InfoSys Launches Graph-Based AI Security Solution GOS, Says Enterprise-Wide Deployment at Major Conglomerate Proves Its Technology

InfoSys has officially launched GOS (Graph Optimized Security), a graph-based AI security solution. On the 6th, the company announced that it would supply GOS enterprise-wide to a major domestic semiconductor company. GOS is a graph-based security solution for detecting abnormal behavior.

InfoSys is an ontology-based enterprise AI company. It connects a company's vast data and industrial knowledge through an ontology-based framework and is building industry-specific AI that can understand and infer business context. InfoSys primarily operates through Axiom and DELTAflow; Axiom is an ontology-based enterprise AI solution, and DELTAflow is a solution for reading industrial drawings and converting them into knowledge graphs.

InfoSys said the launch of GOS and its supply to a major conglomerate were completed quickly, demonstrating both its technological capabilities and commercial viability.

InfoSys explained GOS as a tool that analyzes an organization's massive security logs to identify threats. GOS learns organizational behavior patterns and operates by graphing relationships among data.

GOS links the relationships among large-scale security logs — including accounts, processes, files, and access records — into a graph, then identifies anything that deviates from normal behavior as a security threat. This allows it to analyze and alert on attack entry points, movement paths, and related actions that are difficult to identify by checking individual logs alone.

GOS then uses its in-house Graph ML technology to analyze large-scale security logs and sends only highly likely threat candidates to Analyzer LLM. Analyzer LLM carefully verifies the threat likelihood and behavioral context of the candidates it receives.

InfoSys said that analyzing every log with a commercial large language model, or LLM, could incur annual costs in the several trillion won range. To address this, GOS has adopted a method that reduces the share of analysis subject to LLM processing to about 0.068% of the entire behavior graph.

With this approach, GOS significantly reduces operating costs while maintaining detection accuracy. Its structure first filters high-risk candidates with Graph ML and then has Analyzer LLM perform a detailed verification only on those candidates, reducing the cost burden that would arise if all logs were sent directly to an LLM.

The solution is designed to be used alongside SIEM and SOAR without replacing the systems currently in operation at a company. GOS performs graph analysis on the logs collected by SIEM. Through this, GOS provides risk levels, attack paths, and grounds for its judgments.

Verified threat information is linked to SOAR response procedures. Examples of SOAR response procedures include account locking, endpoint isolation, and communication blocking. As a result, companies can preserve existing security investments while adding an AI layer that makes alerts more accurate.

Taek Jung-soo, CEO of InfoSys, said the key to enterprise security lies not in increasing alerts, but in establishing an accurate decision-making system that both prevents missed real threats and avoids disrupting normal operations. He said GOS has been recognized for its accuracy and commercial viability in real enterprise environments. He added that, following this project win, the company plans to expand the solution's application to enterprise clients with large-scale security logs and complex business environments.

Source: TECHWORLD · Lee Gwang-jae
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407774

References

• https://www.epnc.co.kr/news/articleView.html?idxno=407774