Artificial intelligence (AI) has successfully transformed various fields through automation, yet it continues to struggle in some fields, particularly in the medical field. It is due to the lack of large, accessible datasets, because medical institutions often limit data sharing to ensure the confidentiality and safety of critical patient data. An application for multi-class brain tumor classification that uses a privacy-preserving Federated Transfer Learning (FTL) platform is proposed to get over these ethical and privacy concerns. It integrates the TL with the privacy-preserving FL. Within our system, the exchange of information among various healthcare institutions does not need the direct sharing of raw patient data, hence ensuring the protection of patient privacy. In the federated learning context with two transfer learning models, VGG16 and VGG19, the accuracies are 93.11% and 91.8%, respectively. The results of our study indicate that privacy can be maintained without compromising performance and emphasize the potential of our proposed FTL framework to safeguard privacy in healthcare systems that are propelled by AI.