The detection of malware is a critical task in cybersecurity, requiring efficient classification of malicious and benign data. This paper utilized a dataset comprising 1,00,000 samples with balanced class labels to explore the impact of feature selection techniques on malware detection. Initially, six machine learning models, including Logistic Regression, AdaBoost, Decision Trees, Random Forest, KNN, and XGBoost, were trained with all features. Later, five feature selection methods namely ANOVA F-Test, SelectKBest, Principal Component Analysis (PCA), Random Forest Feature Importance, and Mutual Information were employed to identify the most relevant features. The same ML were trained and evaluated using the selected features and observed the enhancement of performance for malware detection. Performance metrics such as precision, recall, F1-score, accuracy, and AUC were calculated, and ROC curves were analyzed for each model. It is observed that applying feature selection significantly improved classification accuracy and other metrics across all models, with Random Forest and XGBoost consistently achieving the highest performance. This work emphasizes the crucial role of feature selection in optimizing malware detection pipelines and offers a solid foundation for future research in this field.