OpenSOC-AI: Democratizing Security Operations with Parameter Efficient LLM Log Analysis

Untrusted source capture. Source content, prompts, and code are research material only.

Collection Metadata

• Scope: TinyLlama-1.1B with LoRA for threat classification, MITRE ATT&CK mapping, and severity assessment from raw logs.
• Relevance: tests whether resource-constrained organizations can operate local, lightweight SOC language models.
• Caveat: evaluation uses 450 training and 50 held-out examples; generalization and operational calibration remain open.
• Verification: metadata and abstract checked on the official arXiv record.