ENSEMBLE DESIGN OF MASQUERADER DETECTION SYSTEMS FOR INFORMATION SECURITY

Masqueraders are a category of intruders who impersonate other people on a computer system and use this entry point to use the information stored in the systems or throw other attacks into the network. This paper focuses on Ensemble Design of a Masquerader Detection System using Decision trees and Support Vector Machines for classification with two kernel functions linear and linear BSpline. The key idea is to find out specific patterns of command sequence that tells about user behaviour on a system, and use them to build classifiers that can perfectly recognize anomalous and normal behaviour. Real time truncated command line data set collected from a debian Linux server is used for performance comparison of the developed classifiers with the standard truncated command line data set of Schonlau[4]. The results show that Ensemble Design of Masquerader Detection Systems is much faster than individual Decision trees or Support Vector Machines.