Graph Structure Based Anomaly Behavior Detection

DOI

10.2991/iccia-17.2017.90How to use a DOI?

Keywords

Anomaly Detection, Graph Mining, Unsupervised Learning, Social Graph.

Abstract

The analysis of malicious user behavior patterns in social networks has important implications for detecting malicious pages, fraudsters, and financial frauds.Traditional anomaly detection technology general based on classification algorithm using content feature and user behavior feature, but these type of methods are often with low efficiency, data acquisition difficulty and ignoring the network topology information.This paper puts forward a network graph structure based, unsupervised anomaly detection algorithm GBKD-Forest, we extracted three types of structure characteristics, within the Bagging method random sampling features to establish KD-Tree Forest, to isolate the abnormal samples.Evaluation through the experiment, the proposed algorithm in terms of accuracy and AUC is superior to other graph based anomaly detection algorithm and classical classification algorithm, at the same time, the time complexity of this algorithm has a linear relation with the number of nodes, low space complexity is suitable for large-scale network anomaly detection datasets.

Copyright

© 2017, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Kai Wang
AU  - Danwei Chen
PY  - 2016/07
DA  - 2016/07
TI  - Graph Structure Based Anomaly Behavior Detection
BT  - Proceedings of the 2nd International Conference on Computer Engineering, Information Science & Application Technology (ICCIA 2017)
PB  - Atlantis Press
SP  - 531
EP  - 538
SN  - 2352-538X
UR  - https://doi.org/10.2991/iccia-17.2017.90
DO  - 10.2991/iccia-17.2017.90
ID  - Wang2016/07
ER  -