MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641107340 A) filed by Dr T Charan Singh; B. Keerthi Chaitanya; Snvasrk Prasad; Kommala Aparna; Dr. Muntimadugu Vijaya Kanth; and Ms. Chakala Navya on September 07, 2026, for Privacy-Preserving Analytics For Differential Privacy Analysis And Utility-Accuracy Trade-Off Improvements.
Inventors include Dr T Charan Singh; B. Keerthi Chaitanya; Snvasrk Prasad; Kommala Aparna; Dr. Muntimadugu Vijaya Kanth; and Ms. Chakala Navya.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: The present invention introduces a practical system for conducting privacy-preserving analytics that carefully examines differential privacy guarantees while steadily improving the balance between data utility and analytical accuracy. Sensitive information is processed through a controlled pipeline that first measures how much any single record can influence query results, then adds carefully calibrated noise according to a chosen privacy level. At the same time, the system continuously checks the quality of the privatized results against useful accuracy measures, allowing an adaptive controller to adjust noise levels or remaining privacy allowances so that later analytical steps keep higher fidelity without violating the overall privacy limit. A clear record of every privacy expenditure is kept for auditing, and experiments on typical workloads show that the adaptive approach reduces estimation errors and improves model performance compared with fixed-noise methods that use the same total privacy budget. The resulting framework therefore supports reliable statistical insights and machine-learning tasks on protected data while maintaining strong formal privacy protections.
Disclaimer: Curated by HT Syndication.
