Post-Quantum Lattice-Based Cryptographic Schemes for Long-Term Data-at-Rest Protection in Enterprise Database Systems

Enterprise databases routinely hold records — financial ledgers, electronic health records, genomic biobank entries, government archives — whose confidentiality must survive for decades, yet the asymmetric key-wrapping operations that protect them today (RSA and elliptic-curve key exchange inside transparent data encryption, TDE) are efficiently broken by a sufficiently large quantum computer running Shor's algorithm. Published resource estimates for that attack have fallen roughly three orders of magnitude in thirteen years — from approximately 10^9 physical qubits in 2012 to under 10^6 in 2025 — and the harvest-now-decrypt-later (HNDL) threat means an adversary can exfiltrate encrypted backups today and decrypt them once such a machine exists. A record's effective confidentiality horizon is therefore bounded by whichever comes first: its retention obligation, or the arrival of a cryptographically relevant quantum computer (CRQC). This paper studies that boundary at the key-management layer specifically, since bulk symmetric encryption (AES-256) already resists Grover's algorithm and is out of scope. We present a lattice-based post-quantum envelope-encryption architecture that replaces the RSA/ECC key-wrapping tier of a conventional TDE key hierarchy with a hybrid ML-KEM/classical key-encapsulation combiner, authenticates key-rotation metadata with ML-DSA or FN-DSA signatures over a hash-chained audit log, and negotiates between classical-only, hybrid, and pure post-quantum modes under policy control. The design is grounded in the Module-LWE and NTRU-lattice hardness assumptions behind NIST's finalized ML-KEM (FIPS 203) and ML-DSA (FIPS 204) and the still-draft FN-DSA (FIPS 206); we give seven explicit algorithms — hybrid encapsulation and decapsulation, enterprise-scale rekeying, signed audit and verification, crypto-agility negotiation, and exposure-ranked migration scheduling — with a composed security argument and an analytical cost model calibrated against published lattice-scheme parameters. That model puts the hybrid combiner's overhead at a modeled 9.0% added latency per key-rewrap relative to an all-classical baseline (12% at the highest NIST security level); shows full-estate rekeying of ten million keys falling from over ten hours at one worker to under ninety seconds at 512 parallel workers; and projects a signed, ML-DSA-authenticated ten-year audit trail for one thousand tenants at roughly 400 MB — about thirteen times an ECDSA baseline — a gap FN-DSA's compact, still-draft signatures narrow to roughly 100 MB. Four illustrative case studies — a 30-year financial ledger, a 50-year health/device archive, a 60-year genomic biobank, and a multinational deployment under CNSA 2.0 and EU deadlines — show migration completing in well under an hour of modeled compute for every estate, yet the two longest-horizon records would carry 41 to 57 years of residual exposure under Mosca's inequality if that migration were skipped: the key-management tier, not the migration's computational cost, is the binding constraint. The paper closes by positioning the framework against current mandates (NIST, NSA CNSA 2.0, the 2026 U.S. federal schedule, and the EU roadmap) and naming the engineering work — constant-time FN-DSA, HSM firmware support, and empirical validation — that separates this analytical architecture from a production deployment.