Identify suspicious URL, a complete application for online learning

Google Tech Talk May 2010 Summary fifth But Justin presented. We explore approaches to learning to detect malicious websites (those involved in the fraud prosecution), with basic functionality to the host and the vocabulary of the URL. It is shown that this application is particularly suitable for on-line algorithms, such as the size of the training data more efficiently than can be processed in batches and distribution functions, writing the malicious URL changing. Withreal-time system, we have asked for collection of elements of the URL, the URL of a real source of great online email providers developed combined revealed That Recently, online algorithms can develop more accurate techniques for classification match at the precision available day, up to 99% in the set of balanced data. Presentation: My cseweb.ucsd.edu Justin is a PhD student at the University of California at San Diego taught by Stefan Savage, Geoff Voelker and Lawrence Saul. His research interests are in systems and networksFocus on network security, and his current focus is the application of machine learning to security problems. He will join the University of Berkeley as a post-doc after graduation. [Home: www.cs.ucsd.edu]



http://www.youtube.com/watch?v=n3iANHusfcY&hl=en