A Study on Predicting Campus Tailgating Intrusion Using Data Fusion Technology Based on IoT and Cloud Computing in the Security Management of Smart Campuses at Private Universities

This study addresses the dynamic prediction and real-time warning of tailgating intrusion in smart campuses at private universities. We propose an integrated solution that combines IoT data acquisition, cloud-based data fusion, and LSTM time-series prediction. IoT devices deployed at key campus entrances and exits collect high-frequency real-time pedestrian data, which is then preprocessed through cleaning, normalization, time synchronization, and dynamic weight adjustment. Bayesian estimation and Kalman filtering fuse these multi-source data to construct high-quality state time series. An LSTM model subsequently captures subtle behavioral changes, enabling precise anomaly detection. Moreover, efficient real-time processing is ensured by a comprehensive monitoring and dynamic anomaly detection algorithm, with time and space complexities analyzed via sliding window statistics, extended Kalman filtering, and distributed task scheduling. Experimental results on the TON_IoT dataset indicate significant improvements in prediction accuracy, precision, recall, F1 score, and response time, as well as reductions in false alarms, missed detections, and training time. These findings offer novel theoretical and practical insights for enhancing security management in smart campuses and other public safety domains.