With the rapid growth of Smart Home users, reaching 360 million as of 2023, security has become a primary concern for modern households, as evidenced by the widespread adoption of Doorbell Cameras. Although companies like Ring, Nest, and Arlo have successfully integrated IoT technology into their products, these camera-based solutions exhibit limited intelligence and struggle in low-power or no-network environments. To address these challenges, we propose a novel Smart Doorkeeper system, a non-imaging visitor monitoring method that leverages UWB, audio, haptic, and motion sensors to assess visitor risk. This paper provides a detailed examination of the doorkeeper system, proposes a novel architecture for a risk assessment framework, and evaluates multiple classifier models to identify risks associated with visitor interactions. Using intelligently extrapolated datasets mirroring real-world scenarios, we enhance home security and further introduce a human-in-the-loop regression feedback model to refine risk assessments. By leveraging advanced machine learning techniques, we optimize various parameters and deliver enhanced security. Our extensive research demonstrates a remarkable accuracy of 92% in assessing visitor interaction risks, surpassing conventional camera-based solutions that rely solely on real-time threat detection. The proposed Smart Doorkeeper solution, a pioneering invention in Smart Home security, has been filed as a patent, with both provisional and complete specifications submitted to the Indian Patent Office.