Exploring the Vulnerabilities of Federated Learning: A Deep Dive into Gradient Inversion Attacks