This study explores the use of Channel State Information for biometric authentication, focusing on addressing the challenges posed by environmental variations. To achieve this, experiments were conducted using off-the-shelf ESP32 devices to collect CSI data across different environments, including urban, suburban, and rural settings. The primary objective was to analyze the influence of external environmental factors on the accuracy of CSI-based biometric systems and to develop methods to mitigate these effects. The significant subcarrier selection method was combined with a weighted Random Forest classifier to improve the system's performance. The results demonstrated that certain subcarriers are more sensitive to environmental changes, and by assigning different weights to these subcarriers the authentication accuracy improved to 93.33%. These findings highlight the potential of CSI-based biometrics to offer reliable and environment-independent authentication, making them suitable for real-world applications in dynamic settings, such as smart homes and vehicular systems. This research lays the groundwork for further studies aimed at developing more resilient biometric systems capable of operating effectively across diverse environments.