Wireless Sensor Networks (WSNs) are prone to different security threats because of their open communication environment, distributed architecture, and resource constraints. For the security and integrity of a network to be ensured, robust intrusion detection systems (IDS) are required. The WSN-DS dataset has been used to provide an effective machine learning (ML) and Deep Learning (DL) based IDS and attack detection technique for WSNs. Several learning models, including K-Nearest Neighbors (KNN), Random Forest (RF), Decision Tree (DT), Convolutional Neural Networks (CNN), Support Vector Machine (SVM), Logistic Regression (LR), and Neural Networks (NN), are compared in terms of performance. Preprocessing methods of data encoding, normalization, and data splitting are applied to the dataset to improve classification performance. The effectiveness of these models is compared using large trials with significant performance metrics such as ROC-AUC, F1-score, accuracy, precision, and recall. The results indicate that the Optimized RF model has been optimized to achieve the optimal accuracy of 99.71\%, which outperforms other state-of-the-art approaches. Apart from pointing out the importance of ML in detecting WSN attacks, this research provides a promising way forward for enhancing network security through effective detection methods.