Scholarlink

The deployment architecture of hybrid clouds associated with public and private infrastructures has emerged as the frame of dynamic containerized micro service apps in contemporary enterprise computing. But this computing elasticity also creates a high-risk dynamic in which container orchestration platforms (i.e: Kubernetes) become vulnerable to advanced persistent threats (APTs), lateral privilege escalation, and runtime exploits. This paper presents the concept of deep learning-enabled security orchestration framework that has the ability to predict possible threats and take an autonomous action in real-time in a micro service cluster that spans the boundaries of hybrid clouds. The unique selling point of the system is the use of a tri-layered architecture to combine LSTM in sequential anomaly detection, CNN in signature identification of API behaviors, and Transformer encoders on contextualization of threats across services. This intelligence is implemented in a monitoring pipeline that is live and constantly consumes system telemetry, such as process calls, inter-container traffic, and system audit logs, and the scoring of risk initiates zero trust countermeasures in the running environment. Test implementation in a Kubernetes-OpenStack hybrid test lab achieved 94.6% in early threat detection, a 37.8% decrease in attack containment and 21.5% decrease in node recovery latency. It also limited untrue alarms by 26.3% compared to establishment IDS and produced a low overhead of resource use through the adjustable-rate container conveyancing of 17.4%. The suggested architecture shows the possibility of the real-time AI-driven cybersecurity incorporation into decision-making in distributed cloud-native mechanisms, and this provides the foundation of autonomous and resilient scalable cloud operations. Moreover, the integration is correlated with the principles of Zero Trust Architecture (ZTA) and configures infrastructure to comply with future edge-cloud security architectures and AI observability layers in Industry 5.0 paradigms.

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

Deep Learning-Based Threat Prediction and Autonomous Response Mechanisms for Containerized Microservices in Hybrid Cloud Deployments

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ahmed Shan-A-Alahi,
  • Harleen Kaur,
  • Emran Hossain,
  • Syed Nazmul Hasan,
  • Clinton Ronjon Barikdar,
  • Jahid Hassan,
  • Jobanpreet Kaur

摘要

The deployment architecture of hybrid clouds associated with public and private infrastructures has emerged as the frame of dynamic containerized micro service apps in contemporary enterprise computing. But this computing elasticity also creates a high-risk dynamic in which container orchestration platforms (i.e: Kubernetes) become vulnerable to advanced persistent threats (APTs), lateral privilege escalation, and runtime exploits. This paper presents the concept of deep learning-enabled security orchestration framework that has the ability to predict possible threats and take an autonomous action in real-time in a micro service cluster that spans the boundaries of hybrid clouds. The unique selling point of the system is the use of a tri-layered architecture to combine LSTM in sequential anomaly detection, CNN in signature identification of API behaviors, and Transformer encoders on contextualization of threats across services. This intelligence is implemented in a monitoring pipeline that is live and constantly consumes system telemetry, such as process calls, inter-container traffic, and system audit logs, and the scoring of risk initiates zero trust countermeasures in the running environment. Test implementation in a Kubernetes-OpenStack hybrid test lab achieved 94.6% in early threat detection, a 37.8% decrease in attack containment and 21.5% decrease in node recovery latency. It also limited untrue alarms by 26.3% compared to establishment IDS and produced a low overhead of resource use through the adjustable-rate container conveyancing of 17.4%. The suggested architecture shows the possibility of the real-time AI-driven cybersecurity incorporation into decision-making in distributed cloud-native mechanisms, and this provides the foundation of autonomous and resilient scalable cloud operations. Moreover, the integration is correlated with the principles of Zero Trust Architecture (ZTA) and configures infrastructure to comply with future edge-cloud security architectures and AI observability layers in Industry 5.0 paradigms.