<![CDATA[The increasing risk of property crime highlights the need for home security systems capable of detecting potential threats and providing information to users remotely. This study aims to develop an Internet of Things (IoT)-based home security system using ESP32-CAM, a Passive Infrared (PIR) sensor, face detection and recognition, Telegram notifications, and remote buzzer control. The system was developed using a Research and Development (R&D) approach with a prototyping model. The developed system employs the Haar Cascade Classifier for face detection and Local Binary Pattern Histogram (LBPH) for face recognition. System testing included functional testing, face recognition testing, distance and lighting variation testing, Telegram image notification testing, and remote buzzer control testing. Face recognition performance was evaluated using five registered identities, with 100 images collected for each identity and 10 final recognition trials conducted for each registered identity. The results showed an overall face recognition success rate of 80% across 50 final recognition trials. Testing with three unregistered subjects resulted in a correct rejection rate (CRR) of 86.67% and a false acceptance rate (FAR) of 13.33% from 30 trials. The highest recognition success rate of 100% was obtained at a distance of 40 cm, while recognition performance generally declined beyond 40 cm with fluctuations at several tested distances, reaching 0% at 180 cm. Under varying lighting conditions, the system successfully detected faces at illumination levels of 312–4679 lux, while detection was unsuccessful at 87 and 136 lux. For unrecognized-face events, five tests were conducted to evaluate image notification delivery through Telegram, and all notifications were successfully delivered within 4.7–8.0 s. Remote buzzer control also operated successfully, with response times ranging from 3.5–7.4 s. These results indicate that the developed prototype successfully performed its intended home security functions within the evaluated conditions. However, the evaluation was limited by the number of subjects, repeated trials, and testing conditions. Further research should involve larger and more diverse subject groups and broader environmental and network conditions to improve the reliability and generalizability of the system.]]>
