Software Defect Prediction (SDP) is crucial for enhancing software quality and minimizing issues after release. The advent of machine learning, particularly in Cross-Project Defect Prediction (CPDP), has garnered significant attention for its potential to enhance defect predictions in one project by leveraging information from another. A critical factor influencing CPDP effectiveness is feature selection, the process of identifying the most relevant features from an available set. This review article thoroughly examines the role of feature selection in CPDP. Existing feature selection methods are systematically analyzed and classified within the CPDP context, encompassing both traditional and state-of-the-art approaches. The review delves into the challenges and opportunities presented by diverse project characteristics, data heterogeneity, and the curse of dimensionality. Additionally, the article underscores how feature selection impacts model performance, generalization, and adaptability across various software projects. Through synthesizing findings from multiple studies, trends, best practices, and potential research directions in the field are identified. In conclusion, this review article provides valuable insights into the significance of feature selection for enhancing the reliability and efficiency of CPDP models.