Critical National Infrastructure (CNI) sectors, including energy, water, and transportation, are increasingly interdependent, yet they remain siloed by centralized Identity and Access Management (IAM) systems that act as single points of failure. This research proposes a novel Federated Identity and Access Management (FIAM) framework that establishes cross-sector trust through a hybrid architecture combining Blockchain-based Decentralized Identity (DID) and Federated Learning (FL). By utilizing a permissioned blockchain, the framework eliminates the reliance on central certificate authorities, providing an immutable and transparent ledger for Verifiable Credentials (VCs). To address the "static trust" limitation of traditional blockchain systems, a Federated Learning layer is integrated to perform real-time anomaly detection. This layer enables individual CNI sectors to collaboratively train a global threat detection model on sensitive access logs without exposing raw data, thus preserving operational privacy while enhancing collective security. Experimental evaluation conducted via a Hyperledger Fabric and Flower (FL) prototype demonstrates that the framework maintains a low authentication latency of <3 seconds and achieves a 98.8% accuracy in detecting identity-based spoofing and lateral movement attacks. The results indicate that the proposed "Trust-but-Verify" logic effectively balances the high-availability requirements of CNI with the need for a decentralized, privacy-preserving security posture. This work provides a scalable blueprint for resilient, cross-sector identity ecosystems capable of defending against sophisticated, multi-stage cyber threats in national infrastructure.