Mark Crosbie - PhD Candidate, Purdue University
Sep 23, 1994
Genetic Algorithms provide a new approach to problem solving based on
the idea of evolving ever improving solutions. They are based on
observations from natural evolution where organisms who best adapt to
an environment will be deemed the "best solution" for that
environment. By defining a problem environment, solutions can be
evolved in ways similar as to that in nature.
An extension to the classic Genetic Algorithm is a Classifier
System. This is a rule based mechanism which analyses a set of inputs
from an environment and performs an action in that environment based
on its internal rules. However, the Genetic Algorithm is used to
search for rules and generate new rules from existing rules, providing
the Classifier System with an ability to adapt quickly to the problem
environment it finds itself in.
The basic operation of both Genetic Algorithms and Classifier Systems
will be explained and then a sample application will be discussed -
Adaptive Firewalls. In this application a standard Firewall would
adapt to the type of traffic flowing across it and determine over time
which types of traffic posed a threat and which were safe to allow
through. The Classifier System would be used to make this decision.
Unless otherwise noted, the security seminar is held on Wednesdays at 4:30P.M.
STEW G52 (Suite 050B), West Lafayette Campus. More information...