BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Adaptive Runtime Verification - Prof  Ezio Bartocci\, University o
 f Wien
DTSTART:20130206T141500Z
DTEND:20130206T151500Z
UID:TALK42369@talks.cam.ac.uk
CONTACT:Stephen Clark
DESCRIPTION:We present Adaptive Runtime Verification (ARV)\, a new approac
 h to\nruntime verification in which overhead control\, runtime verificatio
 n\nwith state estimation\, and predictive analysis are synergistically\nco
 mbined. Overhead control maintains the overhead of runtime\nverification a
 t a specified target level\, by enabling and disabling\nmonitoring of even
 ts for each monitor instance as needed. In ARV\,\npredictive analysis base
 d on a probabilistic model of the monitored\nsystem is used to estimate ho
 w likely each monitor instance is to\nviolate a given temporal property in
  the near future\, and these\ncriticality levels are fed to the overhead c
 ontrollers\, which allocate\na larger fraction of the target overhead to m
 onitor instances with\nhigher criticality\, thereby increasing the probabi
 lity of violation\ndetection. Since overhead control causes the monitor to
  miss events\,\nwe use Runtime Verification with State Estimation (RVSE) t
 o estimate\nthe probability that a property is satisfied by an incompletel
 y\nmonitored run. A key aspect of the ARV framework is a new algorithm\nfo
 r RVSE that performs the calculations offline\, dramatically reducing\nthe
  runtime overhead of RVSE\, at the cost of introducing some\napproximation
  error. We demonstrate the utility of ARV on a\nsignificant case study inv
 olving runtime monitoring of concurrency\nerrors in the Linux kernel.\n
LOCATION:Lecture Theatre 1\, Computer Laboratory
END:VEVENT
END:VCALENDAR