MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641026399 A) filed by Sri Sairam Engineering College, Chennai, Tamil Nadu, on March 6, for 'smart parental control using machine learning.'
Inventor(s) include Barathkumar P; Baranidharan R; Madhankumar S; Niranjan R; and Valarmathy G.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "The present invention discloses a comprehensive, AI-driven, real-time Smart Parental Control System engineered to safeguard minors from exposure to inappropriate, explicit, or harmful digital content without disrupting the usability of the underlying application or browser. Conventional content filtering systems predominantly rely on URL denylists or DNS blocking, which often fail to filter dynamically loaded in-app content, user-generated media, and embedded advertisements, leading to the "glimpse effect" and abrupt, frustrating application closures. To overcome these critical limitations, the disclosed invention operates entirely via Edge Artificial Intelligence directly on the user's device. The system comprises a background interception module that continuously captures screen rendering frames and simultaneously routes them through two localized, GPU-accelerated parallel processing pipelines. The first pipeline utilizes an Optical Character Recognition (OCR) engine to dynamically extract and analyze on-screen text, cross-referencing it against customizable dictionaries to detect cyberbullying, profanity, and sensitive keywords. The second pipeline employs a lightweight, highly optimized Convolutional Neural Network (CNN) to evaluate pixel data and classify explicit, violent, or age-inappropriate visual imagery in real time. Upon positive detection of restricted content exceeding predefined confidence thresholds, a central decision-making engine calculates the precise spatial bounding box coordinates of the offending element. A real-time selective rendering engine subsequently generates and overlays a targeted visual blur mask strictly over the identified coordinates. This selective occlusion ensures that the remainder of the digital interface remains fully visible and interactable, thereby preserving the educational or entertainment experience. Furthermore, by executing all image and text processing locally on the edge device, the system guarantees zero transmission of sensitive raw user data to cloud servers, ensuring strict compliance with stringent privacy frameworks such as GDPR and COPPA. The architecture is augmented by an adaptive federated learning framework integrated with a centralized caregiver dashboard. Caregivers can monitor anonymized incident logs, customize sensitivity parameters, and provide feedback on false positives or negatives. This feedback triggers local model refinements, transmitting only encrypted mathematical weight updates to a central aggregator, continuously enhancing global detection accuracy while maintaining absolute user privacy."
Disclaimer: Curated by HT Syndication.
