Design and Implementation of a Cost-Effective, AI-Enabled Smart Security System for Real-Time Threat Detection in Nigerian Rural Communities

Abstract

This study presented the design and implementation of a cost-effective, AI-enabled smart security system tailored for real-time threat detection in Nigerian rural communities. The system integrated an ESP32-CAM microcontroller, a Passive Infra-Red (PIR) Camera Sensor Module, and a Global System for Mobile Communications (GSM) module, employing a quantized MobileNet model for edge-based classification and Short Message Service (SMS) alert transmission without internet dependency. The methodology involved developing a lightweight MobileNet model trained on a 226-sample dataset, achieving a 97.35% training accuracy, and simulating system performance with 30 test instances, resulting in a 93.3% classification accuracy. The key performance metric included a 0.5-second motion-to-capture time, 60–77ms classification time (post-1s load), and an average 3.2-second SMS alert delivery. Results demonstrated robust threat distinction between threats and non-threats, and was validated by a confusion matrix showing 9 True Positives, 19 True Negatives, 1 False Positive, and 1 False Negative. The system, with an estimated cost of less than ₦30,000, was shown to be more affordable and adaptable to resource-constrained areas than many of the current alternatives. The system's performance confirms its potential as a scalable, intelligent security solution for rural Nigeria, despite the fact that it is simulation-based, which restricted real-world environmental testing.