MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202611090815 A) filed by Noida Institute Of Engineering And Technology Niet on July 26, 2026, for Neuro-Symbolic System And Method For Adversarial Detection And Self-Healing Recovery In Programming-By-Example Synthesis.
Inventors include Ms. Ayushi Gupta; and Mr. Gourav Shelly.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: A neuro-symbolic system (100) for adversarial detection and self-healing recovery in programming-by-example synthesis is disclosed. The system (100) comprises an example ingestion module (110), a cryptographic provenance verification unit (120), a neural anomaly detection module (130) having a neural encoder (132) and attention-based corruption classifier (134), a symbolic margin analyzer (140), an adaptive threshold controller (150), a version-space enumerator (160), a multi-layer consensus engine (170), a self-healing reconstruction module (180), and a program output interface (190). The consensus engine (170) fuses neural corruption scores, provenance flags, and signature-cluster votes to discount adversarial partition majorities, while the reconstruction module (180) iteratively repairs corrupted examples until a bounded-pool robustness radius exceeds a safety constant, thereby producing a synthesized program resilient to worst-case example corruption.
Disclaimer: Curated by HT Syndication.
