Distributed Denial of Service (DDoS) attacks are still one of the focal cybersecurity concerns, primarily due to their complex nature and the sheer number of devices attached. Traditional detection lags in keeping up with emerging attack behaviors, and machine learning-based models have difficulty handling high-dimensional data, like the CICIDS2019 dataset with over 80 features. Feature selection (FS) has thus emerged as an essential preprocessing step to boost detection performance, decrease the cost of computation, and facilitate improved interpretation of the model. In this current study, the GreyLag Goose Algorithm (GGA), a bio-inspired metaheuristic optimization algorithm derived from greylag geese's migratory and foraging behavior, is utilized for the first time to solve the problem of feature selection in DDoS detection. GGA's balanced exploration-exploitation allows effective exploration of high-dimensional spaces of features to find the most informative subset of features. The extracted features were evaluated using the CICIDS2019 dataset with Support Vector Machine (SVM) and Random Forest (RF) classifiers. Experimental results show that GGA reduced the dimensionality from over 80 to 29 features with better classification performance. RF achieved closely perfect with 99.95% accuracy and AUC of 0.9999, although slightly better than SVM's 99.47% accuracy and 0.9987 AUC value. The findings underscore the strength of GGA in feature selection optimization of DDoS detection to achieve an efficient and stable approach that is a balance between efficacy and high prediction accuracy.