Scholarlink

<p>Zero-Trust Network Architectures (ZTNA) demand innovative cybersecurity mechanisms capable of adapting to more complex attacks and independently healing system failures with minimal human intervention. Conventional intrusion detection and recovery systems are typically plagued with static trust assumptions, response latencies, limited contextual awareness, and high false positive rates in dynamic threat environments. To mitigate these challenges, this research proposes a new framework consisting of Artificial Intelligence-driven fault detection and autonomous recovery systems, integrated with Autoencoder based anomaly reconstruction, contextual trust scoring, reinforcement learning-driven adaptive recovery, and policy-aware decision making in the context of Zero-Trust environments. The framework considers streaming telemetry data to perform anomaly detection and response orchestration in near real time and at low latency using Apache Kafka and Apache Flink. Trust is evaluated dynamically based on behavioral metrics, consistency and reliability of feedback, and fault history in order to facilitate intelligent node classification and secure adaptive policy. Traffic rerouting, node isolation, adaptive policy, and service redeployments are considered autonomous recovery actions optimized by reinforcement learning. The framework yielded 97% anomaly detection accuracy and 96% autonomous recovery success rate, with a 0.965 F1-Score for DDoS detection and a 93% reduction in operational downtime. The Mean Time to Repair (MTTR) was achieved at 1.8s, with significantly diminished false positive rates in comparison to other memory based frameworks such as Deep SVDD, DAGMM, USAD, TranAD, and the Transformer-based framework and traditional rule-based methods. The proposed framework also achieved a 910&#xa0;MB memory usage optimization with a high fault simulation of 10,000 nodes. The results indicate the framework provided next generation self-healing Zero-Trust Cybersecurity Infrastructure with scalable, robust, and adaptive intelligent Cybersecurity Solutions.</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

Artificial Intelligence driven fault detection and autonomous recovery in zero trust network architectures

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月7日

查看全文
  • Sami Morsi,
  • Rajit Nair,
  • Hasan Alkahtani,
  • Sultan Ahmad,
  • Theyazn H. H. Aldhyani,
  • Hikmat A. M. Abdeljaber

摘要

Zero-Trust Network Architectures (ZTNA) demand innovative cybersecurity mechanisms capable of adapting to more complex attacks and independently healing system failures with minimal human intervention. Conventional intrusion detection and recovery systems are typically plagued with static trust assumptions, response latencies, limited contextual awareness, and high false positive rates in dynamic threat environments. To mitigate these challenges, this research proposes a new framework consisting of Artificial Intelligence-driven fault detection and autonomous recovery systems, integrated with Autoencoder based anomaly reconstruction, contextual trust scoring, reinforcement learning-driven adaptive recovery, and policy-aware decision making in the context of Zero-Trust environments. The framework considers streaming telemetry data to perform anomaly detection and response orchestration in near real time and at low latency using Apache Kafka and Apache Flink. Trust is evaluated dynamically based on behavioral metrics, consistency and reliability of feedback, and fault history in order to facilitate intelligent node classification and secure adaptive policy. Traffic rerouting, node isolation, adaptive policy, and service redeployments are considered autonomous recovery actions optimized by reinforcement learning. The framework yielded 97% anomaly detection accuracy and 96% autonomous recovery success rate, with a 0.965 F1-Score for DDoS detection and a 93% reduction in operational downtime. The Mean Time to Repair (MTTR) was achieved at 1.8s, with significantly diminished false positive rates in comparison to other memory based frameworks such as Deep SVDD, DAGMM, USAD, TranAD, and the Transformer-based framework and traditional rule-based methods. The proposed framework also achieved a 910 MB memory usage optimization with a high fault simulation of 10,000 nodes. The results indicate the framework provided next generation self-healing Zero-Trust Cybersecurity Infrastructure with scalable, robust, and adaptive intelligent Cybersecurity Solutions.