Scholarlink

<p>Modern Big Data-driven and Cloud based infrastructures face major threats from ever more sophisticated forms of cyber threats, especially Distributed Denial-of-Service (DDoS) attacks. Conventional Intrusion Detection Systems (IDS) have limited scalability and generalization, static rule-based logic which fails to meet the challenge of an evolving threat landscape. The paper proposes a unified detection and response pipeline which combines CloudSim-based simulation, Multi-Objective Optimization (MOO), Subgraph Federated Learning (SFL), Variational Autoencoding (VAE), Adaptive Boosting (AdaBoost), and Dynamic Role-Based Access Control (RBAC) leading towards a framework which is called as Scalable Cloud Adaptative Federated Learning Optimized Detection (SCA-FLOD). Our framework is evaluated against large-scale and heterogeneous DDoS datasets (CICIDS2017, CIC-DDoS2019, BoT-IoT, and CAIDA) that are considered realistic for cloud and Internet of Things (IoT) attack situations. SCA-FLOD outperforms contemporary approaches on every dataset, attaining up to 99.23% accuracy (98.78%–99.23%) and up to 99.70% F1-score (99.18%–99.70%)(SCA-FLOD has reached state-of-the-art performance)).</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

SCA-FLOD: a federated multi-objective learning framework for big data-driven threat detection and adaptive security policy enforcement in cloud infrastructure

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • Richa Verma,
  • Ravindara Bhatt

摘要

Modern Big Data-driven and Cloud based infrastructures face major threats from ever more sophisticated forms of cyber threats, especially Distributed Denial-of-Service (DDoS) attacks. Conventional Intrusion Detection Systems (IDS) have limited scalability and generalization, static rule-based logic which fails to meet the challenge of an evolving threat landscape. The paper proposes a unified detection and response pipeline which combines CloudSim-based simulation, Multi-Objective Optimization (MOO), Subgraph Federated Learning (SFL), Variational Autoencoding (VAE), Adaptive Boosting (AdaBoost), and Dynamic Role-Based Access Control (RBAC) leading towards a framework which is called as Scalable Cloud Adaptative Federated Learning Optimized Detection (SCA-FLOD). Our framework is evaluated against large-scale and heterogeneous DDoS datasets (CICIDS2017, CIC-DDoS2019, BoT-IoT, and CAIDA) that are considered realistic for cloud and Internet of Things (IoT) attack situations. SCA-FLOD outperforms contemporary approaches on every dataset, attaining up to 99.23% accuracy (98.78%–99.23%) and up to 99.70% F1-score (99.18%–99.70%)(SCA-FLOD has reached state-of-the-art performance)).