Cyber-Physical Systems (CPS) represent a critical class of intelligent systems where physical processes are tightly integrated with computational and communication capabilities. Their role in enhancing security in public life is particularly relevant when combined with artificial intelligence techniques for real-time recognition of people, objects, and potential hazards. This paper presents a comparative experimental study of two CPS modules: face recognition and object recognition. The experiments are conducted on two computational platforms with different hardware capabilities: Gigabyte GA-A320M-H and ASUS ROG Zephyrus M16. Both modules are implemented using deep learning models, trained and evaluated under identical conditions. Performance was assessed through multiple metrics, including accuracy, F1-score, training and validation loss, and computational efficiency. The comparative results highlight significant differences in execution time and convergence behavior between the two machines, while maintaining stable recognition accuracy. The analysis provides insights into the trade-offs between hardware configurations and algorithmic performance, offering guidelines for deploying CPS modules in real-world security-critical scenarios. Future work will extend the comparative analysis to other CPS modules, such as dangerous objects and smoke recognition, in order to provide a broader evaluation of system reliability and scalability.