security: [T1-05] fix oracle signature bypass (DGB-SEC-002)#374
Closed
gto90 wants to merge 9 commits intofeature/digidollar-v1from
Closed
security: [T1-05] fix oracle signature bypass (DGB-SEC-002)#374gto90 wants to merge 9 commits intofeature/digidollar-v1from
gto90 wants to merge 9 commits intofeature/digidollar-v1from
Conversation
Bind authorized pubkey from chainparams before Schnorr signature verification in all code paths. Previously, Phase One validation called VerifyPhase2() using the message's embedded pubkey, allowing an attacker to forge oracle price messages with their own keypair. Fix applied to 4 verification sites: - IsValidOracleMessage() Phase One path (critical P2P gate) - IsValidOracleMessage() Phase Two path (already had binding, kept) - OracleDataValidator::ValidateOracleMessage() (defense-in-depth) - ValidatePhaseOneBundle() (defense-in-depth) - ValidatePhaseTwoBundle() (defense-in-depth) The pattern is: copy the message, overwrite oracle_pubkey with the chainparams-authorized key, then verify. This ensures signatures are always checked against trusted keys, never attacker-supplied ones.
There was a problem hiding this comment.
Pull request overview
This pull request addresses a critical security vulnerability (DGB-SEC-002) in oracle signature verification. The vulnerability allowed attackers to forge oracle price messages by providing their own keypair, as the signature verification used the message's embedded public key instead of the authorized key from chainparams.
Changes:
- Fixed 5 oracle signature verification code paths to bind the authorized public key from chainparams before verification
- Refactored price bounds validation to apply to both Phase One and Phase Two in IsValidOracleMessage()
- Added detailed security comments explaining the fix at each location
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
OpenSSL 1.1.1w Configure misparses multi-word flags like "-arch arm64" when passed as positional args — the shell splits them and Configure treats the second word as a target name, causing "target already defined" on ARM64 macOS. Pass CFLAGS/CPPFLAGS as VAR=value assignments per OpenSSL's official INSTALL docs. libcurl 8.5.0 fopen.c uses fileno/fdopen (POSIX) which are hidden by the depends system's strict -std=c11. Add -D_GNU_SOURCE for Linux.
Address Copilot review comments on SEC-002 PR: - Extract duplicated chainparams pubkey binding into a single static helper function, reducing 5 inline copies to 1 definition + 5 calls - Standardize variable naming (remove _p2 suffix) - Unify comment style across all verification sites
8 tasks
Use $$ deferred evaluation for CFLAGS/CPPFLAGS assignments so OS-specific flags (cflags_linux, cppflags_linux) appended by funcs.mk after set_vars are included. Route -fPIC and -D_GNU_SOURCE through proper cflags/cppflags variables instead of bare config_opts to avoid OpenSSL's "Mixing make variables" rejection.
The depends system never defines $(package)_arflags, so ARFLAGS=$($(package)_arflags) in config_env exported an empty ARFLAGS="" to the environment. When CFLAGS/CPPFLAGS are passed as VAR=value assignments (not positional flags), OpenSSL 1.1.1w's Configure sets $anyuseradd=false and falls back to reading env vars. The empty ARFLAGS overrides the target default "r" from Configurations/00-base-templates.conf, producing a Makefile with ARFLAGS= (empty). This causes "ar: two different operation options specified" on Linux and "ar: illegal option -- /" on macOS during build_libs, as ar interprets the archive path as flags. Fix: remove ARFLAGS from config_env entirely, letting Configure use its target default ARFLAGS="r". Verified locally: - Linux x86_64: ar r apps/libapps.a ... (ARFLAGS=r in Makefile) - macOS ARM64: ar r apps/libapps.a ... (ARFLAGS=r in Makefile) - Full build_libs completes successfully on darwin64-arm64-cc
Replace BOOST_CHECK with BOOST_CHECK_MESSAGE in 5 mint validation tests that fail on CI but pass locally. Diagnostic output captures the exact reject reason, script sizes, collateral amounts, and script type identification to help debug the CI-specific failure.
Replace ClearFreeze() with ClearHistory() in the validation test fixture. ClearFreeze() only clears freeze flags, but UpdateState() (called during ValidateDigiDollarTransaction) recalculates volatility from stale price history left by earlier test suites (e.g., digidollar_health_tests) and re-sets the freeze. This caused all "valid mint" tests to fail with "minting-frozen-volatility" on CI where the test suite ordering (linker-dependent) places health_tests before validation_tests. ClearHistory() resets price history, volatility state, and all freeze flags, ensuring each test starts from a clean state.
Keep diagnostic BOOST_CHECK_MESSAGE over base branch's BOOST_TEST_MESSAGE for better CI failure reporting.
The merge with feature/digidollar-v1 silently dropped critical code: 1. DD amount double-counting prevention in ValidateMintTransaction() The OP_RETURN and DD token output both encode ddAmount. Without the consistency check, totalDD was counted twice (e.g. 10000 + 10000 = 20000), requiring 2x the collateral and failing all valid mint tests. 2. Missing #include <digidollar/health.h> 3. Improved log format for insufficient-collateral diagnostic Also adopts the base branch's test file which includes: - DD OP_RETURN outputs required for T1-04b NUMS verification - Proper NUMS key usage via GetCollateralNUMSKey() - ClearHistory() fix for cross-suite volatility state pollution
|
Closing as duplicate — already fixed on Why this is superseded:This PR creates a The refactoring into a single helper function is a nice cleanup, but the security fix itself is already in place at all verification sites. The approach differs slightly (we fixed each site individually vs a centralized helper), but the security outcome is identical. Thanks @gto90 for the thorough analysis of the oracle signature bypass — this was one of the more critical findings. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
IsValidOracleMessage()Phase One path verified Schnorr signatures against the message's embedded pubkey instead of the chainparams-authorized key, allowing attackers to forge oracle price messages with their own keypairVerifyPhase2()in all 4 verification code paths (P2P gate + 3 defense-in-depth sites)Test plan
IsValidOracleMessage()rejects messages signed with unauthorized keystest_digibyte --run_test=oracle_tests)