Wireless Sensor Networks (WSNs) play a vital role in diverse technological environments, facilitating data collection across fields like industrial automation and environmental monitoring. Their widespread use underscores the importance of robust security measures, particularly for intrusion detection. This study explores leveraging machine learning techniques, including Random Forests, XGBoost, and k-Nearest Neighbors (k-NN), to enhance intrusion detection system precision and effectiveness in WSNs. It begins by elucidating the limitations of traditional intrusion detection approaches, emphasizing the constraints of anomaly detection and signature-based methods. Through comprehensive evaluation using metrics like F1 score, accuracy, precision, and recall, we compare the performance of Random Forests, XGBoost, and k-NN. The study utilizes a dataset derived from network traffic patterns, incorporating energy-conscious features critical for WSN longevity. A detailed review of prior research on WSN attack detection highlights both shortcomings and the unique contributions of our investigation. By integrating customized feature selection with machine learning for intrusion detection, we aim to bolster WSN resilience against various security risks and provide valuable insights for practitioners, researchers, and stakeholders invested in WSN dependability and security.