Abstract
Protecting residential properties through CCTV surveillance has become a crucial aspect of contemporary living. In the age of data-driven security, detecting unusual and infrequent patterns within CCTV footage is of significant importance. Conventional anomaly detection methods often need help to handle the intricacies and distinctiveness of home environments, where family members and potential intruders co-exist. To address these challenges, this research incorporates extensive datasets, various machine-learning techniques, and thorough evaluation measures to pinpoint rare patterns that might signal threats to residential premises. These findings provide valuable insights for homeowners, security professionals, and developers working on surveillance systems.