MUMBAI, India, Feb. 13 -- Intellectual Property India has published a patent application (202541125032 A) filed by Saveetha Engineering College, Chennai, Tamil Nadu, on Dec. 11, 2025, for 'neuro-symbolic cyber threat analyzer: hybrid llm ontology reasoning for automated cyber threat intelligence.'

Inventor(s) include Dr. Gowri Ganeshs; Dhinesh Kumar K S; Mohammed Faizal J; Mohanish K; and Anandhamoorthy K.

The application for the patent was published on Feb. 13, under issue no. 07/2026.

According to the abstract released by the Intellectual Property India: "Modem cybenittacks employ highly sophisticated and complex multi-staged maneuvers that can easily confuse and overpower traditional rule based security instruments as well as security operations centers (SOCs) that are heavily dependent on such tools and arc run by security analysts. Most of the time, threat intelligence is provided in loosely structured textual formats, and therefore only trained analysts can manually extract techniques, indicators of compromise, vulnerabilities, and adversary behaviors from them. Large Lan guage Models have good natural language understanding abilities, but they are not always consistent with the domain, whereas ontologies provide semantic strictness but are not able to understand natural text. In order to . tackle these problems, this research is about the Neuro-Symbolic Cyber Threat Analyzer, a hybrid system that combines local LLM-driven extraction with OWL-based on tology reasoning to provide automated cyber threat interpretation. The system uses a custom A TT &CK -aligned ontology, Owlready2 reasoning, and LLa MA3:8B (Ollama)-powered extraction that allows it to link the hack methods, techniques, malware, threat actors, vulnerabilities, CAPEC patterns, and Indica tors of Compromise (IOCs). Besides that, it also assigns risk scores, confidence weights, and gives defense recommendations based on MITRE D3FEND. The user interface based on Streamlit allows quick access to the data being pro cessed along with attack summaries and knowledge-graph mappings. The ex periments show that the system can achieve reliable extraction accuracy and consistent semantic validation, indicating the promise of neuro-symbolic intelligence in handling a large volume of cyber threats in a scalable manner."

Disclaimer: Curated by HT Syndication.