A Vendor-Neutral Analysis of HSM Vendors and Their Support for PQC
Top Seven Vendors and Ease of Upgrading to NIST PQC Algorithms
| Vendor/Product | Top Network-Attached HSM | Approach to PQC Upgrade | Ease of Upgrade (RSA/ECDSA to ML-DSA/ML-KEM) | Practicality of Vendor Claims |
| Entrust (nCipher) | nShield Connect | Post-Quantum SDK + firmware updates | Moderate (requires SDK coding + testing) | Claims of flexibility overstated; app rework significant |
| Thales SafeNet | Luna Network HSM | Functionality Modules (FMs) via firmware | Moderate (modular but slow/customization heavy) | “Crypto-agile” claim optimistic; FMs cumbersome |
| Utimaco | SecurityServer Se Gen2 | Firmware updates + PKCS#11 | Moderate to High (firmware smooth, config tweaks) | Claims solid; minor adjustments manageable |
| Futurex | Vectera Plus | Firmware updates + REST API/KMES tools | Moderate to High (firmware + integration effort) | Claims practical but setup complexity tempers ease |
| Marvell via AWS Cloud HSM | LiquidSecurity (AWS CloudHSM) | AWS-managed firmware updates + PKCS#11 | High (AWS handles it, app integration only) | Claims hold; ease relies on AWS pace, not user control |
| Thales via Azure Dedicated HSM | Luna 7 A790 (Azure Dedicated HSM) | Firmware updates with Thales tools | Moderate (firmware + manual integration) | Claims fair; user-managed effort higher than cloud-managed |
| Fortanix | Data Security Manager (DSM) | Software updates via REST API + SGX enclaves | High (seamless software updates, app tweaks) | Claims robust; ease real but SGX trust a factor |
Vendor Breakdown: Ease of Upgrade and Practicality
1. Entrust (nCipher) – nShield Connect
- Approach: Post-Quantum SDK (CodeSafe) and firmware updates support ML-DSA/ML-KEM.
- Ease of Upgrade: Moderate. Firmware adds PQC, but legacy RSA/ECDSA apps need SDK-driven code changes to handle larger keys (e.g., ML-DSA-44: 1312-byte public key vs. ECDSA P-256: 64 bytes). Hybrid certificates ease phased transitions.
- Practicality: “Future-proof” claims overpromise—app rework and performance tuning for PQC’s slower operations (e.g., ML-DSA signing) demand significant effort.
2. Thales SafeNet – Luna Network HSM
- Approach: Functionality Modules (FMs) via firmware enable ML-DSA/ML-KEM.
- Ease of Upgrade: Moderate. Pre-built FMs cover NIST PQC, but custom FMs for specific needs are slow and costly. Legacy apps must adapt to larger keys (e.g., ML-KEM-512: 800-byte public key vs. RSA-2048: 256 bytes).
- Practicality: “Crypto-agile” sounds great, but FMs are less practical than advertised—customization lags, and integration isn’t plug-and-play.
3. Utimaco – SecurityServer Se Gen2
- Approach: Firmware updates with PKCS#11 support ML-DSA/ML-KEM.
- Ease of Upgrade: Moderate to High. Firmware upgrades are efficient, and PKCS#11 helps legacy integration. Config tweaks (e.g., timeouts) address larger PQC signatures (e.g., ML-DSA-65: 3307 bytes vs. ECDSA: 72 bytes).
- Practicality: Claims of smooth transitions hold—firmware and standards support minimize friction, though minor adjustments are needed.
4. Futurex – Vectera Plus
- Approach: Firmware updates with REST API and KMES tools enable ML-DSA/ML-KEM.
- Ease of Upgrade: Moderate to High. Firmware adds PQC, and KMES aids key management, but legacy apps need API updates for bigger keys (e.g., ML-KEM-768: 1088-byte public key). Hybrid mode supports gradual shifts.
- Practicality: Claims of scalability are practical, but complex initial setup and app integration temper the ease for legacy systems.
5. Marvell via AWS Cloud HSM – LiquidSecurity (AWS Cloud HSM)
- Approach: AWS-managed firmware updates with PKCS#11 support ML-DSA/ML-KEM.
- Ease of Upgrade: High. AWS deploys PQC via firmware, so legacy users only update apps to handle larger keys/signatures (e.g., ML-DSA-87: 4595-byte signature). Cloud scaling manages performance.
- Practicality: Claims of seamless upgrades are realistic—AWS’s control simplifies everything. The caveat is reliance on AWS’s rollout schedule, which lacks firm dates as of March 03, 2025.
6. Thales via Azure Dedicated HSM – Luna 7 A790 (Azure Dedicated HSM)
- Approach: Customer-managed firmware updates with Thales tools support ML-DSA/ML-KEM.
- Ease of Upgrade: Moderate. Firmware enables PQC, but users manually update and reconfigure apps for larger data (e.g., ML-KEM-512: 1632-byte ciphertext). More effort than AWS’s managed model.
- Practicality: Claims of flexibility are fair—updates work, but user-managed effort and testing reduce practicality compared to cloud-managed options.
7. Fortanix – Data Security Manager (DSM)
- Approach: Software updates via REST API, leveraging Intel SGX enclaves, support ML-DSA/ML-KEM.
- Ease of Upgrade: High. No hardware swaps—software patches add PQC post-NIST standardization (already supports LMS, ML-DSA/ML-KEM). Legacy apps need minor tweaks via REST API for larger keys (e.g., ML-DSA-44: 1312 bytes).
- Practicality: Claims of “crypto-agility” are robust—software-driven updates are fast and scalable. SGX reliance might raise trust concerns, and PQC performance needs testing, but ease is real for cloud-friendly setups.
Key Insights: Upgrading Legacy to ML-DSA/ML-KEM
- Easiest Upgrades: Fortanix DSM and Marvell/AWS Cloud HSM lead—software/cloud management minimizes user effort to app-level tweaks. Utimaco and Futurex follow with firmware simplicity, needing slight config adjustments.
- Moderate Effort: Entrust, Thales SafeNet, and Thales/Azure Dedicated HSM require more work—SDK coding, custom FMs, or manual integration slow the process for legacy systems.
- Practicality Gaps: Vendors like Entrust and Thales SafeNet overhype “flexibility” and “agility”—SDKs and FMs sound elegant but bog down in real-world legacy uplift. Fortanix, Utimaco, and Marvell/AWS align better with their promises.
- Universal Challenges: PQC’s larger keys (e.g., ML-KEM-768: 1088 bytes vs. RSA-2048: 256 bytes) and slower operations (ML-DSA signing ~10x slower than ECDSA) mean all vendors face app updates and performance tuning. Hybrid certificates (RSA+PQC) help, but full PQC isn’t “drop-in” anywhere.