DAO-FL: Enabling Decentralized Input and Output Verification in Federated Learning with Decentralized Autonomous Organizations

Federated Learning (FL) has emerged as a decentralized machine learning paradigm that facilitates collaborative training of a global model (GM) across multiple devices while maintaining data privacy. Traditional FL systems suffer from centralized validation of local models and GM updates, compromising transparency and security. In this paper, we propose DAO-FL, a smart contract-based framework that leverages the power of Decentralized Autonomous Organizations (DAOs) to address these challenges. DAO-FL introduces the concept of DAO Membership Tokens (DAOMTs) as a governance tool within a DAO. DAOMTs play a crucial role within the DAO, facilitating members' enrollment and expulsion. Our framework incorporates a Validation-DAO for decentralized input verification of the FL process, ensuring reliable and transparent validation of local model uploads. Additionally, DAO-FL employs a multi-signatures approach facilitated by an Orchestrator-DAO to achieve partially decentralized GM updates, and thus decentralized output verification of the FL process. We present a comprehensive system architecture, detailed execution workflow, implementation specifications, and qualitative evaluation for DAO-FL. Evaluation under threat models highlights DAO-FL's out-performance against traditional-FL (FedAvg), effectively countering input and output attacks. DAO-FL excels in scenarios where decentralized input and output verification are crucial, offering enhanced transparency and trust. In conclusion, DAO-FL provides a compelling solution for FL,  reinforcing the integrity of the FL ecosystem through decentralized decision-making and validation mechanisms.