Scholarlink

<p>The present research introduces GORGO, a unique penetration testing tool that utilizes Reinforcement Learning. It performs vulnerability assessments on 5G—and beyond—infrastructures against sophisticated Denial of Service (DoS) attacks driven by Artificial Intelligence (AI). GORGO consists of multiple Deep Q-Learning agents working in collaboration to assess the resilience of target infrastructure, making it the only penetration testing tool capable of performing such comprehensive evaluations. The objective is to construct the optimal attack strategy and provide the information to understand and mitigate AI-driven DoS attacks. The rationale is to provide detailed insights into the extent of Quality of Service degradation and to identify which strategic choices had the most significant impact on the target infrastructure. GORGO was evaluated in one real Software Defined Network (SDN)-based Cloud Native 5G infrastructure and one Beyond 5G (B5G) testbed. The experimental results demonstrate that GORGO is a highly efficient Reinforcement Learning model and successfully uncovered vulnerabilities in both testing environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GORGO: A Reinforcement Learning Model for DoS Vulnerability Analysis on 5G and Beyond Networks

  • 期刊论文

  • 发表日期：

    2025年10月22日

查看全文
  • Konstantinos Giapantzis,
  • Apostolis Zarras,
  • Sarantis Kalafatidis,
  • Antonios Lalas,
  • Konstantinos Votis,
  • Christos Xenakis

摘要

The present research introduces GORGO, a unique penetration testing tool that utilizes Reinforcement Learning. It performs vulnerability assessments on 5G—and beyond—infrastructures against sophisticated Denial of Service (DoS) attacks driven by Artificial Intelligence (AI). GORGO consists of multiple Deep Q-Learning agents working in collaboration to assess the resilience of target infrastructure, making it the only penetration testing tool capable of performing such comprehensive evaluations. The objective is to construct the optimal attack strategy and provide the information to understand and mitigate AI-driven DoS attacks. The rationale is to provide detailed insights into the extent of Quality of Service degradation and to identify which strategic choices had the most significant impact on the target infrastructure. GORGO was evaluated in one real Software Defined Network (SDN)-based Cloud Native 5G infrastructure and one Beyond 5G (B5G) testbed. The experimental results demonstrate that GORGO is a highly efficient Reinforcement Learning model and successfully uncovered vulnerabilities in both testing environments.