Compare commits
No commits in common. "leora/fix-r3-reinstate-requires-rekey" and "main" have entirely different histories.
leora/fix-
...
main
64
DESIGN.md
64
DESIGN.md
@ -123,9 +123,11 @@ means a future route cannot pass a string where a proof belongs.
|
||||
|
||||
`revoke <agent>` with no end time is permanent: the enrolled public key is
|
||||
cleared and the identity stays dead until someone bootstraps a fresh secret and
|
||||
rekeys (or re-enrols). `reinstate` alone lifts the ban flags but does **not**
|
||||
restore the compromised key — that was the footgun. `revoke <agent> --until
|
||||
<unix>` is a pause that keeps the key, and it lapses on its own.
|
||||
calls `enrol_rekey` (then `reinstate` if still banned). `enrol` will not work —
|
||||
the agent row still exists — and would only burn the seed. `reinstate` alone
|
||||
lifts the ban flags but does **not** restore the compromised key — that was the
|
||||
footgun. `revoke <agent> --until <unix>` is a pause that keeps the key, and it
|
||||
lapses on its own.
|
||||
|
||||
The distinction matters because the two failure modes are different. A key
|
||||
compromise wants death — a lapsing revocation would hand the attacker back in
|
||||
@ -134,10 +136,17 @@ wrong moment. A noisy neighbour or a holiday wants a pause, and re-enrolling for
|
||||
that would destroy the audit history that says it was deliberate.
|
||||
|
||||
Revoking kills live sessions, drops outstanding challenges, and invalidates
|
||||
unused view codes, so a revocation that races an in-flight authentication
|
||||
cannot be overtaken by it, and a pre-issued code cannot mint a fresh human
|
||||
session after the agent is suspended or dead. Rekey invalidates codes the same
|
||||
way: rotating a key must close every door the old key could open.
|
||||
unused view codes. Rekey invalidates codes the same way: rotating a key must
|
||||
close every door the old key could open.
|
||||
|
||||
A narrow check-then-insert race can still insert a session or code after the
|
||||
revoke/rekey statement has run. The choke point is `sessions.resolve`: every
|
||||
use re-checks that the agent is not revoked, still has a live key, and that
|
||||
the session fingerprint (when present — always for agent sessions, and for
|
||||
human_via_agent sessions minted with a fingerprint) matches the current key
|
||||
epoch. View codes store the issuing fingerprint and refuse redeem after rekey
|
||||
or key wipe. That is what makes revocation overtake in-flight
|
||||
verify/issue/redeem, not the first-pass invalidation alone.
|
||||
|
||||
### Existing controls stay
|
||||
|
||||
@ -195,9 +204,15 @@ history, forges replacements, and rebuilds the chain — `verify_chain` passes.
|
||||
So the head digest also gets **signed checkpoints**, by an enrolled agent whose
|
||||
private key the database does not contain. `verify-chain` runs both checks, and
|
||||
the rewrite is caught by the signature the attacker cannot produce. A forged
|
||||
checkpoint fails signature verification, a checkpoint from a stranger is refused
|
||||
at signing time, and a rekey invalidates every earlier checkpoint — which is
|
||||
correct, and says so rather than failing quietly.
|
||||
checkpoint fails signature verification, and a checkpoint from a stranger is
|
||||
refused at signing time.
|
||||
|
||||
Public keys are kept in an append-only `agent_keys` history. Revoke and rekey
|
||||
*retire* a key rather than destroy it, so honest pre-compromise checkpoints and
|
||||
off-box anchors stay verifiable against the fingerprint they name. Checkpoints
|
||||
timestamped after a key's `retired_at` are flagged (not failed). Prefer a
|
||||
**dedicated checkpoint-signer identity** that is never a working agent, so a
|
||||
routine agent compromise-revoke does not touch the tip-signing key at all.
|
||||
|
||||
What this still does not do is survive an attacker who can also delete the
|
||||
checkpoints, because they live in the same database. Set `anchor_path` /
|
||||
@ -216,7 +231,7 @@ because the review looked for it.
|
||||
|
||||
## Status
|
||||
|
||||
106 passing tests, and two hardening passes that closed real defects. The worst were
|
||||
158 passing tests, and two hardening passes that closed real defects. The worst were
|
||||
concurrency: one-time codes were redeemable repeatedly (753 sessions from 200
|
||||
codes under load), and a bootstrap secret could enrol more than one key. Those
|
||||
are now atomic compare-and-set operations, each covered by a test that races
|
||||
@ -229,7 +244,7 @@ sessions mintable from a bare agent name, no revocation operation at all, an
|
||||
unkeyed chain that a database-level attacker could rebuild, and SQLite without
|
||||
a busy timeout.
|
||||
|
||||
Leora's contracted prod pass (Quarry 18a5d762) closed three remaining review blockers on this branch: rekey no longer un-revokes, dual-bootstrap enrol is exactly-one under race, and redeem rate limits are per-agent (anonymous fails isolated from success). Off-box checkpoint tip publish/verify ships when `anchor_path` / `SIGIL_ANCHOR` is set. SealedStore / `SIGIL_SEAL_PASS` seals the DB file with AES-256-GCM (incremental path short of SQLCipher). Independent review and platform glue remain.
|
||||
Leora's contracted prod pass (Quarry 18a5d762) closed three remaining review blockers on this branch: rekey no longer un-revokes, dual-bootstrap enrol is exactly-one under race, and redeem rate limits are per-agent (anonymous fails isolated from success). Off-box checkpoint tip publish/verify ships when `anchor_path` / `SIGIL_ANCHOR` is set. SealedStore / `SIGIL_SEAL_PASS` seals the DB file with AES-256-GCM (incremental path short of SQLCipher). Independent review returned GO-WITH-FIXES; R1–R8/R11 fixes on this branch. Platform glue remains off.
|
||||
|
||||
**Still not production ready**, and the reason is not a to-do list:
|
||||
|
||||
@ -241,7 +256,8 @@ Leora's contracted prod pass (Quarry 18a5d762) closed three remaining review blo
|
||||
rewrite that in-DB checkpoints alone would miss after deletion. Leave it unset
|
||||
and the tip stays in-DB only.
|
||||
- **The service is unwritten.** TLS, CSRF, transport, routes. DB file seal is
|
||||
available (`SealedStore`); SQLCipher page-level encryption is not.
|
||||
available (`SealedStore`, plaintext working copy `0600`); SQLCipher
|
||||
page-level encryption is deferred (not yet available).
|
||||
- **Rate limits are per-process**, so N processes means N times the quota.
|
||||
- **No key recovery.** Rotate and old sessions die. Deliberate.
|
||||
- **Humans remain unattributable** to themselves, which is accepted and is a
|
||||
@ -251,11 +267,27 @@ The tests show the logic is correct. They do not make it safe.
|
||||
|
||||
## Next
|
||||
|
||||
1. Independent security review before this guards anything real. Two hardening
|
||||
passes made it reviewable; neither substituted for one.
|
||||
1. ~~Independent security review~~ — independent review returned GO-WITH-FIXES
|
||||
(tip c0027bb); fix PRs for R1–R8/R11 landed on `sigil-initial`. Platform glue
|
||||
remains off until integration review. See `docs/PROD_RUNBOOK.md`.
|
||||
2. ~~Anchor the checkpoint head digest outside the database~~ — done via
|
||||
`Store(..., anchor_path=)` / `SIGIL_ANCHOR` (signed tip file + `.log`).
|
||||
**Prod must set it** (see runbook).
|
||||
3. Secret broker on top — scoped, short-lived leases bound to a key fingerprint.
|
||||
Its issuance and redemption logs *are* the work-attestation substrate, so it
|
||||
is the natural second piece.
|
||||
4. Deprecate passcode login across the platform, then rotate live credentials.
|
||||
4. Deprecate passcode login across the platform, then rotate live credentials.
|
||||
|
||||
### Info / follow-ups (no code change required yet)
|
||||
|
||||
- **R9 — `issue_agent_session`:** library helper mints `agent:act` from a name
|
||||
after the caller claims to have verified. Keep internal; integration must
|
||||
never mint from headers/names. Consider renaming to `_issue_agent_session`.
|
||||
- **R10 — challenge domain separation:** signatures currently cover
|
||||
`nonce.encode()` with no context prefix. Adding
|
||||
`b"sigil-challenge-v1\0" + nonce` is not backward-compatible with in-flight
|
||||
challenges; schedule with a protocol version bump (follow-up).
|
||||
- **R12 — multi-process rate limits:** quotas are per-`Store` / per-process.
|
||||
Single writer until a shared limiter exists (documented in README + runbook).
|
||||
- **R13 — opt-in anchor:** without `SIGIL_ANCHOR`, deleting in-DB checkpoints
|
||||
allows silent history rewrite. Runbook requires the anchor on separate storage.
|
||||
|
||||
13
README.md
13
README.md
@ -55,7 +55,7 @@ python3 sigil.py sweep # prune stale rows
|
||||
python3 -m unittest discover -s tests -v
|
||||
```
|
||||
|
||||
130 tests. They cover the refusals, not just the happy paths: replayed
|
||||
158 tests. They cover the refusals, not just the happy paths: replayed
|
||||
signatures, wrong keys, expired nonces, single-use codes, scope escalation, and
|
||||
the invariant below. `test_hardening.py`, `test_production.py`, `test_anchor.py` and
|
||||
`test_sealed.py` have one test per defect found in review, so any of them
|
||||
@ -137,7 +137,9 @@ Still open, and these are the reasons not to ship it yet:
|
||||
- **The service around this is unwritten.** TLS, CSRF, session transport and the
|
||||
HTTP routes. DB encryption at rest is available via `SealedStore` /
|
||||
`SIGIL_SEAL_PASS` (AES-256-GCM file seal; not SQLCipher page-level). The
|
||||
library refuses to mint authority from a name, so a route that forgets to
|
||||
plaintext working copy is mode `0600` in a private temp dir while open;
|
||||
SQLCipher remains deferred for hostile co-resident hosts. The library
|
||||
refuses to mint authority from a name, so a route that forgets to
|
||||
authenticate fails closed, but that is only as good as the integration.
|
||||
- **Off-box anchor is opt-in.** Set `SIGIL_ANCHOR` (or `Store(..., anchor_path=)`)
|
||||
and every checkpoint also writes a signed tip outside the SQLite file.
|
||||
@ -156,4 +158,9 @@ The tests assert the logic is correct. They do not make it safe.
|
||||
## See also
|
||||
|
||||
`DESIGN.md` — the design page: rationale, flows, session stamps, the invariant
|
||||
and its reasoning, and what is deferred.
|
||||
and its reasoning, and what is deferred.
|
||||
|
||||
`docs/PROD_RUNBOOK.md` — production checklist: require `SIGIL_SEAL_PASS` and
|
||||
`SIGIL_ANCHOR` on separate storage, verify-chain alerts, integration sharp edges.
|
||||
|
||||
`requirements.txt` — `cryptography>=44.0.1,<45`; see `requirements.lock.txt` / `constraints.txt`.
|
||||
|
||||
3
constraints.txt
Normal file
3
constraints.txt
Normal file
@ -0,0 +1,3 @@
|
||||
# Exact pin matching requirements.lock.txt floor. Use with:
|
||||
# pip install -r requirements.txt -c constraints.txt
|
||||
cryptography==44.0.1
|
||||
46
docs/PROD_RUNBOOK.md
Normal file
46
docs/PROD_RUNBOOK.md
Normal file
@ -0,0 +1,46 @@
|
||||
# Sigil production runbook
|
||||
|
||||
Short checklist before Sigil guards anything real. Library-only today — do **not**
|
||||
enable platform glue until an integration review covers session minting.
|
||||
|
||||
## Required environment
|
||||
|
||||
| Variable | Purpose | Notes |
|
||||
|----------|---------|--------|
|
||||
| `SIGIL_DB` | Path to the store file | Prefer a dedicated volume; backups are sensitive (roster + audit). |
|
||||
| `SIGIL_SEAL_PASS` | AES-256-GCM seal passphrase | **Required in prod.** ≥12 chars. Without it the DB file is cleartext on disk. Seal-on still uses a plaintext `0600` working copy while the process is open — not SQLCipher. |
|
||||
| `SIGIL_ANCHOR` | Off-box signed tip path | **Required in prod.** Must be on storage **separate** from `SIGIL_DB` (different volume/host). Every checkpoint publishes here; `verify_external_anchors` / `verify-chain` must alert on failure. |
|
||||
|
||||
Optional: agent private keys stay **off** the platform host whenever possible; checkpoint signer key preferably offline.
|
||||
|
||||
## Dedicated checkpoint-signer identity
|
||||
|
||||
Enrol a separate agent id used **only** for `checkpoint` / `publish_anchor`
|
||||
(for example `sigil-checkpoint`). Do not use a working agent as the tip signer:
|
||||
compromising and permanently revoking that agent would otherwise retire its key
|
||||
(history still verifies, but ops alerting on "key retired" gets noisy). Never
|
||||
store the checkpoint signer's private key in the Sigil DB — public history only.
|
||||
|
||||
## Ops loop
|
||||
|
||||
1. Open with seal + anchor configured (CLI respects the env vars above).
|
||||
2. Periodically run chain verify **and** external-anchor verify; page on failure.
|
||||
3. After permanent revoke (compromise): bootstrap + `enrol_rekey`, then `reinstate`. Reinstate alone does not restore a wiped key.
|
||||
4. Single writer process until a shared rate-limit store exists (in-process quotas multiply across processes).
|
||||
|
||||
## Integration sharp edges (do not ignore)
|
||||
|
||||
- Never mint `issue_agent_session` from a client-supplied name or header. Only after a verified challenge-response. Prefer keeping that helper internal to verified paths.
|
||||
- Human view codes are ~40 bits; keep TTL short; never log raw codes (already hashed in audit).
|
||||
- Challenge signatures today sign the raw nonce (no domain separation). Changing the signed payload needs a coordinated version bump — tracked as a follow-up, not a silent break.
|
||||
|
||||
## Deferred
|
||||
|
||||
- SQLCipher / page-level encryption for multi-tenant co-resident hosts.
|
||||
- Domain-separated challenge payloads (`sigil-challenge-v1` prefix) — backward-incompatible without a protocol version.
|
||||
- Shared cross-process rate limiter.
|
||||
|
||||
## Dependencies
|
||||
|
||||
Install with `pip install -r requirements.txt -c constraints.txt`
|
||||
(or `--require-hashes -r requirements.lock.txt`). Floor is `cryptography>=44.0.1`.
|
||||
13
requirements.lock.txt
Normal file
13
requirements.lock.txt
Normal file
@ -0,0 +1,13 @@
|
||||
# Hash-pinned lock for Sigil runtime crypto deps.
|
||||
# Install: pip install --require-hashes -r requirements.lock.txt
|
||||
# Floor cryptography==44.0.1 addresses CVE-2024-12797 in older wheels.
|
||||
|
||||
cryptography==44.0.1 \
|
||||
--hash=sha256:53f23339864b617a3dfc2b0ac8d5c432625c80014c25caac9082314e9de56f41
|
||||
|
||||
cffi==2.1.1 \
|
||||
--hash=sha256:a931079504ecc49efed7744c476a5c343a92fabf66dec2db95edb1b2fdc770e2
|
||||
|
||||
pycparser==3.0 \
|
||||
--hash=sha256:b727414169a36b7d524c1c3e31839a521725078d7b2ff038656844266160a992
|
||||
|
||||
5
requirements.txt
Normal file
5
requirements.txt
Normal file
@ -0,0 +1,5 @@
|
||||
# Runtime dependency for Ed25519 + AES-GCM (SealedStore / optional key files).
|
||||
# Lower bound 44.0.1: older 43.x/44.0.0 wheels ship OpenSSL affected by
|
||||
# CVE-2024-12797 (GHSA-79v4-65xg-pq4g). Upper bound <45 keeps upgrades deliberate.
|
||||
# For hash-pinned install see requirements.lock.txt.
|
||||
cryptography>=44.0.1,<45
|
||||
@ -78,6 +78,10 @@ def enrol(
|
||||
_validate_agent_id(agent)
|
||||
if not store.allow(f"enrol:{agent}"):
|
||||
raise errors.RateLimited(f"too many enrolment attempts for {agent}")
|
||||
# Already enrolled (e.g. after permanent revoke): refuse *before* spending
|
||||
# the seed so ops following a wrong "re-enrol" path do not burn it.
|
||||
if store.get_agent(agent) is not None:
|
||||
raise errors.EnrolmentFailed(f"{agent} is already enrolled; use enrol_rekey")
|
||||
# Spend the secret first so a burned secret cannot be retried. The agent
|
||||
# row insert below is INSERT OR IGNORE: if two secrets race for the same
|
||||
# unenrolled agent, exactly one put_agent wins and the loser fails closed
|
||||
@ -295,6 +299,10 @@ def enrol_rekey(store: Store, agent: str, public_b64: str, secret: str) -> str:
|
||||
raise errors.UnknownAgent(f"{agent} is revoked")
|
||||
owner = store.spend_bootstrap(_hash_secret(secret))
|
||||
if owner != agent:
|
||||
# Rate-limit only failed attempts so junk cannot burn the ops path, and
|
||||
# use a dedicated rekey: bucket so exhausting enrol: cannot lock rekey.
|
||||
if not store.allow(f"rekey:{agent}"):
|
||||
raise errors.RateLimited(f"too many rekey attempts for {agent}")
|
||||
raise errors.EnrolmentFailed("bootstrap secret invalid or already spent")
|
||||
public = keys.import_public_b64(public_b64)
|
||||
fingerprint = keys.fingerprint(public)
|
||||
|
||||
@ -11,6 +11,11 @@ working copy is plaintext in memory / a temp file while open). It is a real
|
||||
working seal: the file on disk is ciphertext, and tests assert a stolen seal
|
||||
has no SQLite header and no recoverable agent rows without the passphrase.
|
||||
|
||||
The plaintext working copy is created mode 0600 inside a private temp dir.
|
||||
SQLCipher (or equivalent page-level encryption) remains a deferred prod
|
||||
follow-up for multi-tenant / hostile co-resident hosts — never treat seal-on
|
||||
as "encrypted DB" while the process is running.
|
||||
|
||||
Format on disk (version 1):
|
||||
|
||||
b"SIGILSEAL1" + 16-byte salt + 12-byte nonce + AESGCM(ciphertext)
|
||||
@ -83,6 +88,30 @@ def is_sealed_file(path: str) -> bool:
|
||||
return handle.read(len(MAGIC)) == MAGIC
|
||||
|
||||
|
||||
def _write_private_file(path: str, data: bytes) -> None:
|
||||
"""Write `data` to `path` with mode 0600, fsync, unlink on failure."""
|
||||
flags = os.O_WRONLY | os.O_CREAT | os.O_TRUNC
|
||||
fd = os.open(path, flags, 0o600)
|
||||
try:
|
||||
with os.fdopen(fd, "wb") as handle:
|
||||
fd = -1 # ownership transferred
|
||||
handle.write(data)
|
||||
handle.flush()
|
||||
os.fsync(handle.fileno())
|
||||
os.chmod(path, 0o600)
|
||||
except Exception:
|
||||
if fd >= 0:
|
||||
try:
|
||||
os.close(fd)
|
||||
except OSError:
|
||||
pass
|
||||
try:
|
||||
os.unlink(path)
|
||||
except OSError:
|
||||
pass
|
||||
raise
|
||||
|
||||
|
||||
class SealedStore(Store):
|
||||
"""Store whose on-disk file is AES-GCM sealed.
|
||||
|
||||
@ -107,6 +136,8 @@ class SealedStore(Store):
|
||||
self.seal_path = path
|
||||
self._passphrase = passphrase
|
||||
self._tmpdir = tempfile.TemporaryDirectory(prefix="sigil-seal-")
|
||||
# TemporaryDirectory is already 0700; pin the dir mode explicitly.
|
||||
os.chmod(self._tmpdir.name, 0o700)
|
||||
self._plain_path = os.path.join(self._tmpdir.name, "store.db")
|
||||
if os.path.exists(path):
|
||||
with open(path, "rb") as handle:
|
||||
@ -117,9 +148,16 @@ class SealedStore(Store):
|
||||
"refusing to treat a cleartext DB as sealed"
|
||||
)
|
||||
plain = unseal_bytes(blob, passphrase)
|
||||
with open(self._plain_path, "wb") as handle:
|
||||
handle.write(plain)
|
||||
_write_private_file(self._plain_path, plain)
|
||||
else:
|
||||
# Touch an empty 0600 file so SQLite does not create a world-readable DB.
|
||||
_write_private_file(self._plain_path, b"")
|
||||
super().__init__(self._plain_path, anchor_path=anchor_path)
|
||||
# Re-pin after SQLite open (WAL/SHM sidecars inherit umask otherwise).
|
||||
try:
|
||||
os.chmod(self._plain_path, 0o600)
|
||||
except OSError:
|
||||
pass
|
||||
|
||||
def flush(self) -> str:
|
||||
"""Checkpoint WAL, read plaintext bytes, write sealed blob to seal_path."""
|
||||
@ -130,14 +168,24 @@ class SealedStore(Store):
|
||||
dest_path = os.path.join(self._tmpdir.name, "export.db")
|
||||
if os.path.exists(dest_path):
|
||||
os.unlink(dest_path)
|
||||
# Pre-create 0600 so SQLite backup does not leave a world-readable export.
|
||||
_write_private_file(dest_path, b"")
|
||||
dest = sqlite3.connect(dest_path)
|
||||
try:
|
||||
self.conn.backup(dest)
|
||||
dest.commit()
|
||||
finally:
|
||||
dest.close()
|
||||
try:
|
||||
os.chmod(dest_path, 0o600)
|
||||
except OSError:
|
||||
pass
|
||||
with open(dest_path, "rb") as handle:
|
||||
plain = handle.read()
|
||||
try:
|
||||
os.unlink(dest_path)
|
||||
except OSError:
|
||||
pass
|
||||
blob = seal_bytes(plain, self._passphrase)
|
||||
directory = os.path.dirname(os.path.abspath(self.seal_path))
|
||||
if directory:
|
||||
|
||||
@ -105,9 +105,16 @@ class GrantedCode:
|
||||
agent session (issue_code) or a valid code (redeem_code).
|
||||
"""
|
||||
|
||||
__slots__ = ("agent", "scopes", "ttl", "_sigil_internal")
|
||||
__slots__ = ("agent", "scopes", "ttl", "fingerprint", "_sigil_internal")
|
||||
|
||||
def __init__(self, agent: str, scopes: List[str], ttl: int, _internal=None):
|
||||
def __init__(
|
||||
self,
|
||||
agent: str,
|
||||
scopes: List[str],
|
||||
ttl: int,
|
||||
_internal=None,
|
||||
fingerprint: Optional[str] = None,
|
||||
):
|
||||
if _internal is not _ISSUER:
|
||||
raise errors.SigilError(
|
||||
"GrantedCode cannot be constructed directly; it comes from"
|
||||
@ -116,6 +123,7 @@ class GrantedCode:
|
||||
self.agent = agent
|
||||
self.scopes = list(scopes)
|
||||
self.ttl = ttl
|
||||
self.fingerprint = fingerprint
|
||||
self._sigil_internal = _internal
|
||||
|
||||
def __repr__(self) -> str:
|
||||
@ -151,6 +159,7 @@ def issue_view_session(store: Store, grant: "GrantedCode") -> dict:
|
||||
granted,
|
||||
expires,
|
||||
granted_by=agent,
|
||||
fingerprint=grant.fingerprint,
|
||||
)
|
||||
store.record(
|
||||
"session.issued",
|
||||
@ -176,7 +185,14 @@ def issue_view_session(store: Store, grant: "GrantedCode") -> dict:
|
||||
|
||||
|
||||
def resolve(store: Store, token: str) -> dict:
|
||||
"""Look up a session and reject it if revoked or expired."""
|
||||
"""Look up a session and reject it if revoked, expired, or the agent died.
|
||||
|
||||
The session row alone is not enough: an in-flight verify/issue_code can
|
||||
insert a session or code after revoke/rekey has already run. Re-check the
|
||||
agent on every resolve so a race-surviving bearer cannot keep acting.
|
||||
"""
|
||||
from . import identity as _identity
|
||||
|
||||
row = store.get_session(hash_token(token))
|
||||
if row is None:
|
||||
raise errors.SessionUnknown("no such session")
|
||||
@ -184,6 +200,24 @@ def resolve(store: Store, token: str) -> dict:
|
||||
raise errors.SessionExpired("session revoked")
|
||||
if time.time() >= row["expires_at"]:
|
||||
raise errors.SessionExpired("session expired")
|
||||
|
||||
agent = row["on_behalf_of"]
|
||||
agent_row = store.get_agent(agent)
|
||||
if agent_row is None or _identity.is_revoked(agent_row):
|
||||
raise errors.SessionExpired("agent revoked")
|
||||
if not _identity.has_live_key(agent_row):
|
||||
raise errors.SessionExpired("agent has no live key")
|
||||
# Every session that carries a fingerprint is bound to that key epoch.
|
||||
# After rekey, race-surviving agent:act *and* human_via_agent bearers die
|
||||
# even if revoke_sessions missed them (N7).
|
||||
sess_fp = row["fingerprint"] or ""
|
||||
live_fp = agent_row["fingerprint"] or ""
|
||||
if sess_fp or row["kind"] == AGENT:
|
||||
# Agent sessions always require a matching fingerprint; human sessions
|
||||
# require a match whenever a fingerprint was recorded at mint time.
|
||||
if not sess_fp or sess_fp != live_fp:
|
||||
raise errors.SessionExpired("session key epoch no longer current")
|
||||
|
||||
return {
|
||||
"token": token,
|
||||
"kind": row["kind"],
|
||||
|
||||
233
sigil/store.py
233
sigil/store.py
@ -36,6 +36,20 @@ CREATE TABLE IF NOT EXISTS agents (
|
||||
revoked_reason TEXT
|
||||
);
|
||||
|
||||
-- Append-only public-key history. Revoke and rekey *retire* a key instead of
|
||||
-- destroying it, so historical checkpoints and off-box anchors stay verifiable.
|
||||
-- Never stores private keys.
|
||||
CREATE TABLE IF NOT EXISTS agent_keys (
|
||||
agent TEXT NOT NULL,
|
||||
fingerprint TEXT NOT NULL,
|
||||
pubkey BLOB NOT NULL,
|
||||
active_from INTEGER NOT NULL,
|
||||
retired_at INTEGER,
|
||||
retired_reason TEXT,
|
||||
PRIMARY KEY (agent, fingerprint)
|
||||
);
|
||||
CREATE INDEX IF NOT EXISTS agent_keys_agent ON agent_keys(agent, active_from);
|
||||
|
||||
-- Enrolment bootstrap secrets. Spent secrets are deleted, not marked, so a
|
||||
-- spent secret cannot be replayed even if the table is read.
|
||||
CREATE TABLE IF NOT EXISTS bootstrap (
|
||||
@ -77,7 +91,11 @@ CREATE TABLE IF NOT EXISTS codes (
|
||||
redeem_expires_at INTEGER NOT NULL, -- short: the code is a key
|
||||
view_ttl INTEGER NOT NULL, -- the door opens a full hour, but only
|
||||
-- once the key is used
|
||||
used_at INTEGER
|
||||
used_at INTEGER,
|
||||
-- Fingerprint of the agent key at issue time. Redeem refuses if the agent
|
||||
-- has since rekeyed (or been wiped), so an in-flight issue_code cannot
|
||||
-- outlive a revoke/rekey that lands before put_code commits.
|
||||
fingerprint TEXT
|
||||
);
|
||||
|
||||
-- Append-only, hash-chained audit log. The chain is what makes this an
|
||||
@ -130,6 +148,9 @@ RATE_LIMITS = {
|
||||
"challenge": (30, 60), # 30 per 60s
|
||||
"verify": (60, 60),
|
||||
"enrol": (10, 300),
|
||||
# Separate from enrol: junk pre-auth enrol_rekey must not lock out ops
|
||||
# rotating a compromised agent during incident response.
|
||||
"rekey": (10, 300),
|
||||
"code_issue": (20, 300),
|
||||
"code_redeem": (20, 60),
|
||||
"code_redeem_fail": (10, 300),
|
||||
@ -217,6 +238,22 @@ class Store:
|
||||
self.conn.execute(
|
||||
"UPDATE codes SET view_ttl = 3600 WHERE view_ttl IS NULL"
|
||||
)
|
||||
code_cols = {
|
||||
row["name"] for row in self.conn.execute("PRAGMA table_info(codes)")
|
||||
}
|
||||
if "fingerprint" not in code_cols:
|
||||
self.conn.execute("ALTER TABLE codes ADD COLUMN fingerprint TEXT")
|
||||
# Backfill agent_keys from current agents rows (live or wiped-empty skipped).
|
||||
for row in self.conn.execute(
|
||||
"SELECT agent, pubkey, fingerprint, enrolled_at FROM agents"
|
||||
" WHERE fingerprint IS NOT NULL AND fingerprint != ''"
|
||||
" AND length(pubkey) = 32"
|
||||
):
|
||||
self.conn.execute(
|
||||
"INSERT OR IGNORE INTO agent_keys"
|
||||
" (agent, fingerprint, pubkey, active_from) VALUES (?, ?, ?, ?)",
|
||||
(row["agent"], row["fingerprint"], row["pubkey"], row["enrolled_at"]),
|
||||
)
|
||||
self.conn.commit()
|
||||
|
||||
def close(self) -> None:
|
||||
@ -253,12 +290,19 @@ class Store:
|
||||
which preserves revocation columns.
|
||||
"""
|
||||
with self._write_lock:
|
||||
now = int(time.time())
|
||||
cur = self.conn.execute(
|
||||
"INSERT OR IGNORE INTO agents"
|
||||
" (agent, pubkey, fingerprint, enrolled_at)"
|
||||
" VALUES (?, ?, ?, ?)",
|
||||
(agent, pubkey, fingerprint, int(time.time())),
|
||||
(agent, pubkey, fingerprint, now),
|
||||
)
|
||||
if cur.rowcount == 1:
|
||||
self.conn.execute(
|
||||
"INSERT OR IGNORE INTO agent_keys"
|
||||
" (agent, fingerprint, pubkey, active_from) VALUES (?, ?, ?, ?)",
|
||||
(agent, fingerprint, pubkey, now),
|
||||
)
|
||||
self.conn.commit()
|
||||
return cur.rowcount == 1
|
||||
|
||||
@ -270,10 +314,33 @@ class Store:
|
||||
used to un-revoke. An UPDATE of only the key columns keeps the ban intact.
|
||||
"""
|
||||
with self._write_lock:
|
||||
now = int(time.time())
|
||||
prev = self.conn.execute(
|
||||
"SELECT pubkey, fingerprint, enrolled_at FROM agents WHERE agent = ?",
|
||||
(agent,),
|
||||
).fetchone()
|
||||
cur = self.conn.execute(
|
||||
"UPDATE agents SET pubkey = ?, fingerprint = ? WHERE agent = ?",
|
||||
(pubkey, fingerprint, agent),
|
||||
)
|
||||
if cur.rowcount == 1 and prev is not None:
|
||||
old_fp = prev["fingerprint"] or ""
|
||||
if old_fp and len(prev["pubkey"] or b"") == 32:
|
||||
self.conn.execute(
|
||||
"INSERT OR IGNORE INTO agent_keys"
|
||||
" (agent, fingerprint, pubkey, active_from) VALUES (?, ?, ?, ?)",
|
||||
(agent, old_fp, prev["pubkey"], prev["enrolled_at"] or now),
|
||||
)
|
||||
self.conn.execute(
|
||||
"UPDATE agent_keys SET retired_at = ?, retired_reason = ?"
|
||||
" WHERE agent = ? AND fingerprint = ? AND retired_at IS NULL",
|
||||
(now, "rekey", agent, old_fp),
|
||||
)
|
||||
self.conn.execute(
|
||||
"INSERT OR IGNORE INTO agent_keys"
|
||||
" (agent, fingerprint, pubkey, active_from) VALUES (?, ?, ?, ?)",
|
||||
(agent, fingerprint, pubkey, now),
|
||||
)
|
||||
self.conn.commit()
|
||||
return cur.rowcount == 1
|
||||
|
||||
@ -285,20 +352,39 @@ class Store:
|
||||
"""Suspend an agent. `until` of None is permanent.
|
||||
|
||||
A permanent revoke clears the enrolled public key so `reinstate` alone
|
||||
cannot restore a compromised identity — ops must bootstrap + rekey
|
||||
(or enrol) before the agent can authenticate again. Timed suspension
|
||||
keeps the key: when the pause lapses, the same key works.
|
||||
cannot restore a compromised identity — ops must bootstrap + enrol_rekey
|
||||
before the agent can authenticate again. Timed suspension keeps the
|
||||
key: when the pause lapses, the same key works.
|
||||
|
||||
Returns how many live sessions were killed.
|
||||
"""
|
||||
with self._write_lock:
|
||||
if until is None:
|
||||
# Permanent: wipe key material. Empty fingerprint means
|
||||
# "pending rekey" — see identity.has_live_key.
|
||||
# Permanent: retire the public key into agent_keys (so historical
|
||||
# checkpoints stay verifiable), then wipe the live slot.
|
||||
now = int(time.time())
|
||||
prev = self.conn.execute(
|
||||
"SELECT pubkey, fingerprint, enrolled_at FROM agents WHERE agent = ?",
|
||||
(agent,),
|
||||
).fetchone()
|
||||
if prev is not None:
|
||||
old_fp = prev["fingerprint"] or ""
|
||||
if old_fp and len(prev["pubkey"] or b"") == 32:
|
||||
self.conn.execute(
|
||||
"INSERT OR IGNORE INTO agent_keys"
|
||||
" (agent, fingerprint, pubkey, active_from)"
|
||||
" VALUES (?, ?, ?, ?)",
|
||||
(agent, old_fp, prev["pubkey"], prev["enrolled_at"] or now),
|
||||
)
|
||||
self.conn.execute(
|
||||
"UPDATE agent_keys SET retired_at = ?, retired_reason = ?"
|
||||
" WHERE agent = ? AND fingerprint = ? AND retired_at IS NULL",
|
||||
(now, reason or "permanent revoke", agent, old_fp),
|
||||
)
|
||||
self.conn.execute(
|
||||
"UPDATE agents SET revoked_at = ?, revoked_until = ?, revoked_reason = ?,"
|
||||
" pubkey = ?, fingerprint = ? WHERE agent = ?",
|
||||
(int(time.time()), until, reason, b"", "", agent),
|
||||
(now, until, reason, b"", "", agent),
|
||||
)
|
||||
else:
|
||||
self.conn.execute(
|
||||
@ -335,8 +421,8 @@ class Store:
|
||||
with self._write_lock:
|
||||
now = int(time.time())
|
||||
self.conn.execute(
|
||||
"UPDATE bootstrap SET spent_at = ? WHERE agent = ? AND spent_at IS NULL",
|
||||
(now, agent),
|
||||
"DELETE FROM bootstrap WHERE agent = ? AND spent_at IS NULL",
|
||||
(agent,),
|
||||
)
|
||||
self.conn.execute(
|
||||
"INSERT INTO bootstrap (secret_hash, agent, created_at) VALUES (?, ?, ?)",
|
||||
@ -345,7 +431,7 @@ class Store:
|
||||
self.conn.commit()
|
||||
|
||||
def clear_bootstrap_for_agent(self, agent: str, *, keep_hash: Optional[str] = None) -> int:
|
||||
"""Spend (or delete) every unspent bootstrap secret for an agent.
|
||||
"""Delete every unspent bootstrap secret for an agent.
|
||||
|
||||
After a successful enrol / enrol_rekey, leftover seeds must not remain
|
||||
usable. keep_hash, when set, is left alone (already consumed by CAS).
|
||||
@ -353,14 +439,14 @@ class Store:
|
||||
with self._write_lock:
|
||||
if keep_hash is None:
|
||||
cur = self.conn.execute(
|
||||
"UPDATE bootstrap SET spent_at = ? WHERE agent = ? AND spent_at IS NULL",
|
||||
(int(time.time()), agent),
|
||||
"DELETE FROM bootstrap WHERE agent = ? AND spent_at IS NULL",
|
||||
(agent,),
|
||||
)
|
||||
else:
|
||||
cur = self.conn.execute(
|
||||
"UPDATE bootstrap SET spent_at = ? WHERE agent = ? AND spent_at IS NULL"
|
||||
"DELETE FROM bootstrap WHERE agent = ? AND spent_at IS NULL"
|
||||
" AND secret_hash != ?",
|
||||
(int(time.time()), agent, keep_hash),
|
||||
(agent, keep_hash),
|
||||
)
|
||||
self.conn.commit()
|
||||
return cur.rowcount
|
||||
@ -368,23 +454,26 @@ class Store:
|
||||
def spend_bootstrap(self, secret_hash: str) -> Optional[str]:
|
||||
"""Atomically consume an enrolment secret. Returns the agent or None.
|
||||
|
||||
Compare-and-set: the UPDATE only matches an unspent row, so exactly one
|
||||
concurrent caller can win. Check-then-update would let a secret enrol
|
||||
more than one key.
|
||||
Compare-and-set then delete: exactly one concurrent caller can win, and
|
||||
the spent row is removed so a table dump cannot offline-crack a
|
||||
low-entropy custom secret. Matches the schema comment.
|
||||
"""
|
||||
with self._write_lock:
|
||||
row = self.conn.execute(
|
||||
"SELECT agent FROM bootstrap WHERE secret_hash = ? AND spent_at IS NULL",
|
||||
(secret_hash,),
|
||||
).fetchone()
|
||||
if row is None:
|
||||
return None
|
||||
agent = row["agent"]
|
||||
cur = self.conn.execute(
|
||||
"UPDATE bootstrap SET spent_at = ? WHERE secret_hash = ?"
|
||||
" AND spent_at IS NULL",
|
||||
(int(time.time()), secret_hash),
|
||||
"DELETE FROM bootstrap WHERE secret_hash = ? AND spent_at IS NULL",
|
||||
(secret_hash,),
|
||||
)
|
||||
self.conn.commit()
|
||||
if cur.rowcount != 1:
|
||||
return None
|
||||
row = self.conn.execute(
|
||||
"SELECT agent FROM bootstrap WHERE secret_hash = ?", (secret_hash,)
|
||||
).fetchone()
|
||||
return row["agent"] if row else None
|
||||
return agent
|
||||
|
||||
# -- nonces ------------------------------------------------------------
|
||||
def put_nonce(self, nonce: str, agent: str, expires_at: int) -> None:
|
||||
@ -492,10 +581,11 @@ class Store:
|
||||
scopes: List[str],
|
||||
redeem_expires_at: int,
|
||||
view_ttl: int,
|
||||
fingerprint: Optional[str] = None,
|
||||
) -> None:
|
||||
self._exec(
|
||||
"INSERT INTO codes (code_hash, agent, scopes, issued_at, redeem_expires_at,"
|
||||
" view_ttl) VALUES (?, ?, ?, ?, ?, ?)",
|
||||
" view_ttl, fingerprint) VALUES (?, ?, ?, ?, ?, ?, ?)",
|
||||
(
|
||||
code_hash,
|
||||
agent,
|
||||
@ -503,6 +593,7 @@ class Store:
|
||||
int(time.time()),
|
||||
redeem_expires_at,
|
||||
view_ttl,
|
||||
fingerprint,
|
||||
),
|
||||
)
|
||||
|
||||
@ -510,14 +601,17 @@ class Store:
|
||||
return self._one("SELECT * FROM codes WHERE code_hash = ?", (code_hash,))
|
||||
|
||||
def claim_code(self, code_hash: str) -> bool:
|
||||
"""Consume a one-time code exactly once. False if already used.
|
||||
"""Consume a one-time code exactly once. False if already used or expired.
|
||||
|
||||
Atomic, so a code cannot mint two sessions under a race.
|
||||
Atomic CAS includes the expiry predicate so a race at the TTL boundary
|
||||
cannot mint a session slightly after redeem_expires_at.
|
||||
"""
|
||||
with self._write_lock:
|
||||
now = int(time.time())
|
||||
cur = self.conn.execute(
|
||||
"UPDATE codes SET used_at = ? WHERE code_hash = ? AND used_at IS NULL",
|
||||
(int(time.time()), code_hash),
|
||||
"UPDATE codes SET used_at = ? WHERE code_hash = ? AND used_at IS NULL"
|
||||
" AND redeem_expires_at > ?",
|
||||
(now, code_hash, now),
|
||||
)
|
||||
self.conn.commit()
|
||||
return cur.rowcount == 1
|
||||
@ -646,7 +740,26 @@ class Store:
|
||||
}
|
||||
)
|
||||
return out
|
||||
def get_key_by_fingerprint(self, agent: str, fingerprint: str):
|
||||
"""Return a historical (or live) public key row for verify.
|
||||
|
||||
Prefers agent_keys history. Falls back to the live agents row when the
|
||||
fingerprint still matches (covers DBs created before agent_keys).
|
||||
Never returns private key material — pubkey only.
|
||||
"""
|
||||
row = self._one(
|
||||
"SELECT * FROM agent_keys WHERE agent = ? AND fingerprint = ?",
|
||||
(agent, fingerprint),
|
||||
)
|
||||
if row is not None:
|
||||
return row
|
||||
live = self.get_agent(agent)
|
||||
if live is not None and (live["fingerprint"] or "") == fingerprint:
|
||||
return live
|
||||
return None
|
||||
|
||||
# -- checkpoints -------------------------------------------------------
|
||||
|
||||
@staticmethod
|
||||
def _checkpoint_payload(at: int, label: str, head: str, events: int,
|
||||
signer: str, fingerprint: str) -> bytes:
|
||||
@ -675,8 +788,19 @@ class Store:
|
||||
|
||||
public = _keys.public_from_private(private)
|
||||
fingerprint = _keys.fingerprint(public)
|
||||
if self.get_agent(signer) is None:
|
||||
from . import identity as _identity
|
||||
|
||||
agent_row = self.get_agent(signer)
|
||||
if agent_row is None:
|
||||
raise ValueError(f"{signer} is not enrolled, so cannot sign a checkpoint")
|
||||
if _identity.is_revoked(agent_row):
|
||||
raise ValueError(f"{signer} is revoked, so cannot sign a checkpoint")
|
||||
enrolled_fp = agent_row["fingerprint"] or ""
|
||||
if not enrolled_fp or fingerprint != enrolled_fp:
|
||||
raise ValueError(
|
||||
f"private key fingerprint does not match enrolled key for {signer}; "
|
||||
"refusing to sign a checkpoint with a stranger key"
|
||||
)
|
||||
with self._write_lock:
|
||||
last = self.conn.execute(
|
||||
"SELECT id, hash FROM events ORDER BY id DESC LIMIT 1"
|
||||
@ -740,17 +864,20 @@ class Store:
|
||||
if not notices or notices[-1].get("head") != tip.get("head"):
|
||||
notices = list(notices) + [tip]
|
||||
results = []
|
||||
warnings = []
|
||||
for notice in notices:
|
||||
agent_row = self.get_agent(notice["signer"])
|
||||
if agent_row is None:
|
||||
if self.get_agent(notice["signer"]) is None:
|
||||
results.append({"ok": False, "head": notice.get("head"),
|
||||
"reason": f"signer {notice['signer']} no longer enrolled"})
|
||||
continue
|
||||
if agent_row["fingerprint"] != notice.get("fingerprint"):
|
||||
key_row = self.get_key_by_fingerprint(
|
||||
notice["signer"], notice.get("fingerprint") or ""
|
||||
)
|
||||
if key_row is None:
|
||||
results.append({"ok": False, "head": notice.get("head"),
|
||||
"reason": "signer key changed since the tip was published"})
|
||||
"reason": "signer key for this fingerprint is unknown"})
|
||||
continue
|
||||
public = _keys.from_public_bytes(agent_row["pubkey"])
|
||||
public = _keys.from_public_bytes(key_row["pubkey"])
|
||||
checked = _anchor.verify_notice(notice, public)
|
||||
if not checked["ok"]:
|
||||
results.append({"ok": False, "head": notice.get("head"),
|
||||
@ -770,11 +897,17 @@ class Store:
|
||||
"(database rewrite cannot erase the off-box tip)",
|
||||
})
|
||||
continue
|
||||
results.append({"ok": True, "head": notice["head"],
|
||||
"events": notice["events"], "signer": notice["signer"]})
|
||||
entry = {"ok": True, "head": notice["head"],
|
||||
"events": notice["events"], "signer": notice["signer"]}
|
||||
retired_at = key_row["retired_at"] if "retired_at" in key_row.keys() else None
|
||||
if retired_at is not None and notice.get("at", 0) > retired_at:
|
||||
entry["flagged"] = True
|
||||
entry["reason"] = "notice timestamp is after the signing key was retired"
|
||||
warnings.append(entry)
|
||||
results.append(entry)
|
||||
bad = [r for r in results if not r["ok"]]
|
||||
return {"ok": not bad, "anchors": len(results), "failures": bad,
|
||||
"tip": tip, "path": self.anchor_path}
|
||||
"warnings": warnings, "tip": tip, "path": self.anchor_path}
|
||||
|
||||
def checkpoints(self) -> List[Dict[str, Any]]:
|
||||
rows = self.conn.execute("SELECT * FROM checkpoints ORDER BY id ASC").fetchall()
|
||||
@ -790,22 +923,23 @@ class Store:
|
||||
from . import keys as _keys
|
||||
|
||||
results = []
|
||||
warnings = []
|
||||
for row in self.conn.execute("SELECT * FROM checkpoints ORDER BY id ASC"):
|
||||
agent_row = self.get_agent(row["signer"])
|
||||
if agent_row is None:
|
||||
if self.get_agent(row["signer"]) is None:
|
||||
results.append({"id": row["id"], "ok": False,
|
||||
"reason": f"signer {row['signer']} no longer enrolled"})
|
||||
continue
|
||||
if agent_row["fingerprint"] != row["fingerprint"]:
|
||||
key_row = self.get_key_by_fingerprint(row["signer"], row["fingerprint"])
|
||||
if key_row is None:
|
||||
results.append({"id": row["id"], "ok": False,
|
||||
"reason": "signer key changed since the checkpoint was made"})
|
||||
"reason": "signer key for this fingerprint is unknown"})
|
||||
continue
|
||||
payload = self._checkpoint_payload(
|
||||
row["at"], row["label"], row["head"], row["events"],
|
||||
row["signer"], row["fingerprint"],
|
||||
)
|
||||
try:
|
||||
public = _keys.from_public_bytes(agent_row["pubkey"])
|
||||
public = _keys.from_public_bytes(key_row["pubkey"])
|
||||
# verify() returns None on success and raises on failure, so it
|
||||
# must be called, not inspected.
|
||||
public.verify(base64.b64decode(row["signature"]), payload)
|
||||
@ -823,9 +957,16 @@ class Store:
|
||||
results.append({"id": row["id"], "ok": False,
|
||||
"reason": "chain head at that position no longer matches"})
|
||||
continue
|
||||
results.append({"id": row["id"], "ok": True})
|
||||
entry = {"id": row["id"], "ok": True}
|
||||
retired_at = key_row["retired_at"] if "retired_at" in key_row.keys() else None
|
||||
if retired_at is not None and row["at"] > retired_at:
|
||||
entry["flagged"] = True
|
||||
entry["reason"] = "checkpoint timestamp is after the signing key was retired"
|
||||
warnings.append(entry)
|
||||
results.append(entry)
|
||||
bad = [r for r in results if not r["ok"]]
|
||||
return {"ok": not bad, "checkpoints": len(results), "failures": bad}
|
||||
return {"ok": not bad, "checkpoints": len(results),
|
||||
"failures": bad, "warnings": warnings}
|
||||
|
||||
# -- housekeeping ------------------------------------------------------
|
||||
def sweep(self) -> Dict[str, int]:
|
||||
|
||||
@ -91,9 +91,18 @@ def issue_code(
|
||||
if not store.allow(f"code_issue:{agent}", limit=limit, window=window):
|
||||
raise errors.RateLimited(f"too many view codes for {agent}")
|
||||
|
||||
from . import identity as _identity
|
||||
|
||||
agent_row = store.get_agent(agent)
|
||||
if agent_row is None:
|
||||
raise errors.UnknownAgent(f"{agent} is not enrolled")
|
||||
# Re-check after resolve: revoke/rekey can land between resolve and put_code.
|
||||
if _identity.is_revoked(agent_row) or not _identity.has_live_key(agent_row):
|
||||
raise errors.UnknownAgent(f"{agent} is revoked or has no live key")
|
||||
live_fp = agent_row["fingerprint"] or ""
|
||||
sess_fp = principal.get("fingerprint") or ""
|
||||
if not live_fp or sess_fp != live_fp:
|
||||
raise errors.SessionExpired("session key epoch no longer current")
|
||||
|
||||
granted = list(scopes if scopes is not None else sessions.HUMAN_VIEW_SCOPES)
|
||||
# Refuse to bake write or act scopes into a human code at issue time.
|
||||
@ -115,12 +124,13 @@ def issue_code(
|
||||
granted,
|
||||
redeem_expires_at=now + code_ttl,
|
||||
view_ttl=view_ttl,
|
||||
fingerprint=live_fp,
|
||||
)
|
||||
store.record(
|
||||
"view_code.issued",
|
||||
actor=agent,
|
||||
subject=f"{agent}'s human",
|
||||
fingerprint=agent_row["fingerprint"],
|
||||
fingerprint=live_fp,
|
||||
detail={
|
||||
"code_ref": _code_ref(code),
|
||||
"scopes": granted,
|
||||
@ -134,7 +144,6 @@ def issue_code(
|
||||
"code_ttl": code_ttl,
|
||||
"view_ttl": view_ttl,
|
||||
"redeem_expires_at": now + code_ttl,
|
||||
"view_ttl": view_ttl,
|
||||
}
|
||||
|
||||
|
||||
@ -155,19 +164,25 @@ def redeem_code(store: Store, code: str) -> dict:
|
||||
|
||||
agent = row["agent"]
|
||||
|
||||
# Defense in depth: revoke/rekey already invalidate unused codes, but refuse
|
||||
# here too if the agent is currently revoked or missing so a missed
|
||||
# invalidation cannot mint a human_via_agent session.
|
||||
# Defense in depth: refuse if agent revoked, wiped, or rekeyed since issue.
|
||||
from . import identity as _identity
|
||||
agent_row = store.get_agent(agent)
|
||||
if agent_row is None or _identity.is_revoked(agent_row):
|
||||
code_fp = row["fingerprint"] if "fingerprint" in row.keys() else None
|
||||
live_fp = (agent_row["fingerprint"] or "") if agent_row is not None else ""
|
||||
bad = (
|
||||
agent_row is None
|
||||
or _identity.is_revoked(agent_row)
|
||||
or not _identity.has_live_key(agent_row)
|
||||
or (code_fp is not None and code_fp != live_fp)
|
||||
)
|
||||
if bad:
|
||||
if not store.allow(f"code_redeem_fail:{agent}"):
|
||||
raise errors.RateLimited("too many redemption attempts")
|
||||
store.record(
|
||||
"view_code.redeem_refused",
|
||||
actor=agent,
|
||||
subject=f"{agent}'s human",
|
||||
detail={"reason": "agent revoked or unknown"},
|
||||
detail={"reason": "agent revoked, wiped, or rekeyed"},
|
||||
)
|
||||
raise errors.CodeInvalid("that code is not redeemable")
|
||||
|
||||
@ -216,6 +231,7 @@ def redeem_code(store: Store, code: str) -> dict:
|
||||
_json.loads(row["scopes"]),
|
||||
row["view_ttl"],
|
||||
sessions._ISSUER,
|
||||
fingerprint=live_fp,
|
||||
)
|
||||
return sessions.issue_view_session(store, grant)
|
||||
|
||||
|
||||
@ -359,4 +359,216 @@ class TestAgentIdValidation(unittest.TestCase):
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
unittest.main()
|
||||
|
||||
|
||||
class TestClaimCodeExpiryCAS(unittest.TestCase):
|
||||
"""R6: claim_code CAS must include redeem_expires_at predicate."""
|
||||
|
||||
def test_claim_code_after_expiry_returns_false(self):
|
||||
store, private = helpers.fresh()
|
||||
issued = helpers.issue(store, private, code_ttl=1, view_ttl=60)
|
||||
digest = viewcodes._hash_code(issued["code"])
|
||||
# Force expiry in the row.
|
||||
store.conn.execute(
|
||||
"UPDATE codes SET redeem_expires_at = ? WHERE code_hash = ?",
|
||||
(int(time.time()) - 1, digest),
|
||||
)
|
||||
store.conn.commit()
|
||||
self.assertFalse(store.claim_code(digest))
|
||||
row = store.get_code(digest)
|
||||
self.assertIsNone(row["used_at"])
|
||||
|
||||
|
||||
class TestEnrolRekeyRateLimit(unittest.TestCase):
|
||||
"""N3: enrol_rekey uses a dedicated rekey: bucket; junk cannot lock ops."""
|
||||
|
||||
def test_exhausting_enrol_bucket_does_not_block_rekey(self):
|
||||
store, _ = helpers.fresh("limited")
|
||||
for _ in range(20):
|
||||
store.allow("enrol:limited")
|
||||
secret = identity.seed_bootstrap(store, "limited")
|
||||
# Valid rekey must succeed even if enrol: is exhausted.
|
||||
identity.enrol_rekey(
|
||||
store, "limited", keys.export_public_b64(keys.generate()[1]), secret
|
||||
)
|
||||
|
||||
def test_junk_rekey_fills_rekey_bucket_not_enrol(self):
|
||||
store, _ = helpers.fresh("victim")
|
||||
for _ in range(15):
|
||||
try:
|
||||
identity.enrol_rekey(
|
||||
store, "victim", keys.export_public_b64(keys.generate()[1]), "junk"
|
||||
)
|
||||
except (errors.EnrolmentFailed, errors.RateLimited):
|
||||
pass
|
||||
# enrol: bucket still usable
|
||||
self.assertTrue(store.allow("enrol:victim"))
|
||||
# Valid ops rekey still works: failures charge rekey: only after bad secret,
|
||||
# and a valid secret bypasses the limiter.
|
||||
secret = identity.seed_bootstrap(store, "victim")
|
||||
identity.enrol_rekey(
|
||||
store, "victim", keys.export_public_b64(keys.generate()[1]), secret
|
||||
)
|
||||
|
||||
def test_failed_rekey_eventually_rate_limited(self):
|
||||
store, _ = helpers.fresh("hammer")
|
||||
hit = False
|
||||
for _ in range(30):
|
||||
try:
|
||||
identity.enrol_rekey(
|
||||
store, "hammer", keys.export_public_b64(keys.generate()[1]), "nope"
|
||||
)
|
||||
except errors.RateLimited:
|
||||
hit = True
|
||||
break
|
||||
except errors.EnrolmentFailed:
|
||||
pass
|
||||
self.assertTrue(hit, "junk rekey attempts should eventually RateLimited")
|
||||
|
||||
|
||||
|
||||
class TestBootstrapRowDeletedOnSpend(unittest.TestCase):
|
||||
"""R8: spent bootstrap secrets are deleted, not retained as hashes."""
|
||||
|
||||
def test_spend_deletes_the_row(self):
|
||||
store = Store(":memory:")
|
||||
secret = identity.seed_bootstrap(store, "dave")
|
||||
digest = __import__("hashlib").sha256(secret.encode()).hexdigest()
|
||||
self.assertIsNotNone(store.conn.execute(
|
||||
"SELECT 1 FROM bootstrap WHERE secret_hash = ?", (digest,)
|
||||
).fetchone())
|
||||
owner = store.spend_bootstrap(digest)
|
||||
self.assertEqual(owner, "dave")
|
||||
self.assertIsNone(store.conn.execute(
|
||||
"SELECT 1 FROM bootstrap WHERE secret_hash = ?", (digest,)
|
||||
).fetchone())
|
||||
|
||||
|
||||
class TestResolveRechecksAgentState(unittest.TestCase):
|
||||
"""N1: resolve must refuse sessions that outran revoke/rekey."""
|
||||
|
||||
def _plant_agent_session(self, store, agent, fingerprint):
|
||||
"""Insert a live session as if put_session won a race against revoke."""
|
||||
import time as _t
|
||||
token = "race-survivor-token"
|
||||
store.put_session(
|
||||
sessions.hash_token(token),
|
||||
sessions.AGENT,
|
||||
agent,
|
||||
agent,
|
||||
list(sessions.AGENT_SCOPES),
|
||||
int(_t.time()) + 900,
|
||||
fingerprint=fingerprint,
|
||||
)
|
||||
return token
|
||||
|
||||
def test_resolve_refuses_race_survivor_after_permanent_revoke(self):
|
||||
store, private = helpers.fresh()
|
||||
fp = store.get_agent("bigpickle")["fingerprint"]
|
||||
identity.revoke(store, "bigpickle", reason="compromise")
|
||||
# Session inserted *after* revoke (race); not marked revoked_at.
|
||||
token = self._plant_agent_session(store, "bigpickle", fp)
|
||||
with self.assertRaises(errors.SessionExpired):
|
||||
sessions.resolve(store, token)
|
||||
|
||||
def test_resolve_refuses_stale_fingerprint_after_rekey(self):
|
||||
store, private = helpers.fresh()
|
||||
old_fp = store.get_agent("bigpickle")["fingerprint"]
|
||||
new_pub = keys.generate()[1]
|
||||
secret = identity.seed_bootstrap(store, "bigpickle")
|
||||
identity.enrol_rekey(
|
||||
store, "bigpickle", keys.export_public_b64(new_pub), secret
|
||||
)
|
||||
token = self._plant_agent_session(store, "bigpickle", old_fp)
|
||||
with self.assertRaises(errors.SessionExpired):
|
||||
sessions.resolve(store, token)
|
||||
|
||||
def test_redeem_refuses_code_after_rekey_even_if_not_invalidated(self):
|
||||
"""Fingerprint binding: plant a code with a stale fingerprint."""
|
||||
store, private = helpers.fresh()
|
||||
agent_row = store.get_agent("bigpickle")
|
||||
old_fp = agent_row["fingerprint"]
|
||||
# Issue normally, then surgically restore used_at and keep old fp after rekey
|
||||
issued = helpers.issue(store, private)
|
||||
digest = viewcodes._hash_code(issued["code"])
|
||||
new_pub = keys.generate()[1]
|
||||
secret = identity.seed_bootstrap(store, "bigpickle")
|
||||
identity.enrol_rekey(
|
||||
store, "bigpickle", keys.export_public_b64(new_pub), secret
|
||||
)
|
||||
# Rekey invalidated the code; un-burn it to simulate put_code after invalidate.
|
||||
store.conn.execute(
|
||||
"UPDATE codes SET used_at = NULL, fingerprint = ? WHERE code_hash = ?",
|
||||
(old_fp, digest),
|
||||
)
|
||||
store.conn.commit()
|
||||
with self.assertRaises(errors.CodeInvalid):
|
||||
viewcodes.redeem_code(store, issued["code"])
|
||||
|
||||
def test_redeem_refuses_without_live_key(self):
|
||||
store, private = helpers.fresh()
|
||||
issued = helpers.issue(store, private)
|
||||
digest = viewcodes._hash_code(issued["code"])
|
||||
identity.revoke(store, "bigpickle", reason="compromise")
|
||||
identity.reinstate(store, "bigpickle") # ban lifted, key still wiped
|
||||
store.conn.execute(
|
||||
"UPDATE codes SET used_at = NULL WHERE code_hash = ?", (digest,)
|
||||
)
|
||||
store.conn.commit()
|
||||
with self.assertRaises(errors.CodeInvalid):
|
||||
viewcodes.redeem_code(store, issued["code"])
|
||||
|
||||
def test_verify_then_revoke_session_unusable(self):
|
||||
"""Session minted just before revoke must not resolve."""
|
||||
store, private = helpers.fresh()
|
||||
chal = identity.challenge(store, "bigpickle")
|
||||
sig = helpers.sign(private, chal["nonce"])
|
||||
sess = identity.verify(store, chal["nonce"], sig)
|
||||
identity.revoke(store, "bigpickle", reason="compromise")
|
||||
with self.assertRaises(errors.SessionExpired):
|
||||
sessions.resolve(store, sess["token"])
|
||||
|
||||
|
||||
class TestHumanSessionFingerprintEpoch(unittest.TestCase):
|
||||
"""N7: rekey must kill race-surviving human_via_agent sessions too."""
|
||||
|
||||
def test_resolve_refuses_stale_human_fingerprint_after_rekey(self):
|
||||
store, private = helpers.fresh()
|
||||
old_fp = store.get_agent("bigpickle")["fingerprint"]
|
||||
# Plant a human session as if redeem's put_session won a race against rekey.
|
||||
import time as _t
|
||||
token = "human-race-survivor"
|
||||
store.put_session(
|
||||
sessions.hash_token(token),
|
||||
sessions.HUMAN_VIA_AGENT,
|
||||
"bigpickle's human",
|
||||
"bigpickle",
|
||||
list(sessions.HUMAN_VIEW_SCOPES),
|
||||
int(_t.time()) + 3600,
|
||||
granted_by="bigpickle",
|
||||
fingerprint=old_fp,
|
||||
)
|
||||
new_pub = keys.generate()[1]
|
||||
secret = identity.seed_bootstrap(store, "bigpickle")
|
||||
identity.enrol_rekey(
|
||||
store, "bigpickle", keys.export_public_b64(new_pub), secret
|
||||
)
|
||||
# Clear revoked_at that rekey set, to isolate the fingerprint check.
|
||||
store.conn.execute(
|
||||
"UPDATE sessions SET revoked_at = NULL, revoked_reason = NULL"
|
||||
" WHERE token_hash = ?",
|
||||
(sessions.hash_token(token),),
|
||||
)
|
||||
store.conn.commit()
|
||||
with self.assertRaises(errors.SessionExpired) as ctx:
|
||||
sessions.resolve(store, token)
|
||||
self.assertIn("key epoch", str(ctx.exception))
|
||||
|
||||
def test_issue_code_return_has_single_view_ttl(self):
|
||||
"""N8: issue_code return dict must not duplicate view_ttl."""
|
||||
import inspect
|
||||
from sigil import viewcodes as vc
|
||||
src = inspect.getsource(vc.issue_code)
|
||||
# Count literal "view_ttl" keys in the return dict construction region.
|
||||
self.assertEqual(src.count('"view_ttl":'), 1, src[src.rfind("return"):])
|
||||
|
||||
@ -133,15 +133,25 @@ class TestSignedCheckpoints(unittest.TestCase):
|
||||
with self.assertRaises(ValueError):
|
||||
stranger.checkpoint(self.private, "not-enrolled")
|
||||
|
||||
def test_checkpoint_after_rekey_is_flagged(self):
|
||||
def test_checkpoint_refuses_enrolled_name_with_wrong_key(self):
|
||||
# R4: signer name is enrolled, but private key is not the enrolled one.
|
||||
before = len(self.store.checkpoints())
|
||||
other_private, _ = keys.generate()
|
||||
with self.assertRaises(ValueError) as ctx:
|
||||
self.store.checkpoint(other_private, "bigpickle")
|
||||
self.assertIn("does not match enrolled key", str(ctx.exception))
|
||||
self.assertEqual(len(self.store.checkpoints()), before)
|
||||
|
||||
def test_checkpoint_after_rekey_still_verifies_via_history(self):
|
||||
# N2: historical checkpoints verify against retired public keys.
|
||||
secret = identity.seed_bootstrap(self.store, "bigpickle")
|
||||
identity.enrol_rekey(
|
||||
self.store, "bigpickle",
|
||||
keys.export_public_b64(keys.generate()[1]), secret,
|
||||
)
|
||||
result = self.store.verify_checkpoints()
|
||||
self.assertFalse(result["ok"])
|
||||
self.assertIn("key changed", result["failures"][0]["reason"])
|
||||
self.assertTrue(result["ok"], result)
|
||||
self.assertEqual(result["failures"], [])
|
||||
|
||||
def test_empty_log_checkpoints_cleanly(self):
|
||||
store, private = helpers.fresh()
|
||||
@ -482,3 +492,75 @@ class TestRedeemRateLimitIsolation(unittest.TestCase):
|
||||
self.fail("successes must not share the fail bucket")
|
||||
except errors.SigilError:
|
||||
pass
|
||||
|
||||
|
||||
class TestKeyHistoryKeepsCheckpointsVerifiable(unittest.TestCase):
|
||||
"""N2: permanent revoke / rekey must not make honest checkpoints unverifiable."""
|
||||
|
||||
def test_checkpoints_verify_after_signer_permanent_revoke(self):
|
||||
import tempfile, os
|
||||
d = tempfile.mkdtemp()
|
||||
anchor = os.path.join(d, "tip")
|
||||
store = Store(":memory:", anchor_path=anchor)
|
||||
private, _ = helpers.enrol(store, "signer")
|
||||
store.record("identity.enrolled", actor="signer", subject="signer")
|
||||
store.checkpoint(private, "signer", "pre-compromise")
|
||||
before = store.verify_checkpoints()
|
||||
self.assertTrue(before["ok"], before)
|
||||
anchors_before = store.verify_external_anchors()
|
||||
self.assertTrue(anchors_before["ok"], anchors_before)
|
||||
identity.revoke(store, "signer", reason="compromise")
|
||||
after = store.verify_checkpoints()
|
||||
self.assertTrue(after["ok"], after)
|
||||
self.assertEqual(after["failures"], [])
|
||||
anchors_after = store.verify_external_anchors()
|
||||
self.assertTrue(anchors_after["ok"], anchors_after)
|
||||
# Live key wiped; history retained.
|
||||
self.assertFalse(identity.has_live_key(store.get_agent("signer")))
|
||||
hist = store.conn.execute(
|
||||
"SELECT fingerprint, retired_at FROM agent_keys WHERE agent = ?",
|
||||
("signer",),
|
||||
).fetchall()
|
||||
self.assertEqual(len(hist), 1)
|
||||
self.assertIsNotNone(hist[0]["retired_at"])
|
||||
|
||||
def test_private_key_never_in_agent_keys(self):
|
||||
store, private = helpers.fresh("carol")
|
||||
cols = [r["name"] for r in store.conn.execute("PRAGMA table_info(agent_keys)")]
|
||||
self.assertNotIn("private", cols)
|
||||
self.assertNotIn("privkey", cols)
|
||||
row = store.conn.execute("SELECT * FROM agent_keys WHERE agent = ?", ("carol",)).fetchone()
|
||||
self.assertEqual(len(row["pubkey"]), 32)
|
||||
|
||||
|
||||
class TestEnrolAfterPermanentRevoke(unittest.TestCase):
|
||||
"""N4: enrol must not burn the seed when the agent row still exists."""
|
||||
|
||||
def test_enrol_after_permanent_revoke_does_not_burn_seed(self):
|
||||
store, private = helpers.fresh()
|
||||
identity.revoke(store, "bigpickle", reason="compromise")
|
||||
identity.reinstate(store, "bigpickle")
|
||||
secret = identity.seed_bootstrap(store, "bigpickle")
|
||||
with self.assertRaises(errors.EnrolmentFailed) as ctx:
|
||||
identity.enrol(
|
||||
store, "bigpickle", keys.export_public_b64(keys.generate()[1]), secret
|
||||
)
|
||||
self.assertIn("enrol_rekey", str(ctx.exception))
|
||||
# Seed still usable for the correct path.
|
||||
identity.enrol_rekey(
|
||||
store, "bigpickle", keys.export_public_b64(keys.generate()[1]), secret
|
||||
)
|
||||
|
||||
|
||||
class TestSuspendedSignerCannotCheckpoint(unittest.TestCase):
|
||||
"""N5: timed-suspended signer must not checkpoint."""
|
||||
|
||||
def test_timed_suspend_refuses_checkpoint(self):
|
||||
import time
|
||||
store, private = helpers.fresh()
|
||||
identity.revoke(
|
||||
store, "bigpickle", until=int(time.time()) + 3600, reason="pause"
|
||||
)
|
||||
with self.assertRaises(ValueError) as ctx:
|
||||
store.checkpoint(private, "bigpickle")
|
||||
self.assertIn("revoked", str(ctx.exception).lower())
|
||||
|
||||
@ -124,3 +124,30 @@ class TestSealedStore(unittest.TestCase):
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
|
||||
|
||||
class TestPlaintextWorkingCopyHardening(unittest.TestCase):
|
||||
"""R5: plaintext tmp must be 0600; SQLCipher remains deferred."""
|
||||
|
||||
def test_plaintext_working_copy_is_owner_only(self):
|
||||
import tempfile
|
||||
from sigil.sealed import SealedStore
|
||||
from sigil import identity, keys
|
||||
|
||||
d = tempfile.mkdtemp()
|
||||
path = os.path.join(d, "sealed.db")
|
||||
store = SealedStore(path, "test-passphrase-ok")
|
||||
try:
|
||||
mode = os.stat(store._plain_path).st_mode & 0o777
|
||||
self.assertEqual(mode, 0o600, oct(mode))
|
||||
dir_mode = os.stat(store._tmpdir.name).st_mode & 0o777
|
||||
self.assertEqual(dir_mode, 0o700, oct(dir_mode))
|
||||
secret = identity.seed_bootstrap(store, "alice")
|
||||
_, pub = keys.generate()
|
||||
identity.enrol(store, "alice", keys.export_public_b64(pub), secret)
|
||||
store.flush()
|
||||
# export.db must not linger world-readable after flush
|
||||
export = os.path.join(store._tmpdir.name, "export.db")
|
||||
self.assertFalse(os.path.exists(export))
|
||||
finally:
|
||||
store.close()
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user