Repository navigation
Replies: 13 comments 2 replies
|
I understand the rationale for changing the requirement for Class D in CMU-CSO-UVM from MUST to SHOULD. My concern is how this change will affect CSPs and agencies downstream. Today, a Class D designation tells us that a CSP offering uses FIPS-validated cryptographic modules. If the requirement becomes SHOULD, that assurance goes away. CSPs and agencies would no longer be able to tell whether a product uses FIPS-validated modules without access to the CSP’s CR26 implementation details. The SHOULD language also leaves CSPs unsure whether they need to refactor their applications. An agency could still require FIPS-validated modules under FISMA and OMB M-11-33, so a CSP may not find out a refactor is needed until an agency asks for it. That makes long-term technical and financial planning for serving the Federal Government much harder. I’d ask that the Class D requirement stay at MUST. If it does change to SHOULD, please require CSPs to publicly disclose whether their offering uses FIPS-validated modules, so agencies and other CSPs don’t need CR26 implementation details to find out. |
|
Commenting as a FedRAMP 3PAO assessor who also works CMMC with DIB contractors. I support CMU-CSO-TSC and the hybrid carve-out for PQC. A few things I'd like tightened before this is final:
|
|
Commenting as Files.com, recent Class A recipient who is currently working on our escalation to higher classes. We strongly support this decision for all the reasons you've noted in the RFC. |
|
I am in support of modernizing this rule. Forcing CSPs to use FIPS exclusively doesn't provide a tangible security benefit and often furthers fragmentation in service offerings. As FedRAMP has previously discussed as a program goal, getting to the state of parity between commercial and government offerings is far overdue. Removing a "hard" requirement of FIPS exclusively (which tends to lag behind modernized encryption that is far more suited to protect today's threat environments [i.e. post-quantum crypto, elliptic curve, etc.]) is a worthy improvement. I do worry about potential agency adoption of non-FIPS as they evaluate in their RMF/FIPS 199 contexts, so I would suggest guidance to agencies on how to properly evaluate cryptographic protections for their particular use cases. |
|
I support modernizing this rule as well. However, I feel some areas of concerns or atleast needs more clarification as listed below: Gap #1 Gap #2 Gap #3 Gap #4 |
|
@ryangmsi raised it: every FIPS 140-2 certificate moved to the CMVP Historical list on September 21. Under the proposed FRD-CVM, a lot of modules stopped being "CMVP-validated" overnight, and the provider didn't change anything. CMU-CSO-TSC doesn't fit that, because nothing was swapped. My suggestion: treat a certificate moving to Historical or Revoked as a vulnerability. The provider detects it (VDR-CSO-DET), evaluates it and gives it a PAIN rating (VER-EVA), works it within the VDR-TFR-PVR timeframe, and reports it in the monthly report like anything else. That gives it a clock and a reporting path that already exist. Then CMU-CSO-TSC only covers changes the provider actually makes, which is what it reads like it's meant for. |
|
If Class D moves to SHOULD, the public signal people are asking for only works if the CMU-CSO-CMD documentation has a fixed shape. I'd ask for a simple JSON format for the module inventory, in line with FRC-CSO-JSN, with at least:
With that, anyone can derive the "all / some / none" summary and show it on the Marketplace, and assessors can test it without reading a narrative. It gives agencies the signal back without bringing back the hard requirement. |
|
Microsoft's comments on the proposed cryptographic module requirements are below. CMU-CSO-TSC Microsoft does not believe CMU-CSO-TSC is necessary. CMVP already has a methodology to determine the extent of changes in a cryptographic module and FedRAMP’s policy for update streams is sufficient. Moreover, the types of changes described in the draft CMU-CSO-TSC are not commensurate with the definition of “transformative”. Creating an additional policy duplicates parts of the existing policies and creates discrepancies when CMU-CSO-TSC is not aligned with existing policies. CMU-CSO-CAT Microsoft recommends changing SHOULD to MUST in this requirement: “Providers MUST configure agency tenants by default to use cryptographic services that use CMVP-validated modules when such modules are available.” Using SHOULD could lead engineering teams to interpret validated cryptography as optional. MUST would establish a clear expectation while retaining the existing qualification, “when such modules are available.” Any additional exceptions should be explicitly defined so providers, assessors, and agencies apply the requirement consistently. |
|
Commenting as John B Joseph, Co-founder & CPTO, Purplecop Security. I support the direction of RFC-0035. My comment is about CMU-CSO-CMD. Once CMU-CSO-UVM moves Class D to SHOULD, that documentation is the main signal agencies have left. I agree with eddyboundera and ryangmsi that it needs a fixed, testable shape. I suggest FedRAMP reuse CycloneDX (1.6 or 1.7) for that shape. 1. CMU-CSO-CMD: accept the inventory as a CycloneDX cryptography bill of materials (CBOM). CycloneDX has a 2. FRD-CVM note: record validation per algorithm, not only per module. The note puts algorithms added by an update stream outside the validation. A module-level "validated: yes" hides that. A per-algorithm 3. CMU-CSO-TSC hybrid note: make the carve-out checkable. The exception applies "while a CMVP-validated module still protects federal customer data." If the validated classical algorithm and the added ML-KEM parameter set are listed under the same protocol, an assessor can confirm that condition from the data rather than from a narrative. 4. Two fields FedRAMP would need to define. CycloneDX has no native field for CMVP certificate status (Active, Historical, Revoked) or for approved-mode operation. Two FedRAMP-namespaced 5. Post-quantum visibility without new requirements. RFC-0035 sets no PQC requirements, and this proposal adds none. But NIST IR 8547 (final, October 2026) says NIST "will deprecate quantum-vulnerable public-key cryptography after 2030 and disallow its use under NIST standards and guidelines after 2035." It also says organizations should begin "with cryptographic inventories." The inventory can record the parameter set and NIST security category: ML-KEM-768 is category 3 (FIPS 203), ML-DSA-65 is category 3 (FIPS 204), SLH-DSA parameter sets are categories 1, 3 or 5 (FIPS 205), and quantum-vulnerable RSA or ECDH is 0. With that, an agency can read each service's migration status from the file FedRAMP already requires. The "all / some / none" Marketplace summary falls out of the same data. Example algorithm entry (valid against the CycloneDX 1.6 schema; 1.7 also accepts {"type": "cryptographic-asset", "name": "ML-KEM-768", "bom-ref": "alg-mlkem768",
"cryptoProperties": {"assetType": "algorithm", "algorithmProperties": {
"primitive": "kem", "parameterSetIdentifier": "768",
"certificationLevel": ["none"], "classicalSecurityLevel": 192,
"nistQuantumSecurityLevel": 3}}}Request: state in CMU-CSO-CMD that a CycloneDX 1.6+ CBOM satisfies the documentation and FRC-CSO-JSN, with the two FedRAMP properties above. |
|
CSP-AB submits the following comments to RFC-0035: General
Definitions
For Example:
Suggested wording: CMU-CSO-TSC Transformative Significant Changes of Modules: An algorithm that an update stream adds is not within the module’s validation; using one in place of a validated algorithm is such a change. Adding a non-validated algorithm while a CMVP-validated module still protects federal customer data, such as hybrid implementation, is not such a change.
For example: |
|
I strongly commend the FedRAMP PMO for formally acknowledging that the validation cadence of the NIST CMVP lags behind modern commercial software vulnerability management. Incorporating "update streams" into the definition of a validated cryptographic module under FRD-CVM is a vital, forward-looking step that aligns federal security with modern cloud and continuous delivery practices. However, to ensure that these proposed updates achieve their intended purpose without introducing new compliance ambiguities or audit friction, I recommend three critical refinements: 1. Formally Codify the Definition of "Cryptographic Update Stream" While RFC-0035 mentions update streams within definition FRD-CVM (CMVP-Validated Module), the specific operational boundaries that govern what constitutes an acceptable update stream are undefined in CR26. Previously, CSPs relied on Section 1 and requirements FRR1/FRR7 of the Board-Approved Legacy Policy for Cryptographic Module Selection and Use. Because legacy policies will sunset upon full adoption of CR26, leaving "update stream" without a binding definition invites inconsistent interpretations across authorizers and 3PAOs. Establishing a formalized definition directly within the CR26 framework ensures stability and regulatory predictability. Proposed Definition: 2. Establish an Objective Standard for Dormant / Disabled Algorithms ( The October 5 update to rule CMU-CSO-UVM states that any new algorithms added to an update stream fall outside the stream's scope. In practice, upstream OS errata (like RHEL 8 or OpenSSL) will bundle these ciphers whether we want them or not. Recommend that FedRAMP explicitly clarify two scenarios so 3PAOs don't arbitrarily cite CSPs:
3. Provide Explicit 3PAO Assessment Conformance Guidance (Safe Harbor) The rules should clearly state that systems operating within an approved update stream are fully compliant with CMU-CSO-UVM, preventing 3PAOs from issuing false-positive "Update beyond approved module" findings. Recommend providing clarification via a note. Proposed Assessor Clarification Note: |
|
Thank you for the opportunity to comment. SenseBeen supports enabling carefully documented cryptographic modernization while preserving agency visibility into the protection actually provided to federal data. We recommend that the final guidance include a small, practical evidence profile for cryptographic configuration and negotiation. Our perspective comes from development testing of TLS negotiation and policy enforcement. In a local loopback demonstration, we observed a successful X25519MLKEM768 connection, rejection of a classical-only client under a hybrid-only policy, and a successful classical connection when that alternative was explicitly allowed. These are bounded laboratory observations. They do not establish CMVP validation, production security, or post-quantum authentication. The useful distinction is between an algorithm being installed, an algorithm being allowed, and an algorithm actually being used. A cryptographic inventory alone cannot establish all three. 1. Link documentation to an observed configuration. For each relevant service or data flow, we recommend guidance identifying a minimum evidence set: module/provider and version; validation reference and applicable operating conditions where validation is claimed; enabled policy; observed negotiated algorithms; observation time; and a reference to the test result. Unknown or unobserved values should remain explicit. A handshake naming an algorithm should never be presented as proof that the implementation or deployment is validated. These fields should be usable within existing authorization documentation and machine-readable formats. They need not create a separate reporting platform or prescribe a particular commercial tool. 2. Make fallback behavior testable. We recommend a short worked example covering three outcomes: the intended hybrid connection succeeds; an incompatible peer is rejected when policy requires hybrid operation; and a classical connection succeeds only where the documented policy permits it. A failed connection must also be distinguished from a test setup failure, such as an unavailable endpoint. Record the negotiated result rather than inferring it from advertised capabilities or a configured preference list. Classical fallback is not automatically an attack or a policy violation; the evidence should establish whether it matches the declared policy for that service and tenant. 3. Explain the evidence needed for the hybrid provision. For CMU-CSO-TSC, please provide an example showing how a provider demonstrates that the protection supplied by a validated module remains in place when another algorithm is added. Separate key establishment, authentication, and record protection in that example. A successful hybrid key exchange should not imply that authentication is post-quantum, or that every component in the connection falls within a module's validation. 4. Preserve a useful before-and-after record. For cryptographic changes, we recommend a concise comparison of the previous and proposed configuration, affected data flows, test outcomes, permitted fallback, and rollback conditions. References to retained artifacts and integrity hashes can make this record reproducible. Hashes establish file consistency; they do not establish correctness, authenticity, or validation. Please publish this as lightweight implementation guidance with a sanitized example, allowing equivalent evidence and existing formats. Sensitive topology, keys, customer identifiers, and detailed configurations should remain in appropriately controlled authorization materials. The requested outcome is practical: agencies and assessors should be able to distinguish claimed capability from observed behavior, and providers should have a clear, proportionate way to demonstrate that cryptographic changes match their documented policy. Jordan Louis Henderson |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
RFC-0035 Cryptographic Module Use Rules Updates
Summary
This RFC proposes updates to the Cryptographic Module Use (CMU) rules in the Consolidated Rules for 2026 (CR26) for both 20X and Rev 5. The proposed updates remove a previously firm barrier at Class D but do not require currently operating cloud services to do anything differently.
This RFC also does not change agency obligations under FIPS 140-3, or set post-quantum cryptography requirements. Instead, this RFC acknowledges that many agency customers frequently use cloud services where validated cryptographic modules are not in play and FedRAMP’s role is not to block agency access to this type of service when an agency deems it is appropriate or necessary.
Motivation
Agencies decide where they rely on cryptography to protect federal information. Where they do, FIPS 140-3 applies regardless of FedRAMP Certification Class. There exist many valid technical reasons for agencies to bypass the requirements in FIPS 140-3 today, whether because the data is not sensitive enough to be encrypted with NIST Cryptographic Module Validation Program (CMVP) validated modules or because a specific agency use case requires devices or software that do not have validated modules.
Requiring Class D services to use only CMVP-validated modules can force providers to run separate cryptography for government customers. That cryptography can cost more, lag behind commercial security updates, and perform worse than alternative solutions that still provide sufficient protection for specific use cases.
This proposed change to the FedRAMP Certification rules, especially for Class D, will allow cloud services with specific use cases or valid technical reasons to use other types of cryptography to obtain and maintain a FedRAMP Certification. The existing rule of CMU-CSO-CMD (Cryptographic Module Documentation) requires cloud service providers to clearly document their use of cryptographic modules such that a potential agency customer will always know where and how the cloud service offering uses cryptographic modules and the potential risks of placing sensitive agency data into different parts of the cloud service offering. This proposed change will also allow cloud service providers to integrate emerging cryptographic technologies such as Post-Quantum Cryptography in some parts of their offering at the direct request of agency customers without creating vulnerabilities in the FedRAMP Practices.
Finally, the proposed change forces cloud service providers to treat changes where a CMVP-validated module is removed from an aspect of the cloud service offering as a transformative change to ensure that agency customers have sufficient and timely warning of the change so that they can make updated risk determinations about the continued use of the service if a cloud service offering stops using such modules.
Specific Proposed Changes
FedRAMP Definitions
FRD-CVM CMVP-Validated Module
Cryptographic Module Use
CMU-CSO-CMD Cryptographic Module Documentation
CMU-CSO-UVM Using CMVP-Validated Modules
CMU-CSO-TSC Transformative Significant Changes of Modules
CMU-CSO-CAT Configuration of Agency Tenants
All reactions