Intelligent Automation of Linux System Administration via the NeuroSysAI Autonomous Agent: Security and Cost-Efficiency Analysis

Abstrak

Manual Linux system administration frequently suffers from operational inefficiencies and a high risk of human error. As IT infrastructure grows increasingly complex, traditional automation tools often lack the adaptability needed to handle dynamic troubleshooting. To address these limitations, this study introduces NeuroSysAI, an autonomous agent powered by Large Language Models (LLMs) specifically designed to automate server configuration, security management, and system monitoring in Ubuntu 22.04 LTS environments. Our approach implements a hybrid architecture that leverages the Mistral-Nemo API for complex reasoning alongside a local GPT-OSS model via Ollama to optimize operational costs. A primary contribution of this research is the integration of strict tool-use restrictions. This mechanism effectively mitigates the inherent risk of LLM hallucinations, ensuring that terminal command executions remain secure and controlled. Functional validation demonstrates that NeuroSysAI is highly reliable, achieving a 95.8% success rate in resolving administrative tasks with a robust system load capacity of 34 requests per second (RPS). Beyond technical performance, our threat modeling and cost evaluations confirm that this hybrid agent approach provides an optimal balance between infrastructure investment (CapEx) and operational efficiency (OpEx). Ultimately, NeuroSysAI offers system administrators a secure, adaptive, and economically viable solution for modernizing IT operations