Date of Award


Document Type


Degree Name

Master of Science


Department of Electrical and Computer Engineering

First Advisor

Thomas E. Dube, PhD.


Adversaries employ malware against victims of cyber espionage with the intent of gaining unauthorized access to information. To that end, malware authors intentionally attempt to evade defensive countermeasures based on static methods. This thesis analyzes a dynamic analysis methodology for malware triage that applies at the enterprise scale. This study captures behavior reports from 64,987 samples of malware randomly selected from a large collection and 25,591 clean executable files from operating system install media. Function call information in sequences of behavior generate feature vectors from behavior reports from the les. The results of 64 experiment combinations indicate that using more informed behavior features yields better performing models with this data set. The decision tree classifier attained a max performance of 0.999 area under the ROC curve and 99.4% accuracy using argument information with function sequence lengths from 11-14. This methodology contributes to strategic cyber situation awareness by fusion with fast malware detection methods, such as static analysis, to change the game of malware triage in favor of cyber defense. This method of triage reduces the number of false alarms from automatic analysis that allows a 97% workload reduction over using a static method alone.

AFIT Designator


DTIC Accession Number