Leveraging large language models for enhanced threat detection in security operations centers

Kotilingala, Sudheer (2025) Leveraging large language models for enhanced threat detection in security operations centers. World Journal of Advanced Engineering Technology and Sciences, 15 (1). pp. 579-591. ISSN 2582-8266

Security Operations Centers (SOCs) face mounting challenges in effectively identifying and responding to threats amidst overwhelming alert volumes. Traditional rule-based detection systems struggle with contextual understanding, creating high false-positive rates and analyst fatigue. Large Language Models (LLMs) offer a transformative solution through their advanced contextual awareness, pattern recognition capabilities, adaptability, and natural language processing. This technical article proposes a comprehensive framework for integrating LLMs into SOC workflows to enhance threat detection while reducing false positives. The framework addresses four key objectives: scalable context-aware alert classification, high-accuracy false-positive reduction, analyst workload optimization, and seamless integration with existing infrastructure. Technical implementation considerations include data privacy safeguards, latency optimization, explainability techniques, and domain-specific training requirements. The expected outcomes encompass enhanced detection accuracy for sophisticated threats, improved response times, increased analyst satisfaction, more efficient resource allocation, streamlined compliance reporting, and strategic security intelligence for proactive defense.