A Federated AI and DAG-Based Framework for Secure and Scalable E-Voting

In response to the growing need for secure, scalable, and transparent election processes, this study explores an advanced modular e-voting architecture thatleverages the limitations of traditional and blockchain-based voting systems. While some previous systems eliminated many of the challenges of election fraud and manipulation via immutability and decentralization, they are still riddled with high transaction costs, scalability limits, centralized biometric storage, and the lack of a verifiable audit trail. This paper introduces a new framework that integrates several novel technologies, such as Decentralized Identity (DID), Federated Learning (FL), Zero-Knowledge Proofs (ZKP), Homomorphic Encryption, and Directed Acyclic Graph (DAG)-based Distributed Ledger Technology to realize a secure, private, and tamper-evident digital voting solution. The architecture is composed of four important phases, namely: voter onboarding using DIDs and verified credentials, AI-based eligibility assessment using on-device federated agents, privacy-protected vote casting using ZKP and encryption, and verifiable homomorphic tallying during post-election aggregation. Paired with the ability to achieve near-instant finality and zero-cost submissions, this approach addresses one of the major drawbacks of traditional blockchain used as a voting system, the complexity of smart contracts and gas fees, using DAG-based networks like IOTA or Hedera.