Federated Learning
Ensuring privacy in data exchange for IoT environments
Blockchain
Secures agent-to-agent transactions, enhances security and decentralizes
Smart Contracts
Defines clauses between agents, automates and provides transparency
Federated Learning
A new Machine Learning paradigm allows models to be trained in a distributed manner without centralizing the data.
A network made up of edge nodes, where each one is in charge of storing the data generated by itself and training a local model.
Subsequently, a central server aggregates the different local models generated, establishing a global model of better quality than the local ones and without compromising the private information of each edge, increasing security and reducing the possibilities of exposing the data.
This technology continues to assume the robustness of the central server, being vulnerable both to its own failures and to malicious users who try to corrupt the global model or extract information from the data through the shared models.
Fortunately, the additional risks involved in its use can be overcome by applying other technologies.
Blockchain
It allows decentralizing and strengthening Federated Learning processes through its consensus algorithms.
A technology that manages to decentralize the function of the central server among the different edge nodes, where each of the local models are shared with the rest of the nodes.
The nodes perform the aggregation of the models in the established way and propose a new block in the Blockchain chain.
The rest of the nodes can verify the proposed block and add it to the chain or reject it based on the established parameters.
The integration of this technology ensures the contribution and validation of the network nodes for the establishment of the new global model, increasing the robustness of the whole system and being tolerant to the failure of some network nodes.
Smart Contracts
Smart contracts allow the establishment of a code to be executed when certain conditions are met.
Smart contracts make it possible to generate mechanisms that encourage the participation of the different nodes, granting a series of benefits to those who contribute positively or punishing those who want to misuse the correct functioning of the system.
This technology gives the different parties involved greater confidence when participating in the network and ensures the correct inclusion of their data, as it establishes clear, immutable and transparent rules.