With the prevalence of phishing attacks and the resulting consequences, there is a need to accurately detect and prevent such attacks. The aim of this study is to evaluate whether a nature-inspired optimization algorithm boosts the ability of ensemble learning models (the Random Forest & Gradient Boosting classifiers) to detect phishing attacks in cloud-based systems. Data used in this study was obtained from the Mendeley repository. The data was pre-processed in Python version 3.9. Two ensemble models – the Random Forest (RF) and Gradient Boosting (GB) classifiers – were trained on the pre-processed data to obtain generalization accuracies, precision scores, and recall rates. Feature selection was then conducted using the Grey Wolf Optimizer. Additionally, four objective functions were defined to evaluate model fitness: the Logistic Regressor (LR), RF classifier, GB classifier, and Decision Tree (DT) classifier. This optimization process yielded four datasets that were used to train and validate the RF and GB classifiers. Results demonstrated that the RF and GB classifiers had relatively lower recall rates (<97%) compared to the optimized RF and GB classifiers (>97%). These results indicate that using nature-inspired optimization algorithms for feature selection improves the performance of ensemble learning algorithms.