<?xml version="1.0" encoding="UTF-8"?><xml><records><record><source-app name="Biblio" version="6.x">Drupal-Biblio</source-app><ref-type>10</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Blaine Nelson</style></author><author><style face="normal" font="default" size="100%">Marco Barreno</style></author><author><style face="normal" font="default" size="100%">Fuching Jack Chi</style></author><author><style face="normal" font="default" size="100%">Anthony D. Joseph</style></author><author><style face="normal" font="default" size="100%">Benjamin I. P. Rubinstein</style></author><author><style face="normal" font="default" size="100%">Udam Saini</style></author><author><style face="normal" font="default" size="100%">Charles Sutton</style></author><author><style face="normal" font="default" size="100%">J. D. Tygar</style></author><author><style face="normal" font="default" size="100%">Kai Xia</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">Exploiting machine learning to subvert your spam filter</style></title><secondary-title><style face="normal" font="default" size="100%">1st Usenix Workshop on Large-Scale Exploits and Emergent Threats</style></secondary-title></titles><dates><year><style  face="normal" font="default" size="100%">2008</style></year><pub-dates><date><style  face="normal" font="default" size="100%">04/2008</style></date></pub-dates></dates><urls><web-urls><url><style face="normal" font="default" size="100%">http://www.usenix.org/events/leet08/tech/full_papers/nelson/nelson.pdf</style></url></web-urls></urls><publisher><style face="normal" font="default" size="100%">USENIX Association</style></publisher><pub-location><style face="normal" font="default" size="100%">San Francisco, CA</style></pub-location><abstract><style face="normal" font="default" size="100%">Using statistical machine learning for making security decisions introduces new vulnerabilities in large scale systems. This paper shows how an adversary can exploit statistical machine learning, as used in the SpamBayes spam filter, to render it useless--even if the adversary's access is limited to only 1% of the training messages. We further demonstrate a new class of focused attacks that successfully prevent victims from receiving specific email messages. Finally, we introduce two new types of defenses against these attacks.</style></abstract></record></records></xml>