MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641114139 A) filed by Vellore Institute Of Technology on September 23, 2026, for Graph-Based Adaptive Privacy Middleware For Enterprise Llm Systems Using Risk Prediction.
Inventors include Banupriya R; and M. Sai Shobith Reddy.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: [0055] An adaptive privacy-preserving middleware system protects healthcare information before processing by Large Language Models (LLMs) or artificial intelligence systems. It receives patient records and constructs a healthcare relationship graph representing entities like patients, providers, and conditions. A Graph- based Risk Assessment (GRRA) module analyzes the graph to determine a patient-level privacy-risk score based on density, uniqueness, and sensitivity, classifying it into a risk level. An Adaptive Privacy Strategy Selection and Application (APSSA) module then selects and applies field-level privacy transformations, such as pseudonymization, masking, or generalization, tailored to each attribute's characteristics and the assessed risk. The protected information is provided to downstream AI/LLM systems, enabling useful processing while reducing sensitive data exposure. [FIG. 1]
Disclaimer: Curated by HT Syndication.
