MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641106318 A) filed by Dr. G. Palanikumar; Dr. Sivasundar M; Dr. T. Thirumalaikumari; Dr. H. Vennila; Ms. Balambiga G; Ms. Devika M; Ms. Keerthana R; and Dr. S. Nithya on September 03, 2026, for Novel Integrated Ai, Edge Computing And Cybersecurity Framework For Next-Generation Intelligent Systems.
Inventors include Dr. G. Palanikumar; Dr. Sivasundar M; Dr. T. Thirumalaikumari; Dr. H. Vennila; Ms. Balambiga G; Ms. Devika M; Ms. Keerthana R; and Dr. S. Nithya.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: The invention disclosed herein teaches the integration of artificial intelligence, edge computing, and cybersecurity into a unified architecture for next generation intelligent systems. The architecture is suitable for deploying perception, reasoning, actuation, and protection at the edge of a factory, energy grid, transportation corridor, healthcare facility, city infrastructure, or other domain where decisions must be made in close proximity to the source of sensing, within limits of latency, bandwidth, energy, and trust. In current practice, these three domains are treated as separately designed layers, training models in distant facilities and then copying them to edge nodes, and placing security controls outside, rather than inside, the decision loop. Edge nodes are too often passive, simply collecting sensor data, rather than collaborating with other elements at the edge in a continuous reasoning loop. This results in unacceptable response time for control loops, widened attack surface at every hop, out-of-date models that cannot adapt to changing dynamics and adversaries, and security decisions made without operational context, leading to excessive false positives or dangerous oversights.The invention enables a closed loop of sensing, local inference, policy reasoning, cryptographic protection, identity assurance, and coordinated response, executing primarily on a fabric of edge nodes with selective collaboration to regional and central resources when risk, complexity, or governance dictates. Each edge node is a computational and protective element collocated with sensors, actuators, controllers, gateways, or user devices. A light weight runtime consumes heterogeneous signals and projects them to a common observational schema that carries process measurements together with identity attributes, session context, device health, software integrity evidence, communication patterns, and environmental conditions. The same schema provides both operational state estimates and trust estimates, making security a first class input to reasoning. Fast, local models provide event classification, short horizon prediction, anomaly detection, risk scoring, and candidate actions, all under stringent processors limits and intermittent connectivity. These models are refined by slower regional or central level models. Candidate replacements are only admitted after tests of safety, energy cost, latency, and adversarial robustness, over an attested channel, and with the possibility of rollback of unsafe changes if later observations show harm. A context aware policy engine evaluates each proposed action against standing rules, neighborhood summaries, requester identity, path, operational mode, integrity evidence, and predicted physical consequence. Protection of data in motion and data at rest is cryptographic, with algorithms and key lifetimes selected to match the sensitivity of assets and the computational budget of the node. Identity is continuous rather than episodic, enabling privilege revocation in near real time between collaborating machines. Neighboring nodes share compact, privacy preserving summaries rather than raw records, so that confidence increases when multiple observers report related anomalies, and operations continue when wide area connectivity is lost. Actions are taken from a constrained catalog that includes sensing adjustments, model replacement, cryptographic reconfiguration, privilege change, network containment, operator notification, and physical actuation, and are only executed after a safety interlock confirms declared limits. Outcomes are added to local history, models, and policies. Provenance and accountability are carried with every model, rule, and cryptographic configuration, and decisions with safety or privacy impact can emit an explainable trace. Retention and erasure are enforced at the edge.The architecture keeps time critical decisions and first line of defense at the edge of the network, limits export of raw streams, operates in isolation, and treats security as a continuous control problem rather than a periodic audit. Next generation intelligent systems can acquire, reason, act, and protect in the same moment, at the same place, with lower latency, frugal communication, adaptive trust, and resilience to accident as well as attack.
Disclaimer: Curated by HT Syndication.
