Sifting through network data to cull activity patterns with HEAPs

Esam Sharafuddin, Yu Jin, Nan Jiang, Zhi-Li Zhang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Scopus citations

Abstract

Today's large campus and enterprise networks are characterized by their complexity, i.e. containing thousands of hosts, and diversity, i.e. with various applications and usage patterns. To effectively manage and secure such networks, network operators and system administrators are faced with the challenge of characterizing, profiling and tracking activity patterns passing through their networks. Because of the large number of IP addresses and the prevalence of dynamic IP addresses, profiling and tracking individual hosts may not be effective nor scalable. In this paper, we develop a hierarchical extraction of activity patterns (HEAPs), which is a method for characterizing and profiling activity patterns within subnets. By representing activities within a subnet in a host-port association matrix (HPAM) and applying pLSA, we obtain co-clusters that capture the significant and dominant activity patterns of the subnet. Using these co-clusters, we utilize hierarchical clustering to cluster activity patterns to assist network operators and security analysts gain a "big-picture" view of the network activity-patterns. We also develop a novel method to track and quantify changes in activity patterns within subnets over time and demonstrate how to utilize this method to identify major changes and anomalies within the network.

Original languageEnglish (US)
Title of host publicationICDCS 2010 - 2010 International Conference on Distributed Computing Systems
Pages685-696
Number of pages12
DOIs
StatePublished - Aug 27 2010
Event30th IEEE International Conference on Distributed Computing Systems, ICDCS 2010 - Genova, Italy
Duration: Jun 21 2010Jun 25 2010

Publication series

NameProceedings - International Conference on Distributed Computing Systems

Other

Other30th IEEE International Conference on Distributed Computing Systems, ICDCS 2010
CountryItaly
CityGenova
Period6/21/106/25/10

Fingerprint Dive into the research topics of 'Sifting through network data to cull activity patterns with HEAPs'. Together they form a unique fingerprint.

Cite this