Close forged-directness link-rebind DoS on DM routing (#1421)

* Gate connected-link trust on a secure session; deny forged direct announces the connected shortcut

Residual gap after the rotation-heal containment (#1401): "verified
direct" announces prove the signature but not directness — TTL is
unsigned — so a malicious connected peer can replay a victim's fresh
announce with its TTL restored. When the victim has no live link, the
replayer's link rebinds to the victim's ID and reads as "connected",
and MessageRouter's connected fast-path then trusts it outright: every
DM stalls on a Noise handshake the replayer can never complete and is
silently lost while showing "sent".

Router-level trust gate (no wire change):
- Transport gains canDeliverSecurely(to:) — BLE answers with an
  established Noise session; Nostr keeps its prompt-delivery predicate;
  the protocol default forwards to canDeliverPromptly for transports
  without a forgeable link layer.
- MessageRouter.sendPrivate only trusts a connected link outright when
  it can deliver securely; otherwise it still sends (kicking the
  handshake on a genuine link) but retains a copy and hands a sealed
  copy to couriers, like the reachable path. flushOutbox gets the same
  gate so a flush over an insecure link resends instead of dropping the
  retained copy.

Presence hardening (defense in depth): a "direct" announce arriving on
a link already bound to a different peer no longer shortcuts the
claimed peer into "connected" — only real link state does. Genuine
first-contact and direct announces are unaffected; only the ambiguous
heal path loses the forgeable shortcut.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* Harden the insecure-link outbox bound; document the second-replay presence gap and the courier metadata tradeoff

Review follow-ups on the canDeliverSecurely gate:

- flushOutbox no longer counts connected-but-insecure flushes toward the
  maxSendAttempts drop: the message was actually transmitted over a live
  link, so a peer whose Noise handshake stalls across reconnect flapping
  must not burn through the cap and lose the store-and-forward copy the
  gate exists to preserve. Retention stays bounded by the 24h outbox TTL
  and the per-peer FIFO cap; acks still clear it. Attempt-counting stays
  for reachable-only (heuristic) sends. Regression test: >8 connected-
  insecure flushes keep the retained copy, drop callback never fires.

- Document the known second-replay presence gap: linkBoundToOtherPeer
  reads the binding before rebindLinkAfterVerifiedDirectAnnounce steals
  the link, so once a first replay has rebound a link to an absent
  victim's ID, a second replay marks the victim connected. Presence
  display only — DMs stay on the retain+courier path via the router
  gate. Not closed at the announce layer because the post-rebind state
  is indistinguishable from a legitimate rotation/reconnect heal (a
  supported, field-verified flow). Covered by a two-announce test.

- Note the accepted courier-spray metadata tradeoff at the connected-
  insecure send: nearby verified peers receive a sealed copy (they learn
  a DM exists, never its content), cleared on ack.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* Promote a healed rotation to connected; normalize the secure-delivery probe; retain Codable properties in Periphery

Codex review follow-ups on 5017c232:

- P1: a legitimate rotation announce necessarily arrives on a link still
  bound to the OLD peer ID, so the linkBoundToOtherPeer denial stored the
  new identity as disconnected — and the rebind only fixed link state,
  never the registry. A healed rotation then read as disconnected until
  the peer happened to announce again. rebindLinkAfterVerifiedDirect-
  Announce now promotes the rebound identity to connected after the
  containment checks pass (registry markConnected + topology/snapshot/
  peer-list refresh; the .peerConnected UI event already fired from the
  announce path). This consciously moves the forged-presence residue
  from second-replay to first-successful-rebind — bounded by the rebind
  containment (never steals a live identity, one rebind per link per
  cooldown) and still display-only: DMs stay gated on canDeliverSecurely.
  Comments and the pinned tests updated; new regression test covers
  rotate-on-open-link -> rebind -> isPeerConnected(new) == true.

- P2: BLEService.canDeliverSecurely probed the Noise session with the
  peer ID as given, but sessions are keyed by the short wire ID — a send
  keyed by the full 64-hex Noise key (favorites resolution) misread an
  established session as insecure and needlessly retained + couriered
  every DM until ack. Normalize with toShort() like isPeerConnected.
  Test drives a real XX handshake and asserts both ID forms pass.

- Periphery: the PrekeyBundleStore.StoredBundle.noiseKey "assign-only"
  flake fired again and slipped past its baselined USR (read exists in
  loadFromDisk; the indexer intermittently misses it). Replace the
  baseline entry with retain_codable_properties: true — deterministic,
  and a truly-dead Codable field is a persisted-format change anyway.
  Local periphery scan --strict: no unused code detected.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: jack <jackjackbits@users.noreply.github.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
jack
2026-07-09 16:46:33 +02:00
committed by GitHub
co-authored by jack Claude Opus 4.8
parent c8479c15e3
commit b7c6f42b3a
12 changed files with 530 additions and 12 deletions
+138
View File
@@ -253,6 +253,144 @@ struct BLEServiceCoreTests {
#expect(ble.currentPeerSnapshots().map(\.peerID).contains(attackerPeerID))
}
@Test
func replayedDirectAnnounceForAbsentPeerNeverYieldsSecureDelivery() async throws {
// Residual heal-path gap: the victim has NO live link, so the
// identity-owns-a-link containment cannot refuse the rebind. The
// replay steals the link binding, and because a successful rebind
// promotes its new owner to connected (a legitimate rotation heal
// requires that), the absent victim may read as connected. That
// forged presence is display-only and accepted the invariant that
// holds is that the stolen link can never produce an established
// Noise session, so MessageRouter's canDeliverSecurely gate routes
// DMs through retain + courier instead of trusting it outright.
let ble = makeService()
let attackerPeerID = PeerID(str: "1122334455667788")
let attackerLink = "central-attacker-absent-victim"
ble._test_seedConnectedPeer(attackerPeerID, nickname: "attacker")
ble._test_bindCentral(attackerLink, to: attackerPeerID)
// The absent victim's fresh signed announce, replayed on the
// attacker's bound link with its direct TTL restored.
let victimSigner = NoiseEncryptionService(keychain: MockKeychain())
let announcement = AnnouncementPacket(
nickname: "victim",
noisePublicKey: victimSigner.getStaticPublicKeyData(),
signingPublicKey: victimSigner.getSigningPublicKeyData(),
directNeighbors: nil
)
let payload = try #require(announcement.encode(), "Failed to encode announcement")
let victimPeerID = PeerID(publicKey: announcement.noisePublicKey)
let unsigned = BitchatPacket(
type: MessageType.announce.rawValue,
senderID: Data(hexString: victimPeerID.id) ?? Data(),
recipientID: nil,
timestamp: UInt64(Date().timeIntervalSince1970 * 1000),
payload: payload,
signature: nil,
ttl: 7
)
let packet = try #require(victimSigner.signPacket(unsigned), "Failed to sign announce packet")
#expect(ble._test_recordIngressIfNew(packet: packet, linkID: attackerLink))
ble._test_handlePacket(packet, fromPeerID: victimPeerID, preseedPeer: false)
// The rebind steals the link (no live link owns the victim's
// identity, so containment cannot refuse)
let rebound = await TestHelpers.waitUntil(
{ ble._test_centralBinding(attackerLink) == victimPeerID },
timeout: TestConstants.longTimeout
)
#expect(rebound)
// and the promote marks the absent victim connected: the accepted,
// display-only forged-presence residue (documented at
// BLEAnnounceHandler's linkBoundToOtherPeer check)
let forgedPresence = await TestHelpers.waitUntil(
{ ble.isPeerConnected(victimPeerID) },
timeout: TestConstants.longTimeout
)
#expect(forgedPresence)
// but secure delivery stays impossible the DM gate holds, and
// MessageRouter retains + couriers instead of trusting the link.
#expect(!ble.canDeliverSecurely(to: victimPeerID))
}
/// A legitimate rotation announce necessarily arrives on a link still
/// bound to the OLD ID, so its registry upsert stores the new peer
/// disconnected. The successful rebind must promote it: a healed
/// rotation with a live link has to read as connected again for routing
/// and outbox flushes.
@Test
func rotationHealPromotesRotatedPeerToConnected() async throws {
let ble = makeService()
let oldPeerID = PeerID(str: "1122334455667788")
let centralUUID = "central-rotation-promote"
ble._test_seedConnectedPeer(oldPeerID, nickname: "alice")
ble._test_bindCentral(centralUUID, to: oldPeerID)
let signer = NoiseEncryptionService(keychain: MockKeychain())
let announcement = AnnouncementPacket(
nickname: "alice",
noisePublicKey: signer.getStaticPublicKeyData(),
signingPublicKey: signer.getSigningPublicKeyData(),
directNeighbors: nil
)
let payload = try #require(announcement.encode(), "Failed to encode announcement")
let newPeerID = PeerID(publicKey: announcement.noisePublicKey)
let unsigned = BitchatPacket(
type: MessageType.announce.rawValue,
senderID: Data(hexString: newPeerID.id) ?? Data(),
recipientID: nil,
timestamp: UInt64(Date().timeIntervalSince1970 * 1000),
payload: payload,
signature: nil,
ttl: 7
)
let packet = try #require(signer.signPacket(unsigned), "Failed to sign announce packet")
#expect(ble._test_recordIngressIfNew(packet: packet, linkID: centralUUID))
ble._test_handlePacket(packet, fromPeerID: newPeerID, preseedPeer: false)
let rebound = await TestHelpers.waitUntil(
{ ble._test_centralBinding(centralUUID) == newPeerID },
timeout: TestConstants.longTimeout
)
#expect(rebound)
let connected = await TestHelpers.waitUntil(
{ ble.isPeerConnected(newPeerID) },
timeout: TestConstants.longTimeout
)
#expect(connected)
}
/// Noise sessions are keyed by the short wire ID, but routers may key
/// sends by the full 64-hex Noise key (favorites resolution does). The
/// secure-delivery gate must normalize like isPeerConnected, or an
/// established session is misread as insecure and every DM needlessly
/// retains + couriers until an ack.
@Test
func canDeliverSecurelyNormalizesFullNoiseKeyPeerIDs() async throws {
let ble = makeService()
let remote = NoiseEncryptionService(keychain: MockKeychain())
let remoteKey = remote.getStaticPublicKeyData()
let shortID = PeerID(publicKey: remoteKey)
let fullKeyID = PeerID(hexData: remoteKey)
#expect(fullKeyID.toShort() == shortID)
#expect(!ble.canDeliverSecurely(to: shortID))
// Full XX handshake; the local side keys the session by the short
// wire ID, exactly as packets present it in production.
let m1 = try ble._test_noiseInitiateHandshake(with: shortID)
let m2 = try #require(try remote.processHandshakeMessage(from: ble.myPeerID, message: m1))
let m3 = try #require(try ble._test_noiseProcessHandshakeMessage(from: shortID, message: m2))
_ = try remote.processHandshakeMessage(from: ble.myPeerID, message: m3)
#expect(ble.canDeliverSecurely(to: shortID))
#expect(ble.canDeliverSecurely(to: fullKeyID))
}
@Test
func ingressRejectsSelfLoopbackBeforeSpoofChecks() async throws {
let ble = makeService()