Scholarlink

<p>The conventional paradigm is inadequate to address the ever-evolving Cyber Threats resulting from the swift expansion of Cloud Computing, Remote Access, and Distributed Digital Infrastructures. Zero Trust Environments (ZTE) necessitate continuously updated and adaptive trust assessment procedures in response to emerging cyber threats and insider attacks. This study introduces a Hybrid DQN Driven Risk Adaptive Identity Trust Engine (HD-RAITE) platform that integrates unsupervised behavioural anomaly detection to facilitate a continuous and dynamic approach to trust decision-making for each identity. The HD-RAITE consistently gathers identity, device, session, and behavioural characteristics for each user and computes a risk score utilising the results from Isolation Forest and Autoencoders. The computed risk score is transmitted to facilitate optimal trust decisions, including allowing, limiting, escalating authentication, and denying access over Deep Q-Networks (DQN). The assessment of the Risk-based Authentication Dataset underpinned the experimental outcomes of the HD-RAITE, which attained an accuracy of 98.68%, precision of 98.60%, sensitivity of 98.77%, specificity of 98.59%, and an F1 score of 98.68%, thereby surpassing all existing Zero Trust and AI-based models. The results indicate that the HD-RAITE model has substantially enhanced the operational security resilience of the evaluated companies with negligible effects on genuine users.</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

HD-RAITE: a hybrid deep reinforcement learning framework for Risk-Adaptive Identity trust management in Zero Trust Environments

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月23日

查看全文
  • Vanajaroselin Chirchi,
  • M. Prakash,
  • N. Duraimutharasan,
  • C. Viswanathan,
  • P. Gururama Senthilvel,
  • S. Balasubramaniam,
  • Frie Ayalew

摘要

The conventional paradigm is inadequate to address the ever-evolving Cyber Threats resulting from the swift expansion of Cloud Computing, Remote Access, and Distributed Digital Infrastructures. Zero Trust Environments (ZTE) necessitate continuously updated and adaptive trust assessment procedures in response to emerging cyber threats and insider attacks. This study introduces a Hybrid DQN Driven Risk Adaptive Identity Trust Engine (HD-RAITE) platform that integrates unsupervised behavioural anomaly detection to facilitate a continuous and dynamic approach to trust decision-making for each identity. The HD-RAITE consistently gathers identity, device, session, and behavioural characteristics for each user and computes a risk score utilising the results from Isolation Forest and Autoencoders. The computed risk score is transmitted to facilitate optimal trust decisions, including allowing, limiting, escalating authentication, and denying access over Deep Q-Networks (DQN). The assessment of the Risk-based Authentication Dataset underpinned the experimental outcomes of the HD-RAITE, which attained an accuracy of 98.68%, precision of 98.60%, sensitivity of 98.77%, specificity of 98.59%, and an F1 score of 98.68%, thereby surpassing all existing Zero Trust and AI-based models. The results indicate that the HD-RAITE model has substantially enhanced the operational security resilience of the evaluated companies with negligible effects on genuine users.