How Ant Group’s Aspect‑Fusion AI Boosts Threat Detection and Reduces False Alarms

Recently, the 2025 Beijing Cybersecurity Conference listed Ant Group’s “Aspect‑Fusion Intelligence in Threat Detection” as an outstanding case.

The case, selected by the “practical security operation” track, focuses on using artificial intelligence to enhance security operations, threat intelligence, and incident response, addressing real‑world network attacks.

Aspect‑Fusion Intelligence System was proposed by Ant Group and applied to cybersecurity. It combines a security large model’s deep semantic understanding with expert knowledge, following the DKCF (Data/Knowledge/Collaboration/Feedback) paradigm to address issues such as vague professional concepts, reasoning gaps, inefficient feedback loops, and limited attention in collaborative chain security.

In practice, the system first gathers multi‑dimensional data (system, network, application) via aspect techniques, extracts credentials, and builds complete attack chains. The large model then merges expert and machine intelligence to improve unknown threat discovery and establish application‑centric security baselines, reducing false alarms. Finally, a security knowledge graph is constructed from system, network, application, and intelligence data to enhance explainability and enable attack pattern discovery and tracing.

Currently, the system provides intrusion detection for Ant Group’s business, handling trillion‑scale real‑time and offline analysis, quantifying intrusion detection levels, and meeting rapid response requirements in strong adversarial scenarios. It achieves financial‑grade stability and reliability, delivering excellent results in drills and real‑world attacks while significantly improving risk‑operation efficiency.

The “Aspect‑Fusion Intelligence” project has earned multiple industry recognitions, including the 2024 World Intelligence Expo “Find Smart Technology Innovation Application Typical Case”, the 2024 AI Pioneer Case by the China AI Industry Development Alliance, and the China Academy of Information and Communications Technology’s “Model Application Sample Project”.