EUVIMEDEUVIMED entdecken
Crossref

Autonomous Penetration Testing Framework Based on Reinforcement Learning and Intrinsic Motivation

Rasha A. Alsakran, Sultan S. Alqahtani

International Journal of Advanced Computer Science and Applications · 2026 · 17 (10)

DOI: 10.14569/ijacsa.2026.0171009 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Advanced Computer Science and Applications, 2026. Verfasst von Rasha A. Alsakran und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Ethical penetration testing (PT) is a structured security assessment in which cybersecurity experts simulate real attacks on an organization’s digital assets in a controlled testing environment. The primary objective is to evaluate the effectiveness of cybersecurity controls by detecting vulnerabilities in systems, networks, or applications. Subsequently, report these vulnerabilities to facilitate remediation prior to exploitation by malicious adversaries. In most cases, manual penetration testing relies mostly on expert knowledge and requires substantial human resources, particularly in cases where the network infrastructure is large and complex. Hence, the development of autonomous penetration testing emerges as a promising avenue. Recently, reinforcement learning (RL) has gained attention as an effective approach for automating penetration testing by enabling intelligent agents to interact with target environments and make adaptive attack decisions. This research study proposes an automated penetration testing framework termed PTiRL. powered by reinforcement learning, particularly the Deep Q-Network (DQN) algorithm. Unlike prior work, the proposed framework enhances agent learning by incorporating intrinsic motivation (IM) mechanisms, such as Random Network Distillation (RND), to address challenges associated with sparse and delayed rewards. A comparative evaluation is conducted between our PTiRL agents that utilize intrinsic motivation and those agents relying solely on standard reward signals, demonstrating the effectiveness of intrinsic motivation in improving environment exploration and learning efficiency in attack path discovery. The experiments were conducted using the Network Attack Simulator (NASim), which is an abstract network attack simulator that has benchmark scenarios designed to be used for rapid testing of autonomous penetration testing agents based on reinforcement learning.

Zitieren

APA

Alsakran, R. A., & Alqahtani, S. S. (2026). Autonomous Penetration Testing Framework Based on Reinforcement Learning and Intrinsic Motivation. International Journal of Advanced Computer Science and Applications, 17(10). https://doi.org/10.14569/ijacsa.2026.0171009

Vancouver

Alsakran RA, Alqahtani SS. Autonomous Penetration Testing Framework Based on Reinforcement Learning and Intrinsic Motivation. International Journal of Advanced Computer Science and Applications. 2026;17(10). doi:10.14569/ijacsa.2026.0171009

BibTeX
@article{Alsakran2026,
  author = {Alsakran, Rasha A. and Alqahtani, Sultan S.},
  title = {Autonomous Penetration Testing Framework Based on Reinforcement Learning and Intrinsic Motivation},
  journal = {International Journal of Advanced Computer Science and Applications},
  year = {2026},
  volume = {17},
  number = {10},
  doi = {10.14569/ijacsa.2026.0171009}
}
Kostenlos starten →