Adithya Manthena

Decentralized and Secured Machine Learning through Blockchains.

Machine learning has become an integral part of our day-to-day lives in recent years with applications ranging from phone cameras to cancer diagnosis. Despite this the creation of the model is majorly performed in a centralized environment where the data is present. This is not ideal for scenarios where multiple pockets of data are spread across a network that cannot be shared due to privacy concerns. We propose a Decentralized and secured machine model learning through Blockchains. Blockchain facility secures transfer of the model across the network with validation through consensus. In our framework multiple peers train the model locally using their data and this trained model is shared across the network for validation. Each peer would validate the updated model in their own system and the update is accepted if maximum peers accept the newer model.