BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Detecting Network Traffic Anomalies - Paul Barford\, University of
  Wisconsin-Madison
DTSTART:20101117T141500Z
DTEND:20101117T151500Z
UID:TALK26388@talks.cam.ac.uk
CONTACT:Stephen Clark
DESCRIPTION:Unwanted events such as attacks\, misconfigurations and failur
 es can\ncause significant disruptions in day-to-day network operations.\nE
 ffective management and mitigation of these events is predicated on\nfast 
 and accurate identification.  One way to identify these events is\nto appl
 y an anomaly detection algorithm to network traffic streams.\nIn this talk
 \, I will describe the basic framework for anomaly\ndetection in network t
 raffic\, and provide perspective on standard\nanomaly detection methods an
 d why they have not been widely deployed.\nI will then describe a new flex
 ible but precise anomaly detection\nmethod that we have recently developed
  called BasisDetect.  Using a\nsmall dataset with labeled anomalies\, our 
 framework uses a novel basis\npursuit algorithm to enable detection of a l
 arge class of anomalies in\ndifferent types of network data\, both from si
 ngle source and a network\nwide perspective.  Using a combination of synth
 etic and real world\ndata\, I will show that BasisDetect significantly red
 uces false alarms\nversus other anomaly detection methods.\n\n\n
LOCATION:Lecture Theatre 1\, Computer Laboratory
END:VEVENT
END:VCALENDAR