The accurate detection of network intrusions is essential for protecting digital infrastructures from cyber threats. Traditional machine learning models, such as Random Forest (RF), Decision Tree (DT), AdaBoost, XGBoost, Logistic Regression, Support Vector Machine (SVM), and K-Nearest Neighbors (KNN), have shown promising performance in intrusion detection tasks. However, these models often produce uncalibrated probability estimates, reducing their reliability in critical decision-making scenarios. This paper introduces an enhanced approach that combines three calibration techniques namely Platt Scaling, Isotonic Regression, and Beta Calibration with these ML models to improve the reliability and interpretability of predicted probabilities. The calibrated models outperformed their traditional counterparts in terms of key metrics, including accuracy, precision, recall, and F1 score. Experimental results demonstrate that calibration not only refines prediction probabilities but also enhances overall performance, making these models more effective for real-world intrusion detection.