Scholarlink

<p>The increasing sophistication of cyber threats necessitates intelligent, adaptive, and mathematically grounded solutions for modern Security Operations Centers (SOCs). This paper presents an AI-enhanced SOC framework that integrates twisted group algebra with graph neural networks (GNNs) to achieve efficient and interpretable threat detection and mitigation. In the proposed model, network entities and events are represented as vertices within an algebraically structured graph, where edges are governed by twisted group operations capturing non-commutative relationships between system components and security states. This algebraic formalism enables dynamic encoding of contextual dependencies, allowing the GNN to learn complex correlation patterns across multi-domain security data. Reinforcement learning mechanisms are employed to optimize autonomous decision-making for alert prioritization and adaptive response. The theoretical foundation is supported by a rigorous analysis of the algebraic graph Laplacian under twisted operations, ensuring stability and computational soundness. Experimental evaluations using benchmark SOC datasets demonstrate superior accuracy, reduced false-positive rates, and improved resilience against evolving threat vectors compared to conventional AI-based SOC systems. The proposed framework thus bridges advanced algebraic theory and modern AI, providing a scalable, explainable, and quantum-resilient architecture for next-generation cybersecurity infrastructures.</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

A graph-based approach to adaptive threat detection for security operations centers

  • 期刊论文

  • 发表日期：

    2026年3月18日

查看全文
  • R. Krishnaprabha

摘要

The increasing sophistication of cyber threats necessitates intelligent, adaptive, and mathematically grounded solutions for modern Security Operations Centers (SOCs). This paper presents an AI-enhanced SOC framework that integrates twisted group algebra with graph neural networks (GNNs) to achieve efficient and interpretable threat detection and mitigation. In the proposed model, network entities and events are represented as vertices within an algebraically structured graph, where edges are governed by twisted group operations capturing non-commutative relationships between system components and security states. This algebraic formalism enables dynamic encoding of contextual dependencies, allowing the GNN to learn complex correlation patterns across multi-domain security data. Reinforcement learning mechanisms are employed to optimize autonomous decision-making for alert prioritization and adaptive response. The theoretical foundation is supported by a rigorous analysis of the algebraic graph Laplacian under twisted operations, ensuring stability and computational soundness. Experimental evaluations using benchmark SOC datasets demonstrate superior accuracy, reduced false-positive rates, and improved resilience against evolving threat vectors compared to conventional AI-based SOC systems. The proposed framework thus bridges advanced algebraic theory and modern AI, providing a scalable, explainable, and quantum-resilient architecture for next-generation cybersecurity infrastructures.