Privacy-Preserving AI Framework for Child Suspicious Activity Recognition With Parental Control and Digital Protection
Vinod Kumar Mahor, Jaytrilok Choudhary, Dhirendra Pratap Singh · 2025
Ensuring child safety in more and more linked digital contexts provides a double difficulty: properly monitoring questionable behavior while maintaining user privacy and allowing adaptive parental control. Facilitating real-time identification of abnormalities in children's digital activities, the framework uses a hybrid CNN-BiLSTM model to precisely capture both spatial and temporal behavioral patterns. hence addressing important privacy issues., hence delivering real-time notifications and risk evaluations customised to the environment and intensity of the identified actions. Achieving 95% accuracy and good precision, recall, and F1-score, experimental tests using the Kinetics-700 dataset confirm the efficacy of the suggested model. With improved computational efficiency appropriate for real-time applications, the model shows better performance than traditional methods. The inclusion of privacy-preserving technologies does not noticeably affect performance, hence stressing the framework's appropriateness for use in actual digital platforms.