Multi-class Problem Decomposition Using Genetic Programming

Behavioural detectors for intrusion detection require training in order to correctly characterize the operation of a service – protocol combination. Implicit in this is the assumption that the learning algorithm will scale to large datasets and provide simple solutions. This work will address both requirements under a Genetic Programming context through the use of a combined multi-objective, host-parasite model. It has already been demonstrated that both schemes are appropriate independently. This project will integrate the two schemes to provide a single, holistic solution and benchmark under representative datasets.

Faculty Supervisor:

Dr. Malcolm Heywood

Student:

Andrew McIntyre

Partner:

Telecom Applications Research Alliance (TARA)

Discipline:

Computer science

Sector:

Information and communications technologies

University:

Dalhousie University