Performance Evaluation of SGX-hardened Federated Learning

The proposed study, being an analysis of a concrete implementation of Federated Learning, is indirectly placed in the field of artificial intelligence. The importance of the analysis carried out derives from the ever-increasing need to process information safely and in the context of machine learning. The challenge undertaken is to be able to evaluate the interaction between federated learning with the Gramine-SGX framework, obtaining an accurate analysis of which problems and needs we might encounter in a concrete implementation of these technologies. To achieve these ob- jectives, an application was implemented that allowed it to be modular and easily configurable. The design and implementation phase preceded the testing phase. the testing phase was consecutive to the development phase. The tests carried out fo- cused on the performances achieved by the various configurations and conceptually, we could divide them into preliminary tests and detailed tests. The preliminary tests had the task of providing clues to establish which aspects deserved in-depth examination. The detailed tests, however, made it possible to refute or confirm the hypotheses formulated in the previous phases.