Federated learning (FL) is a decentralized machine learning approach where devices train models on their data without directly sharing it. This ensures privacy by transmitting only encrypted updates of the model to a central server for aggregation. FL offers advantages like stronger privacy, lower communication costs, and working with diverse data sources. However, FL can still be vulnerable to attacks compromising privacy, disrupting model performance, or weakening system security. Researchers are developing methods to address these security and privacy concerns. This survey explores current FL techniques and known security and privacy issues with their mitigation strategies. Finally, it highlights the challenges and future directions for ensuring user privacy and model effectiveness, which is essential for the broader adoption of FL.