Fraud intelligence generated by one Kenyan bank or mobile money provider is rarely shared with peer institutions in a timely, trusted, and auditable way, leaving fraud schemes that move across institutional boundaries only partially visible to any single organisation. This study designs, implements, and evaluates a consortium blockchain framework that lets banks and mobile money providers submit, validate, share, and retrieve fraud intelligence within a permissioned, auditable environment. The framework is realized as a Hyperledger Fabric prototype spanning five simulated organisations, combining organisation-specific cryptographic identities, mutual TLS, chaincode-enforced access control, an append-only audit trail, and client-side hashing of account identifiers. Functional, security, and performance testing, using synthetic fraud-intelligence data and workloads of 50 and 100 transactions, show that the prototype supports the full fraud-alert lifecycle with a 100 percent transaction success rate, chaincode-enforced confidentiality, and a fully traceable audit trail, while transaction-volume performance improves rather than degrades over the tested range (mean response time falling from 390.5 ms to 304.0 ms). Adversarial testing confirms that identity-spoofing and unauthorized-access attempts are rejected, though a ledger-tampering test exposes a boundary condition in post-commit corruption detection. The main contribution is a working, empirically tested demonstration that permissioned consortium blockchain is a technically viable mechanism for controlled, auditable fraud-intelligence sharing among competing financial institutions, distinct from, and complementary to, institution-level fraud detection itself.