3 Commits

Author SHA1 Message Date
romanett 7fe7794a19 Enable AES-GCM and ChaCha20-Poly1305 policies on net48 via BouncyCastle (stacked on #4703) (#4705)
> Follow-up to #4703 (now merged to master; this PR has been retargeted
to `master`).

Part of #4510. #4703 enabled the CBC ECC policies on net48 through a
BouncyCastle raw-ECDH polyfill. The AEAD policies were still unavailable
there because .NET Framework has no `AesGcm` or `ChaCha20Poly1305`. This
PR adds those ciphers the same way, so on net48 all of these are now
offered:
`ECC_{nistP256,nistP384,brainpoolP256r1,brainpoolP384r1}_AesGcm` /
`_ChaChaPoly`, `RSA_DH_AesGcm`, `RSA_DH_ChaChaPoly`.

## Changes
- **Polyfill**
`src/Opc.Ua.Security.Certificates/Org.BouncyCastle/AeadCiphers.cs`
(internal, `#if NETFRAMEWORK`, next to `EcdhRawAgreement`): `AesGcm` and
`ChaCha20Poly1305` types with the .NET 8 BCL surface. They wrap
BouncyCastle's `GcmBlockCipher(AesEngine)` and
`Org.BouncyCastle.Crypto.Modes.ChaCha20Poly1305`.
- A tag mismatch throws `CryptographicException`, never BC's
`InvalidCipherTextException`, so
`PlatformSymmetricCryptoProvider.DecryptAuthenticated` still returns
`false` and the channel still raises a protocol error.
- GCM releases plaintext blocks before it checks the tag, so plaintext
stays in a scratch buffer and is copied out only after the tag verifies.
On failure the destination is cleared, as the BCL does. Scratch buffers
and the key copy are zeroed.
- **Opc.Ua.Core**: aliases the polyfill types on `NETFRAMEWORK` and
removes the `NET8_0_OR_GREATER` gates in `SecurityPolicies`
(`SupportsAesGcmPolicy` / `SupportsChaCha20Poly1305Policy`),
`CryptoUtils` (`SymmetricEncryptAndSign` / `SymmetricDecryptAndVerify`,
AEAD helpers) and `PlatformSymmetricCryptoProvider`. The channel, the
provider facet and the encrypted ECC user tokens (`EncryptedSecret`,
through `CryptoUtils`) now run the same code on every target. .NET 8+
keeps the BCL ciphers.
- **FIPS** (requested by the #4703 session, after a2ecf13b1): under
`CryptoCompliancePolicy.FipsOnly`, `CryptoCompliance.IsPolicyPermitted`
now also withholds every AEAD policy on .NET Framework. That adds
`RSA_DH_AesGcm`, which is FIPS-approved and not ECC. The BouncyCastle
cipher runs outside any validated module and the user-token path
bypasses the provider registry, so the auditor cannot see it. Tests and
`docs/CryptoProvider.md` are updated to match.
- **Docs**: EccProfiles (Known Limitations), migration `certificates.md`
(and the bundled skill copy), ProductionChecklist, Profiles,
WhatsNewIn2.0, CryptoProvider. They state that on net48 these primitives
run in managed BouncyCastle, which is not a validated (FIPS) module and
not hardware-accelerated or hardened against cache-timing attacks like
CNG AES.

## Tests
- New `AeadPolyfillTests` (Core.Tests):
- Known answers: NIST GCM test cases 2/4 (AES-128) and 14/16 (AES-256),
plus RFC 8439 §2.8.2 for ChaCha20-Poly1305. They run through
`PlatformSymmetricCryptoProvider` on every TFM and through the polyfill
directly.
- Tampering with the tag, the ciphertext or the AAD gives `false` from
the provider, and `CryptographicException` with a cleared plaintext from
the polyfill.
  - Argument validation.
- **BCL cross-check:** on .NET 8+, Core.Tests also compiles
`AeadCiphers.cs` (`AEAD_POLYFILL_CROSSCHECK`, test-only
`BouncyCastle.Cryptography` reference). Polyfill and BCL must produce
identical ciphertext and tag, and each must decrypt the other's output:
3 algorithms × 8 message lengths (0 to 8192) × 3 AAD lengths.
- `CryptoProviderRegistryTests`: `RSA_DH_AesGcm` is permitted under
FipsOnly only off .NET Framework and withheld on it.
- Tests that `Assert.Ignore` when AEAD is unavailable now run on net48,
e.g. `EccSecretRejectsDifferentNegotiatedPolicyAsync` and the
`RSA_DH_AesGcm` server session tests.

### Local results
| Run | Result |
|---|---|
| Core.Tests net48, full (before the FIPS commit) | 6049 passed, 0
failed, 88 skipped |
| Core.Tests crypto/AEAD/ECC/EncryptedSecret/Nonce subset, net48 /
net10.0 | 278 / 343 passed, 0 skipped |
| Server.Tests session-security/ECC subset, net48 | 128 passed, 0
skipped |
| `EccInteropTests` net48 vs **open62541 1.5.8 (OpenSSL)** peer | all 16
AEAD cases pass, 4 of them two-renewal token-renewal runs
(`ECC_nistP256_AesGcm`/`_ChaChaPoly` × Sign/SignAndEncrypt ×
Anonymous/UserName, both directions). Remaining 8 failures are the
documented open62541 CBC Sign-only HKDF differences in
`expected-differences.json` |
| `EccInteropTests` net48 vs 1.5.378 LegacyPeer | unchanged 26 passed /
28 skipped |

Note on the acceptance criterion: the 1.5.378 LegacyPeer does not
implement the AEAD policies, so its
`_AesGcm`/`_ChaChaPoly`/`*AeadTokenRenewalAsync` cases skip on every TFM
("The peer (UA-.NETStandard) does not declare …"). Before this PR they
skipped on the 2.0 side; now they get past it and stop at the peer
check. The open62541 peer is the real cross-stack check of the net48
AEAD path against a non-BouncyCastle implementation.

## PubSub DTLS on net48
The DTLS 1.3 transport (`src/Opc.Ua.PubSub.Udp/Dtls`) was compiled out
on .NET Framework, so net48 registered no DTLS profiles. It now reuses
the same polyfills (`Opc.Ua.Security.Certificates` gains
`InternalsVisibleTo Opc.Ua.PubSub.Udp`):
- The `NET8_0_OR_GREATER` gates in the handshake, ECDHE, record
protection and profile probe are removed.
- Two .NET 8 APIs get net48 fallbacks: the AES-ECB record-number mask
uses `CreateEncryptor`, and the handshake random uses
`RandomNumberGenerator.Create`.
- The HKDF probe now checks the managed `DtlsHkdf` that the key schedule
actually uses.
- The public `IsAeadSupported` / `IsRawSharedSecretSupported` probes are
kept and now return `true` everywhere.
- The DTLS tests now also run on net48. The net48 "registers no
profiles" test now asserts that the NIST AEAD profiles are registered.
The socket transport-peer tests and the benchmarks stay .NET 8-only
because they use .NET 8 socket APIs.
- `docs/PubSub.md` DTLS profile table: the net48 column is updated.

Results: PubSub.Udp.Tests net48, full suite: **229 passed, 0 skipped**.
It includes complete NIST AEAD, brainpool and integrity-only handshakes
and ChaCha20 records. The DTLS subset is 76 on net48 vs 81 on net10.0;
the 5 still-gated socket-peer tests are the difference.

Note: DTLS profile selection does not consult `CryptoCompliancePolicy`
on any target. That is pre-existing and unchanged here.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-09 06:30:14 +02:00
romanett d552634258 Enable ECC security policies on net48 via a BouncyCastle raw-ECDH polyfill (#4703)
Closes #4510.

## Summary

OPC UA Part 6 feeds the **raw** ECDH shared secret into HKDF. .NET
Framework only exposes `ECDiffieHellman.DeriveKeyMaterial`, which always
hashes it, so the ECC SecureChannel and user-token policies were
unavailable on `net48` (`SecurityPolicies.GetInfo` returned `null`).

This adds a BouncyCastle polyfill for just that primitive:

-
**`Opc.Ua.Security.Certificates/Org.BouncyCastle/EcdhRawAgreement.cs`**
(`net48` only): an internal `DeriveRawSecretAgreement(this
ECDiffieHellman, ECDiffieHellmanPublicKey)` extension using
`ECDHBasicAgreement`. It returns the X coordinate padded to the field
size, exactly like the .NET 8 BCL method. Keys stay CNG keys; ECDSA
signatures and HKDF stay on the platform — only the agreement step is
managed code. The point is validated (on curve, not infinity) and
mismatched curves are rejected.
- Because the extension mirrors the .NET 8 signature,
`Nonce.GenerateSecret` now has one code path on every target.
- `SupportsRawEccSecretAgreement()` is removed; the CBC-based ECC
policies (`ECC_nistP256`, `ECC_nistP384`, `ECC_brainpoolP256r1`,
`ECC_brainpoolP384r1`) are offered on every target (still subject to OS
curve support).
- BouncyCastle was already a `net48` dependency of
`Opc.Ua.Security.Certificates`; no new package. (Only `net48` is
downlevel now — `netstandard2.1`/`net472` from the issue text are no
longer targets.)

**Not in scope:** the `_AesGcm` / `_ChaChaPoly` ECC variants (and PubSub
DTLS) still need .NET 8 because .NET Framework has no AEAD ciphers —
follow-up stacked on this branch.

## Docs

`EccProfiles.md` (Known limitations), `migrate/2.0.x/certificates.md`,
`ProductionChecklist.md`, `Profiles.md`, `WhatsNewIn2.0.md`. Calls out
that on `net48` the agreement step does not run in a FIPS/validated
module.

## Tests (local)

| Suite | Result |
|---|---|
| `Opc.Ua.Core.Tests` full, net48 | 5956 passed, 0 failed |
| `EccInteropTests` net48 vs 1.5.378 peer (net8.0, native raw ECDH) | 26
passed; 28 skipped (AEAD + curve25519/448) — all four CBC curves, both
directions, UserName tokens, token renewal |
| `Opc.Ua.Sessions.Tests` ECC filter, net48 | 52 passed, 0 failed |
| Server `SessionCoverageTests`/`EphemeralNonce*`, Configuration builder
tests, net48 | all pass |
| Nonce/ECC/EncryptedSecret + session tests, net10.0 | all pass |

New tests: symmetric agreement on all four curves; on net48 an oracle
against CNG (`SHA256(raw) == DeriveKeyFromHash`); curve mismatch and
off-curve rejection. Tests that asserted the old net48 limitation were
updated.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-08 21:52:44 +02:00
Marc Schier c4e0721a29 Add a documentation learning path and fix issues found while verifying it (#4555)
## Summary

### Documentation
- Stage the documentation index as a learning path: a first application,
independent client, server, and PubSub paths, then data and services,
companion models, production, advanced integration, and draft or
experimental capabilities.
- Add Getting started (repository samples or NuGet packages), OPC UA
concepts, Your first information model, and a production readiness
checklist.
- Add Native AOT application publishing, a first PubSub publisher and
subscriber exchange, and the UDP transport options.
- Add or expand contents navigation throughout `docs/`, and keep
contributor documentation separate.
- Correct outdated SDK examples and capability guidance, consolidate
repeated explanations into canonical guides, and repair local links and
heading structure.
- Rewrite What's New in 2.0 as a concise feature overview with short
linked details; keep migration instructions in the Migration Guide.
- Expand the feature catalogue with specification IDs, GDS, LDS, the OPC
10000-21 onboarding registrar, ISA-95, xRegistry, OpenUSD, and
sample-local models; distinguish draft status, explain crypto
pluggability, and bring performance guidance forward.
- Address review feedback on wording, active monitored items during
shadow reload, and the `Samples.md` filename and references.
- Regenerate the `opcua-v20-migration` plugin's bundled copy of the
migration guides so it matches the edited `docs/migrate/2.0.x` guides.
- Merge the latest `master`, preserving upstream resource-isolation
guidance within the task-based documentation index.
- Retain deferred Lingity findings where appropriate; this PR does not
claim that every document meets the Lingity score thresholds.

### Fixes found while verifying the guides
Each fix has regression tests that fail without it.
- **Client:** V2 subscriptions compared their lifetime with the
operation timeout instead of the session timeout, and logged a spurious
warning. `ISubscriptionEngineContext`, which is new in 2.0, gains a
`SessionTimeout` property; implementations must add it.
- **PubSub:** a subscriber now registers the `DataSetMetaData`
configured on its readers (OPC 10000-14 6.2.9.4), and the UADP decoder
decodes DataSetMessages that carry no ConfigurationVersion. Before, a
secured subscriber of the reference sample received no DataSets.
- **Source generator:** a ModelDesign method that declares its arguments
inline, without a method type, produced code that did not compile; it
now uses `MethodState`. The generated code of existing models is
unchanged.
- **MinimalClient sample:** waits for its first data-change notification
before it disconnects.

### Known issues, not addressed here
- UDP multicast: the publisher does not send on the interface that the
subscriber joins, and `MulticastLoopback` defaults to false. The PubSub
guide uses unicast for the exchange on one computer.
- The UDP metadata announcement is sent unsecured on secured writer
groups; a secured subscriber drops it with a warning.
- A server does not start in globalization-invariant mode, because
`ResourceManager` creates `CultureInfo("en-US")`. The Native AOT guide
cautions against the mode.

## Validation
- `git diff --check` and
`./.azurepipelines/validate-migration-plugin.ps1` (15 bundled migration
docs validated, plugin manifests synchronized).
- Local Markdown links, anchors, file-link casing, and contents lists
checked across all 87 documents under `docs/`. The index links every
top-level guide and every target of the previous index.
- Both Getting started routes, Your first information model, the PubSub
exchange, and the Native AOT publish of the Getting started client and
server were run as documented. The NuGet route and the tutorial use the
public `2.0.0-preview.5` packages.
- The changed libraries build for all target frameworks in Release
without warnings.
- Tests on .NET 10: Opc.Ua.SourceGeneration.Tests 199 passed;
Opc.Ua.SourceGeneration.Core.Tests 3,977 passed, 8 skipped;
Opc.Ua.Server.Tests `Fluent` category 621 passed; Opc.Ua.Client.Tests
subscription-engine tests 70 passed; Opc.Ua.PubSub.Tests 1,374 passed, 2
skipped; PubSub fuzz replay 145 passed. The new and affected tests also
pass on .NET Framework 4.8.
- Earlier merge validation: 174 resource-isolation tests passed on .NET
10.
- The full solution suite was not run locally.

Review and validation artifacts are maintained outside the repository.

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-09-28 19:19:39 +00:00