An Investigation of Machine Learning Techniques for the Detection of Unknown Malicious Code

ABSTRACT\ud AN INVESTIGATION OF MACHINE LEARNING TECHNIQUES FOR\ud THE DETECTION OF UNKNOWN MALICIOUS CODE\ud by\ud ?? Mark Kellogg 2010\ud Master of Science in Computer Science\ud California State University, Chico\ud Fall 2010\ud Traditional forms of anti-virus like signature recognition are lacking in that\ud they can only fight against viruses after they have created a significant enough impact\ud to be identified as malicious. Recently researchers have been applying modern machine\ud learning techniques to the task of detecting malicious programs before they even have a\ud chance to cause any damage. This paper investigates a selection of these techniques in a\ud controlled environment on the Microsoft Windows platform. The selected techniques\ud are tested on the same set of malicious code and the resulting performance and effectiveness\ud of each is analyzed and discussed. The paper concludes with a discussion of\ud potential directions the research could be taken in the future.CSU, Chic