This paper presents a privacy-preserving scheme for checking passwords against breach databases using obfuscated deterministic Bloom filter indices. The approach addresses critical limitations in existing methods: k-anonymity schemes leak partial hash information that can increase attack success rates by an order of magnitude, while cryptographic protocols like Oblivious PRF (OPRF) impose significant computational overhead. The proposed scheme employs deterministic noise generation to obfuscate Bloom filter queries, preventing servers from inferring passwords while maintaining the efficiency of Bloom filter lookups. Analysis demonstrates that deterministic noise provides superior privacy guarantees compared to random noise, particularly against correlation attacks over multiple queries. Experimental evaluation shows that the scheme achieves sub-millisecond query times with minimal bandwidth overhead (typically under 1KB per query) while providing strong privacy guarantees. The proposed method is particularly suitable for integration into password managers, authentication systems, and enterprise security infrastructure.