Nick Duffield, AT&T Labs Research, Florham Park, NJ

Location :

LIP6,
room 549,
104, avenue du President Kennedy,
Paris, 16eme

Directions

Abstract :

Matching packet level features of internet traffic to known signatures is a powerful method for detecting 
traffic anomalies, but is very resource intensive. Flow level information on traffic is more widely collected, 
but provides information at the header level only. This talk describes how machine learning methods can be 
used to associate packet level features internet traffic with its flow level features, and describe a system 
architecture for network-wide flow-alarming built around the ML approach.

Host :

LIP6