Fix #256: per-peer HMAC secrets defeat cross-peer impersonation (#281)

Before this change, every peer-push HMAC was derived from the single
fleet-shared MESH_PEER_PUSH_SECRET. The receiver could prove "this
request was signed by someone who knows the fleet secret" but it could
NOT prove which peer signed it. Any peer that knew the global secret
could compute the expected HMAC for any other peer URL and forge a
push pretending to be that peer.

Fix: introduce MESH_PEER_SECRETS, an optional comma-separated
url=secret map. When a peer URL appears in the map, only the listed
per-peer secret is accepted for it -- the global secret is ignored for
that specific URL. Peer A no longer knows peer B's secret, so peer A
cannot forge a push claiming to be peer B.

The new helper resolve_peer_key_for_url() in mesh_crypto.py wraps the
lookup and is called from every existing peer-push call site:

- backend/auth.py:_verify_peer_push_hmac (receiver)
- backend/main.py:_http_peer_push_loop (Infonet event push)
- backend/main.py:_http_gate_pull_loop (gate event pull)
- backend/main.py:_http_gate_push_loop (gate event push)
- backend/services/mesh/mesh_router.py (two transports, push)
- backend/services/mesh/mesh_hashchain.py (gate wire ref key)
- backend/services/mesh/mesh_wormhole_prekey.py (peer prekey lookup)

Zero hostility, by design:

- Single-peer installs leave MESH_PEER_SECRETS empty -> resolver falls
  back to MESH_PEER_PUSH_SECRET -> behavior is byte-for-byte unchanged.
- Multi-peer installs that haven't migrated yet behave exactly as
  before.
- Multi-peer installs that DO migrate set MESH_PEER_SECRETS on both
  ends of each peering and immediately close the impersonation surface
  for those URLs. Migration is incremental: unlisted peers keep using
  the global secret.

Tests in backend/tests/test_per_peer_secret_resolver.py:
- env parsing (default, override, whitespace, malformed entries, cache)
- precedence: per-peer beats global
- migration window: unlisted peer falls back to global
- IMPERSONATION REFUSAL: peer A with global-secret-only cannot forge
  HMAC for peer B that has a per-peer secret configured
- IMPERSONATION REFUSAL: peer A with its OWN per-peer secret cannot
  forge HMAC for peer B
- positive control: legitimate peer B request verifies
- zero-behavior-change: single-peer install produces the same key bytes
  as before the change

Credit: tg12 (external security audit, P1/High/High confidence)
This commit is contained in:
Shadowbroker
2026-05-21 10:05:29 -06:00
committed by GitHub
parent 084e563412
commit 2e14e75a0e
9 changed files with 544 additions and 43 deletions
+6
View File
@@ -53,6 +53,12 @@ class Settings(BaseSettings):
MESH_RELAY_FAILURE_COOLDOWN_S: int = 120
MESH_BOOTSTRAP_SEED_FAILURE_COOLDOWN_S: int = 15
MESH_PEER_PUSH_SECRET: str = ""
# Issue #256 (tg12): optional per-peer HMAC secret map. Comma-separated
# `url=secret` pairs. When a peer URL appears here, only that per-peer
# secret is accepted for it — the global MESH_PEER_PUSH_SECRET above is
# ignored for that specific URL. Single-peer installs and unmigrated
# multi-peer installs leave this empty and behavior is unchanged.
MESH_PEER_SECRETS: str = ""
MESH_RNS_APP_NAME: str = "shadowbroker"
MESH_RNS_ASPECT: str = "infonet"
MESH_RNS_IDENTITY_PATH: str = ""
+109
View File
@@ -69,6 +69,115 @@ def _derive_peer_key(shared_secret: str, peer_url: str) -> bytes:
).digest()
# ---------------------------------------------------------------------------
# Issue #256 (tg12): per-peer HMAC secrets
# ---------------------------------------------------------------------------
#
# Before this change, ALL peer-push HMACs were derived from a single
# fleet-shared ``MESH_PEER_PUSH_SECRET``. The receiver could prove a
# request was signed by *someone who knows the fleet secret*, but it
# could NOT prove which peer signed it — any peer could compute the
# expected HMAC for any other peer's URL and impersonate that peer.
#
# Fix: an optional ``MESH_PEER_SECRETS`` env var maps specific peer URLs
# to per-peer secrets. When a peer URL is listed there, only that
# per-peer secret is accepted for that URL — the global secret is
# ignored for that peer. Peer A no longer learns peer B's secret, so
# peer A cannot forge a request claiming to be peer B.
#
# Backwards-compatible by design:
#
# - Single-peer installs (``MESH_PEER_SECRETS`` empty) keep using the
# global secret. Zero behavior change. Zero operator action required.
# - Multi-peer installs that haven't migrated yet keep using the global
# secret for every peer. Same behavior as before — same exposure.
# - Multi-peer installs that have migrated configure
# ``MESH_PEER_SECRETS=urlA=secretA,urlB=secretB`` and immediately get
# per-peer identity. Migration is incremental: peers not yet listed
# continue using the global secret until both sides of that peering
# add their entry.
_PEER_SECRETS_CACHE: dict[str, str] = {}
_PEER_SECRETS_CACHE_RAW: str = ""
def _lookup_per_peer_secret(normalized_url: str) -> str:
"""Return the per-peer secret for ``normalized_url`` from MESH_PEER_SECRETS.
Returns "" if no per-peer entry is configured for that URL. The parser
is forgiving:
- Whitespace around items, URLs, and secrets is stripped.
- Items without ``=`` or with empty URL/secret halves are skipped.
- The URL half is normalized via ``normalize_peer_url`` so config
authors don't have to match scheme/port/path quirks exactly.
The cache is invalidated whenever the env var's raw value changes,
which keeps tests' ``monkeypatch.setenv`` calls effective without
forcing a process restart.
"""
import os
raw = str(os.environ.get("MESH_PEER_SECRETS", "") or "").strip()
global _PEER_SECRETS_CACHE, _PEER_SECRETS_CACHE_RAW
if raw != _PEER_SECRETS_CACHE_RAW:
new_cache: dict[str, str] = {}
for chunk in raw.split(","):
chunk = chunk.strip()
if not chunk or "=" not in chunk:
continue
url_part, _, secret_part = chunk.partition("=")
normalized = normalize_peer_url(url_part.strip())
secret = secret_part.strip()
if normalized and secret:
new_cache[normalized] = secret
_PEER_SECRETS_CACHE = new_cache
_PEER_SECRETS_CACHE_RAW = raw
return _PEER_SECRETS_CACHE.get(normalized_url, "")
def resolve_peer_key_for_url(peer_url: str) -> bytes:
"""Return the HMAC key for ``peer_url``, preferring per-peer secret.
Issue #256: this is the function every peer-push call site should
use. It looks up the peer-specific secret first, falling back to the
fleet-shared ``MESH_PEER_PUSH_SECRET`` only when the URL is NOT
listed in ``MESH_PEER_SECRETS``.
Both sender (computing X-Peer-HMAC) and receiver (verifying it) call
this with the SENDER's URL — they must derive the same key, so
operators on both ends of a peering need matching MESH_PEER_SECRETS
entries for that URL to stay in sync.
Returns empty bytes when no usable secret exists. Callers must treat
that as fail-closed (skip the push, reject the verification).
"""
normalized_url = normalize_peer_url(peer_url)
if not normalized_url:
return b""
per_peer_secret = _lookup_per_peer_secret(normalized_url)
if per_peer_secret:
return _derive_peer_key(per_peer_secret, normalized_url)
# No per-peer entry for this URL — fall back to the legacy global
# secret. This is what preserves zero-hostility for single-peer
# installs and the migration window for multi-peer installs.
try:
from services.config import get_settings
global_secret = str(
getattr(get_settings(), "MESH_PEER_PUSH_SECRET", "") or ""
).strip()
except Exception:
return b""
if not global_secret:
return b""
return _derive_peer_key(global_secret, normalized_url)
def _node_digest(public_key_b64: str) -> str:
raw = base64.b64decode(public_key_b64)
return hashlib.sha256(raw).hexdigest()
+8 -7
View File
@@ -216,18 +216,19 @@ def _peer_pair_ref_key(peer_url: str) -> bytes:
Returns an empty key on misconfiguration so callers fail closed.
"""
try:
from services.config import get_settings
from services.mesh.mesh_crypto import _derive_peer_key, normalize_peer_url
secret = str(get_settings().MESH_PEER_PUSH_SECRET or "").strip()
from services.mesh.mesh_crypto import (
normalize_peer_url,
resolve_peer_key_for_url,
)
except Exception:
return b""
if not secret:
return b""
normalized = normalize_peer_url(peer_url or "")
if not normalized:
return b""
peer_key = _derive_peer_key(secret, normalized)
# Issue #256: resolve_peer_key_for_url() prefers per-peer secrets
# from MESH_PEER_SECRETS and falls back to the global
# MESH_PEER_PUSH_SECRET only when the URL has no per-peer entry.
peer_key = resolve_peer_key_for_url(normalized)
if not peer_key:
return b""
# Domain-separate from the transport HMAC key so the two
+16 -11
View File
@@ -26,7 +26,11 @@ from enum import Enum
from typing import Any, Callable, Optional
from collections import deque
from urllib.parse import urlparse
from services.mesh.mesh_crypto import _derive_peer_key, normalize_peer_url
from services.mesh.mesh_crypto import (
_derive_peer_key,
normalize_peer_url,
resolve_peer_key_for_url,
)
from services.mesh.mesh_metrics import increment as metrics_inc
from services.mesh.mesh_privacy_policy import (
TRANSPORT_TIER_ORDER as _TIER_RANK,
@@ -703,7 +707,6 @@ class InternetTransport(_PeerPushTransportMixin):
endpoint_path, padded = self._build_peer_push_request(envelope, self.NAME)
except ValueError as exc:
return TransportResult(False, self.NAME, str(exc))
secret = str(settings.MESH_PEER_PUSH_SECRET or "").strip()
delivered = 0
last_error = ""
@@ -713,10 +716,13 @@ class InternetTransport(_PeerPushTransportMixin):
try:
normalized_peer_url = normalize_peer_url(peer_url)
headers = {"Content-Type": "application/json"}
if secret:
peer_key = _derive_peer_key(secret, normalized_peer_url)
if not peer_key:
raise ValueError("invalid peer URL for HMAC derivation")
# Issue #256: per-peer secret takes precedence over the
# global MESH_PEER_PUSH_SECRET. When neither is set the
# key is empty and we skip the HMAC header entirely so a
# bare (unsigned) push still works on test deployments
# that have not yet configured any secret at all.
peer_key = resolve_peer_key_for_url(normalized_peer_url)
if peer_key:
headers["X-Peer-Url"] = normalized_peer_url
headers["X-Peer-HMAC"] = hmac.new(
peer_key,
@@ -798,7 +804,6 @@ class TorArtiTransport(_PeerPushTransportMixin):
endpoint_path, padded = self._build_peer_push_request(envelope, self.NAME)
except ValueError as exc:
return TransportResult(False, self.NAME, str(exc))
secret = str(settings.MESH_PEER_PUSH_SECRET or "").strip()
delivered = 0
last_error = ""
@@ -808,10 +813,10 @@ class TorArtiTransport(_PeerPushTransportMixin):
try:
normalized_peer_url = normalize_peer_url(peer_url)
headers = {"Content-Type": "application/json"}
if secret:
peer_key = _derive_peer_key(secret, normalized_peer_url)
if not peer_key:
raise ValueError("invalid peer URL for HMAC derivation")
# Issue #256: per-peer secret takes precedence; see the
# other transport above for the rationale.
peer_key = resolve_peer_key_for_url(normalized_peer_url)
if peer_key:
headers["X-Peer-Url"] = normalized_peer_url
headers["X-Peer-HMAC"] = hmac.new(
peer_key,
@@ -91,13 +91,15 @@ def _fetch_dm_prekey_bundle_from_peer_lookup(lookup_token: str) -> dict[str, Any
return {"ok": False, "detail": "lookup token required"}
try:
from services.config import get_settings
from services.mesh.mesh_crypto import _derive_peer_key, normalize_peer_url
from services.mesh.mesh_crypto import (
normalize_peer_url,
resolve_peer_key_for_url,
)
from services.mesh.mesh_router import configured_relay_peer_urls
settings = get_settings()
secret = str(getattr(settings, "MESH_PEER_PUSH_SECRET", "") or "").strip()
if not secret:
return {"ok": False, "detail": "peer prekey lookup unavailable"}
# Issue #256: secret check moved per-peer below. We still bail out
# cleanly when there are no peers configured at all.
peers = configured_relay_peer_urls()
if not peers:
return {"ok": False, "detail": "peer prekey lookup unavailable"}
@@ -121,7 +123,8 @@ def _fetch_dm_prekey_bundle_from_peer_lookup(lookup_token: str) -> dict[str, Any
or os.environ.get("SB_TEST_NODE_URL", "").strip()
or normalized_peer_url
)
peer_key = _derive_peer_key(secret, sender_peer_url)
# Issue #256: prefer per-peer secret keyed by the sender URL.
peer_key = resolve_peer_key_for_url(sender_peer_url)
if not peer_key:
continue
headers = {