Surveillance and controlled access environments are increasingly built around cyber-physical systems (CPS), where secure and dependable identity validation is no longer optional but essential. Authentication in such systems is still largely based on passwords, cards, and tokens. While these mechanisms are convenient, their reliability is frequently compromised by loss, duplication, and misuse. These limitations become more pronounced in open CPS environments that involve repeated and direct human interaction. As a result, biometric authentication has emerged as a practical alternative, as it links access control directly to inherent physiological characteristics and reduces dependence on transferable credentials.This paper provides an application-oriented evaluation of a lightweight biometric authentication module with a specific focus on deployment within cyber-physical security systems. The work does not aim to propose a new biometric descriptor or recognition algorithm. Instead, it examines whether simple and computationally efficient biometric methods can realistically operate as security modules under CPS constraints. Texture-based biometric features are employed together with classical machine learning classifiers to maintain a balance between recognition reliability and limited computational resources. Performance is evaluated using benchmark datasets under conditions that reflect real CPS operation. The results show that stable authentication performance can be achieved with minimal computational overhead. These findings indicate that compact biometric authentication modules can be practically deployed in resource-constrained cyber-physical systems to support smart surveillance and access-control applications without relying on complex deep learning architectures.