetsuya TAKINE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this article, we consider anomaly detection schemes for traffic traces measured in multiple links of a network. Principal Component Analysis (PCA) is a promising technique to detect anomalies in network traffic, and it can detect anomalies by projecting the measured traffic data onto a normal and anomalous subspaces. In the classical PCA-based anomaly detection scheme, however, outliers, anomalies with excessively large traffic volume, may degrade its performance, which is referred to as the subspace contamination problem. In this article, we propose a new traffic anomaly detection scheme based on the MOD (Minimum Covariance Detection) scheme, a robust PCA mechanism. The proposed scheme utilizes the periodicity of network traffic, and before constructing subspaces, outliers are removed by using the measured traffic in the preceding period. Although robust PCA schemes are effecitive against the subspace contamination problem themselves, the proposed scheme can reduce the computational cost by using the traffic periodicity. We evaluate the performance of the proposed scheme with numerical experiments.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Traffic anomaly detection / Principal Component Analysis / Periodicity of network traffic
Paper # NS2013-128
Date of Issue

Conference Information
Committee NS
Conference Date 2013/11/7(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Network Systems(NS)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A Robust Principal Component Analysis for Traffic Anomaly Detection
Sub Title (in English)
Keyword(1) Traffic anomaly detection
Keyword(2) Principal Component Analysis
Keyword(3) Periodicity of network traffic
1st Author's Name Takahiro MATSUDA
1st Author's Affiliation Graduate School of Engineering, Osaka University()
2nd Author's Name Tatsuya MORITA
2nd Author's Affiliation School of Engineering, Osaka University
3rd Author's Name Takanori KUDO
3rd Author's Affiliation Graduate School of Engineering, Osaka University
4th Author's Name Tetsuya TAKINE
4th Author's Affiliation Graduate School of Engineering, Osaka University
Date 2013-11-15
Paper # NS2013-128
Volume (vol) vol.113
Number (no) 292
Page pp.pp.-
#Pages 6
Date of Issue