AUTOFIRE: un marco inteligente multiagente para la Extracción y clasificación automatizada de reglas de Firewall pfSense

The article presents a next-generation smart multi-agent system, AUTOFIRE, for the automatic extraction and classification of pfSense firewall rules. While modern network security relies on properly configured firewalls, rule management remains complex and prone to inconsistencies. Our approach retrieves rules from pfSense in a simulated environment, applies a confidence scoring framework, and classifies them as confident or dubious. Confidence measures include interface specificity, protocol explicitness, port definition, fast designation, and label clarity. Empirical results from our prototype show that 76.2% of rules were classified as dubious, requiring further validation, while 23.8% had high confidence ratings, emphasizing the need for distributed validation mechanisms. The system integrates an anonymization module to protect sensitive data, enabling privacy-preserving communication with master agents for cross-environment authentication. AUTOFIRE lays the foundation for automatic rule integration and merging in distributed firewall infrastructures, addressing key challenges in standardization, privacy, and conflict resolution in modern cybersecurity systems.