Regarding this aspects, there are a lot of frameworks that still don’t have good documentation and tutorials. Among the latter, there are: PySyft, OpenFL and FedML. PySyft is still under construction, as the official repos- itory repository says, and for that reason often the documentation is not up to date and is not complete. OpenFL, on its side, has very meager documentation and only a few tutorials that don’t explore a lot of ML frameworks or a lot of scenarios. The FedML framework also has, like PySyft, incomplete documentation because the project is born very recently and is still under development. Finally, the FATE framework has a complete and well-made documentation but very few tutorials and, because of its complex architec- turearchitecture, would have taken too much time. Because of these reasons, these four frameworks were discarded from the comparison.

In the context of FL, the significance of a framework being ready for commer- cial commercial use cannot be overstated. As businesses increasingly recognise recognize the value of decentralised decentralized ML solutions, the demand for robust and production-ready frameworks has intensified. A FL framework geared for commercial use offers several crucial ad- vantagesadvantages. Firstly, it provides a stable and scalable foundation to deploy large-scale FL systems across diverse devices and platforms. This ensures that businesses can seamlessly integrate the framework into their existing infrastructure, minimising minimizing disruption , and optimising optimizing efficiency. Moreover, a commercially viable framework emphasises emphasizes security and pri- vacy privacy measures, a non-negotiable aspect when dealing with sensitive data across distributed environments. Advanced encryption techniques, secure communication protocols, and differential privacy methods guarantee that user data remains safeguarded, mitigating potential risks of data breaches or unauthorised unauthorized access. Of the frameworks covered, only a few are ready to be used commer- ciallycommercially. Among them are: Flower, NVFlare, FATE, OpenFL , and TFF. On the other hand, there are some frameworks that are not yet ready. Among the latter are: FedML, PySyft and IBMFL. The first two are in fact still under development and not yet ready for commercial-level use, while the third has a private license that does not allow the framework to be used for commercial-level applications. As explained in the previous sub-sections, these frameworks were already discarded from comparison.

analysis, six frameworks were deemed unsuitable due to some requisites not being met. The requirements that were met and not met by considered are summarized in the framework were summarised in Table 3.1following table: