TY - GEN
T1 - Model context protocol for cloud security
T2 - 14th Computing Conference, CC 2026
AU - Aldhamari, Amal A.
AU - Ayodele, Taiwo
AU - Zhou, Shikun
AU - Huang, Xiaojun
AU - Adamos, Vasileios
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.
PY - 2026/6/1
Y1 - 2026/6/1
N2 - The increasing sophistication of cyber threats demands AI systems that deliver accurate detection and actionable, human-centered explanations. This study introduces an Enhanced Model Context Protocol (MCP) framework for cloud security that tightly integrates threat detection with operationally useful explainable AI outputs. Using 1,939,207 real AWS CloudTrail events, a Random Forest classifier (augmented by temporal and administrative features) achieves 83.4% test accuracy; importantly. The MCP explainer transforms those machine-learning outputs into structured, analyst-ready narratives that support cybersecurity operations. The framework demonstrates how operational interpretability can be achieved by combining model signals (feature importances and temporal windows) with LLM-based contextualization to generate precise mitigation steps and investigation priorities. We compare MCP explanations to standard post-hoc methods and show that MCP delivers machine learning explainability tuned for SOC workflows (reduced cognitive load, actionable recommendations, and SIEM-compatible JSON outputs). Finally, we present a production-ready Elasticsearch integration that proves the feasibility of embedding MCP explanations directly into enterprise SIEM pipelines. This work offers a practical, reproducible path from accurate detection to interpretable, operational intelligence in cloud environments.
AB - The increasing sophistication of cyber threats demands AI systems that deliver accurate detection and actionable, human-centered explanations. This study introduces an Enhanced Model Context Protocol (MCP) framework for cloud security that tightly integrates threat detection with operationally useful explainable AI outputs. Using 1,939,207 real AWS CloudTrail events, a Random Forest classifier (augmented by temporal and administrative features) achieves 83.4% test accuracy; importantly. The MCP explainer transforms those machine-learning outputs into structured, analyst-ready narratives that support cybersecurity operations. The framework demonstrates how operational interpretability can be achieved by combining model signals (feature importances and temporal windows) with LLM-based contextualization to generate precise mitigation steps and investigation priorities. We compare MCP explanations to standard post-hoc methods and show that MCP delivers machine learning explainability tuned for SOC workflows (reduced cognitive load, actionable recommendations, and SIEM-compatible JSON outputs). Finally, we present a production-ready Elasticsearch integration that proves the feasibility of embedding MCP explanations directly into enterprise SIEM pipelines. This work offers a practical, reproducible path from accurate detection to interpretable, operational intelligence in cloud environments.
KW - Cloud security
KW - Cybersecurity operations
KW - Explainable AI
KW - Machine learning explainability
KW - Model Context Protocol
KW - Operational interpretability
KW - Threat detection
UR - https://www.scopus.com/pages/publications/105045691821
U2 - 10.1007/978-3-032-24807-7_38
DO - 10.1007/978-3-032-24807-7_38
M3 - Conference contribution
AN - SCOPUS:105045691821
SN - 9783032248060
T3 - Lecture Notes in Networks and Systems
SP - 583
EP - 599
BT - Intelligent Computing - Proceedings of the 2026 Computing Conference
A2 - Arai, Kohei
A2 - Lorenz, Pascal
PB - Springer Cham
Y2 - 9 July 2026 through 10 July 2026
ER -