Smart Anti-Theft Security System with Face Recognition Using Edge AI for Real-Time Access Control

Abstract

With the rapid increase in security threats such as unauthorized entry, theft, and tampering, the need for intelligent and reliable access-control systems has become more important than ever. Conventional security methods, including mechanical locks, passwords, and RFID-based systems, suffer from major limitations because they can be stolen, duplicated, or misused. To overcome these drawbacks, this paper presents a Smart Anti-Theft System based on Raspberry Pi Pico, TensorFlow Lite, OpenCV, Firebase Cloud, and Android application support. The proposed system is designed to provide a low-cost, fast, and privacy-aware security solution using real-time facial recognition as the primary authentication mechanism. The system captures the user’s face through a camera module and processes the image locally using a lightweight quantized Convolutional Neural Network (CNN) model deployed through TensorFlow Lite. By performing face recognition at the edge, the system reduces latency, improves privacy, and minimizes dependence on continuous cloud processing. If the detected face matches an authorized user, the system grants access by activating a solenoid lock. In contrast, if the face is unrecognized or unauthorized, the system denies access and immediately generates alerts and event logs. These logs are synchronized with Firebase Cloud, enabling real-time monitoring and secure data management. An Android companion application is integrated with the system to provide additional functionality such as user face enrollment, real-time security event monitoring, remote lock control, and alert viewing. The combination of embedded edge intelligence and cloud connectivity makes the proposed model efficient, modular, and suitable for practical deployment in homes, offices, laboratories, and other restricted environments. Experimental observations show that the system is capable of making quick authentication decisions and supporting smart anti-theft operations with improved security and usability.

Introduction

The system replaces weak conventional methods like keys, passwords, and RFID cards with facial recognition-based authentication, which is more secure and contactless. Unlike many existing systems that rely on cloud processing, this project performs real-time face recognition locally using a Raspberry Pi Pico, TensorFlow Lite, and OpenCV, which improves speed, privacy, and reliability.

The working process includes capturing a face image, detecting and preprocessing it, recognizing the user using a lightweight CNN model, and then making an authentication decision. If the user is authorized, the system unlocks a solenoid lock; otherwise, it triggers alerts, logs the event in Firebase, and notifies the user through an Android application.

The literature review shows that recent research supports the shift toward edge AI, lightweight models, privacy-preserving biometric systems, and anti-spoofing techniques, but also highlights challenges like spoofing attacks and environmental sensitivity.

In testing, the system performed well with around 90% recognition accuracy, 92% detection success, 0.4–0.6 sec response time, and strong lock reliability. It successfully handled real-time access control, cloud logging, and mobile monitoring, though performance decreased under poor lighting or improper face angles.

Overall, the project demonstrates a fast, low-cost, privacy-aware, and practical smart security system suitable for homes, offices, and restricted areas, while also highlighting areas for future improvement such as robustness and lighting adaptability.

Conclusion

The proposed Smart Anti-Theft System proves that an effective access-control solution can be built by combining facial recognition, edge AI, embedded lock control, Firebase cloud logging, and Android-based monitoring. Unlike traditional keys, passwords, or RFID cards, the system uses biometric identity for authentication, which improves security and reduces the chance of duplication, theft, or misuse. The use of local face recognition also makes the system faster and more privacy-aware, because critical recognition happens near the device instead of depending completely on remote cloud processing. The study also shows that integrating cloud logging with local decision-making is a strong practical approach for smart security applications. Local processing supports low-latency door access, while Firebase-style cloud connectivity enables alert storage, remote monitoring, and event history management through the mobile application. Based on the reviewed literature, this direction is consistent with current work in IoT security, embedded face recognition, decentralized biometric systems, and anti-spoofing-aware authentication. In conclusion, the project provides a strong foundation for a low-cost, intelligent, and scalable anti-theft system suitable for homes, offices, laboratories, and restricted areas. With future improvements such as liveness detection, better low-light handling, larger user enrolment, and stronger spoof resistance, the system can be developed into a more robust real-world smart security product

References

[1] R. Reaño, P. Carrión, and J.-P. Mansilla, “Access Control Using Facial Recognition with Neural Networks for Restricted Zones,” in Proc. 19th Int. Conf. Web Information Systems and Technologies (WEBIST), 2023, pp. 310–318, doi: 10.5220/0012185800003584. [2] G. Chaopeng, L. Zhengqing, and S. Jie, “A Privacy Protection Approach in Edge-Computing Based on Maximized DNN Partition Strategy with Energy Saving,” Journal of Cloud Computing, vol. 12, art. 29, 2023, doi: 10.1186/s13677-023-00404-y. [3] A. S. Mohammad, T. G. Jarullah, M. T. S. Al-Kaltakchi, J. Alshehabi Al-Ani, and S. Dey, “IoT-MFaceNet: Internet-of-Things-Based Face Recognition Using MobileNetV2 and FaceNet Deep-Learning Implementations on a Raspberry Pi-400,” Journal of Low Power Electronics and Applications, vol. 14, no. 3, art. 46, 2024, doi: 10.3390/jlpea14030046. [4] J. Guo, H. Mu, and C. Han, “Federated Learning for Biometric Recognition: A Survey,” Artificial Intelligence Review, vol. 57, art. 208, 2024, doi: 10.1007/s10462-024-10847-7. [5] B. V. Satish Babu, J. Manoranjini, R. Changala, et al., “Biometric-Based Access Control Systems with Robust Facial Recognition in IoT Environments,” in 2024 3rd Int. Conf. Intelligent Techniques in Control, Optimization and Signal Processing (INCOS), 2024, doi: 10.1109/INCOS59338.2024.10527499. [6] M. Marimuthu, G. Mohanraj, J. Akilandeswari, and V. Sathiyapriya, “Facial Recognition Enabled Smart Security Lock System Using Machine Learning Approach,” EAI Endorsed Transactions on Internet of Things, vol. 11, Jun. 2025, doi: 10.4108/eetiot.5657. [7] S. S. F. A. Elnozahy, S. C. Pari, and L. C. Liang, “Raspberry Pi-Based Face Recognition Door Lock System,” IoT, vol. 6, no. 2, art. 31, 2025, doi: 10.3390/iot6020031. [8] E. Abdullahu, H. Wache, and M. Piangerelli, “Secure and Decentralized Hybrid Multi-Face Recognition for IoT Applications,” Sensors, vol. 25, no. 18, art. 5880, 2025, doi: 10.3390/s25185880. [9] H. Xing, S. Y. Tan, F. Qamar, and Y. Jiao, “Face Anti-Spoofing Based on Deep Learning: A Comprehensive Survey,” Applied Sciences, vol. 15, no. 12, art. 6891, 2025, doi: 10.3390/app15126891. [10] Y. Li, Y. Li, and Z. Wang, “FaceCloseup: Enhancing Mobile Facial Authentication with Perspective Distortion-Based Liveness Detection,” Computers, vol. 14, no. 7, art. 254, 2025, doi: 10.3390/computers14070254. [11] M. Kira, Z. Alajamy, A. Soliman, Y. Mesbah, and M. Mazzara, “Trustworthy Face Recognition as a Service: A Multi-Layered Approach for Mitigating Spoofing and Ensuring System Integrity,” Future Internet, vol. 17, no. 10, art. 450, 2025, doi: 10.3390/fi17100450. [12] E. L. Birgisdóttir, M. I. Kunkel, et al., “Exploring the Security of Mobile Face Recognition: Attacks, Defenses, and Future Directions,” Applied Sciences, vol. 15, no. 24, art. 13232, 2025, doi: 10.3390/app152413232. [13] N. Zeeshan, et al., “Continuous Authentication in Resource-Constrained Environments,” Sensors, vol. 25, no. 18, art. 5711, 2025. [14] S. S. U. Hasan, et al., “A Review on Secure Authentication Mechanisms for Mobile and IoT Devices,” Sensors, vol. 25, no. 3, art. 700, 2025. [15] O. Korchenko, et al., “Modular Neural Network Model for Biometric Authentication of Personnel in Critical Infrastructure Facilities Based on Facial Images,” Applied Sciences, vol. 15, no. 5, art. 2553, 2025. [16] H. Sabit, et al., “Artificial Intelligence-Based Smart Security System Using Face Recognition and Internet of Things,” Electronics, vol. 14, no. 3, art. 608, 2025. [17] A. Nurpeisova, et al., “Deep Residual Learning for Face Anti-Spoofing,” Technologies, vol. 13, no. 9, art. 413, 2025. [18] J. Saleem, et al., “Machine Learning-Enhanced Attribute-Based Authentication for Smart IoT Environments,” Sensors, vol. 25, no. 9, art. 2779, 2025. [19] Y. Li, et al., “Face Anti-Spoofing Based on Adaptive Channel Attention,” Journal of Imaging, vol. 11, no. 4, art. 116, 2025. [20] H. Kim, et al., “Anti-Spoofing Method by RGB-D Deep Learning for Robust Face Recognition,” Electronics, vol. 14, no. 11, art. 2182, 2025. [21] B. Seyedi, et al., “Securing IoT Communications via Anomaly Traffic Detection,” Sensors, vol. 25, no. 13, art. 4098, 2025. [22] R. Klinowski, et al., “Face Spoofing Detection with Stacking Ensembles in Work Time Registration Systems,” Applied Sciences, vol. 15, no. 15, art. 8402, 2025. [23] M. H. A. Pratama, et al., “Advancing Secure Face Recognition Payment Systems: A Systematic Literature Review,” Information, vol. 16, no. 7, art. 581, 2025. [24] T. S. Matlala, et al., “A Convolutional Neural Network-Based Vehicle Security Framework with Vision-Based Authentication,” Applied Sciences, vol. 15, no. 19, art. 10584, 2025. [25] J. Al-Nabulsi, et al., “IoT Solutions and AI-Based Frameworks for Masked-Face Recognition: A Review,” Sensors, vol. 23, no. 16, art. 7193, 2023.

Copyright

Copyright © 2026 Prof. Satish C. Cholke, Dhonnar Dipali, Mahale Hemangi , Khalkar Vaishnavi , Salve Harshada . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.