Privacy-Preserving Autonomous Alert Triage and Incident Ticketing Using Local Large Language Models and Wazuh SIEM
Tier-1 Security Operations Center (SOC) analysts face persistent alert fatigue caused by high-volume Security Information and Event Management (SIEM) telemetry, repetitive manual enrichment, and fragmented incident ticket creation. While commercial cloud-hosted Large Language Models (LLMs) offer natural-language summarization capabilities, transmitting raw SIEM alerts containing internal hostnames, IP topologies, user identities, and vulnerability indicators to third-party APIs creates severe confidentiality, intellectual property, and regulatory compliance risks. This paper presents AI-The-Autonomous-SOC, an open-source, self-hosted security automation pipeline that unifies Wazuh SIEM alert ingestion, local LLM risk reasoning, and automated incident ticketing. The architecture continuously polls the Wazuh Indexer REST API, filters high-priority security events, evaluates threat context and MITRE ATT&CK mappings locally via an on-premises Ollama (Llama 3) inference engine using schema-constrained structured outputs, and automatically generates standardized incident tickets in Jira Cloud with zero external telemetry egress.
Authors
- Kavana Arabagatta Shivarudrappa
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-20
- DOI
- https://doi.org/10.5281/zenodo.24382749
- Primary Topic
- Network Security and Intrusion Detection
- Type
- preprint