MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641072277 A) filed by Vellore Institute Of Technology on June 10, 2026, for Quantum Entropy-Anchored Apparatus For Privacy-Preserving Federated Learning.
Inventors include M. Malathi; Shlok Kumar Goenka; and Raghav Tibrewal.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: ABSTRACT QUANTUM ENTROPY-ANCHORED APPARATUS FOR PRIVACY-PRESERVING FEDERATED LEARNING A quantum entropy-anchored dual- mode apparatus for privacy-preserving federated learning comprises a photonic quantum random number generator chip (101) generating random bits via balanced homodyne detection of quantum vacuum fluctuations, an entropy health monitor (102) validating the bits, and a dual-stream entropy splitter (103) partitioning the validated stream into first and second sub-streams using bit-interleaved sampling. A differential privacy module comprises a transform unit (201) converting first sub- stream bits into Gaussian noise samples, a gradient noise injector (202) producing a noised gradient, and a cryptographic attestation unit (203, 204) generating a watermark and digital signature binding the noised gradient to consumed bits. A continual learning module comprises a Rademacher generator (301) mapping second sub-stream bits to perturbation vector elements and a Fisher Information estimator (302) estimating a Fisher Information Matrix diagonal. An aggregation server (401, 402, 403) verifies the watermark and signature and aggregates verified gradients.
Disclaimer: Curated by HT Syndication.
