Detecting Ø-Days Attacks With Learning Intrusion Detection Systems presented at Blackhat USA 2004

by Stefano Zanero, M.s. ,

Tags: Security

Summary : Traditional anomaly-based
Intrusion Detection Systems, relying on pattern matching and static
signatures, are not really able to keep up with the creation of new
forms of attacks, and particularly with zero-day attacks. In this talk
we will analyze the problem, and present new types of misuse detection
systems, based on unsupervised learning techniques, that can complement
well traditional IDS systems and help detect zero-days techniques of
attack and various other misbehaviours. A proof of concept based on our
current research prototypes will be also presented.