2026-05-24T05:26:21Zhttps://keep.lib.asu.edu/oai/request

oai:keep.lib.asu.edu:node-1318842024-12-18T20:58:30Zoai_pmh:alloai_pmh:repo_items

131884 https://hdl.handle.net/2286/R.I.56108 http://rightsstatements.org/vocab/InC/1.0/ 2020-05 17 pages eng Dejarnett, Eric Andrew Huang, Dijiang Chowdhary, Ankur Computer Science and Engineering Program Barrett, The Honors College Text As automation research into penetration testing has developed, several methods have been proposed as suitable control mechanisms for use in pentesting frameworks. These include Markov Decision Processes (MDPs), partially observable Markov Decision Processes (POMDPs), and POMDPs utilizing reinforcement learning. Since much work has been done automating other aspects of the pentesting process using exploit frameworks and scanning tools, this is the next focal point in this field. This paper shows a fully-integrated solution comprised of a POMDP-based planning algorithm, the Nessus scanning utility, and MITRE's CALDERA pentesting platform. These are linked in order to create an autonomous AI attack platform with scanning, planning, and attack capabilities. planning algorithm POMDP autonomous penetration testing Artificial Intelligence network security AI-Based Autonomous Security Assessment Tool