The main contribution of this thesis is to give a novel perspective on Active Malware Analysis modeled as a decision making process between intelligent agents. We propose solutions aimed at extracting the behaviors of malware agents with advanced Artificial Intelligence techniques. In particular, we devise novel action selection strategies for the analyzer agents that allow to analyze malware by selecting sequences of triggering actions aimed at maximizing the information acquired. The goal is to create informative models representing the behaviors of the malware agents observed while interacting with them during the analysis process. Such models can then be used to effectively compare a malware against others and to correctly identify the malware family