MUMBAI, India, Oct. 26 -- Intellectual Property India has published a patent application (202631123359 A) filed by Saumen Bandyopadhyay on October 15, 2026, for An Ai-Enabled Multisource Graph Neural Network System For Merchant, Account And Transaction Fraud Risk Assessment.
Inventor includes Saumen Bandyopadhyay.
The application for the patent was published on October 23, 2026, under issue no. 43/2026.
Abstract: The invention provides a multisource fraud-intelligence system for coordinated merchant, account and transaction risk assessment. Timestamped transaction, merchant, account, identity, device and network events are resolved into canonical keys, quality-qualified and incrementally incorporated into a versioned heterogeneous temporal graph. A relation-aware graph-neural engine aggregates messages over typed, time-bounded neighbourhoods and generates distinct mer- chant, account and transaction risk evidence. A validity evaluator calibrates uncertainty and withholds predictive action authority for incomplete, stale, out- of-distribution or unauthenticated inputs. A policy engine combines valid graph evidence with deterministic transaction safeguards. A constrained interface dispatches a reversible or approval-gated instruction carrying an idempotent token after revalidating graph and transaction versions. A feedback engine verifies connector acknowledgement and subsequent source-system state before updating graph and risk states. The architecture reduces full-graph recomputation, repeated alerts and stale actions while preserving traceable fallback and governed model deployment.
Disclaimer: Curated by HT Syndication.
