Files
PiCloud/crates/manager-core/src/app_secrets_repo.rs
MechaCat02 513c4a2d3c fix(audit-2026-06-11/H-D1): bind AES-GCM AAD on secrets + realtime signing key
At-rest secrets were AES-256-GCM sealed with no Associated
Authentication Data, so anyone with Postgres write access could
ciphertext-swap rows across apps (or rename via row edit) and the
decrypt would silently succeed under the wrong identity, returning
attacker-chosen plaintext. This breaks the cross-app isolation boundary
the moment DB write access is achieved.

Adds an AAD-bound envelope (v1) alongside the legacy no-AAD layout (v0),
discriminated by a per-row `version` column:

* shared::crypto — new encrypt_with_aad / decrypt_with_aad using
  aes_gcm::aead::Payload { msg, aad }. Originals retained for v0 reads.
  Tests: AAD round-trip, AAD-mismatch fails, empty-AAD round-trip.

* migration 0042 — adds `secrets.version SMALLINT NOT NULL DEFAULT 0`
  and `app_secrets.realtime_signing_key_version SMALLINT NOT NULL
  DEFAULT 0`. Existing rows stay v0; new writes are v1.

* secrets (SDK + admin API) — seal() now binds AAD =
  "secret:{app_id}:{name}" and emits v1; open() dispatches on version.
  StoredSecret gains a `version` field; SecretsRepo::set takes it.
  Both secrets_service::set and secrets_api::set_secret go through the
  v1 path. Tests prove a cross-app swap and a cross-name swap both
  surface Corrupted, and that a hand-built v0 row still decrypts.

* app_secrets (realtime signing key) — get_or_create_signing_key writes
  v1 with AAD = "app_secret:{app_id}:realtime_signing_key"; decode
  dispatches on version. Tests cover v0 decode, v1 round-trip, and v1
  decode-under-wrong-app failing.

* email-trigger inbound secret — kept on v0 (seal_legacy/open_legacy)
  and explicitly deferred: email_trigger_details has no version column
  and the trigger_id isn't known at seal time. The audit classes the
  email-trigger AAD gap as Medium; folded into v1.2's key-versioning
  pass per SECURITY_AUDIT.md.

* expected_schema.txt re-blessed by hand (no local Postgres) for the two
  new columns + migration 0042.

Also folds in a let-else clippy fix in auth_api.rs (login Argon2
semaphore acquire, from the H-B1 commit) and two cargo-fmt reflows.

No re-encryption sweep — v0 rows decrypt as-is; the sweep is deferred to
v1.2's key-versioning pass (audit "Notes on remediation methodology").

Audit ref: security_audit/03_crypto_secrets.md (H-D1).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-12 17:15:09 +02:00

262 lines
8.7 KiB
Rust

//! `AppSecretsRepo` — per-app secret material (v1.1.6, encrypted v1.1.7,
//! plaintext column dropped v1.1.8 F1).
//!
//! Holds the HMAC signing key for realtime subscriber tokens. The key is
//! generated lazily (32 random bytes) on the first
//! `pubsub::subscriber_token` call for an app and never changes
//! thereafter (no rotation API yet). The key is never exposed to
//! scripts: the SDK mints tokens, it never returns the key.
//!
//! **v1.1.7 at-rest encryption.** The key is sealed with the process
//! master key (AES-256-GCM). v1.1.8 (this commit) removes the
//! plaintext fallback column — the read path consults only the
//! encrypted columns now. The 0032 migration enforces this by
//! refusing to drop the plaintext column unless every existing row
//! has been encrypted by the v1.1.7 startup sweep first.
use async_trait::async_trait;
use picloud_shared::{crypto, AppId, MasterKey};
use rand::RngCore;
use sqlx::PgPool;
/// Length of a freshly-generated realtime signing key.
pub const SIGNING_KEY_LEN: usize = 32;
#[derive(Debug, thiserror::Error)]
pub enum AppSecretsRepoError {
#[error("database error: {0}")]
Db(#[from] sqlx::Error),
/// A stored encrypted signing key could not be decrypted — corrupted
/// row or a master-key mismatch (e.g. `PICLOUD_SECRET_KEY` changed).
#[error("realtime signing key could not be decrypted (corrupted row or master-key mismatch)")]
Crypto,
}
#[async_trait]
pub trait AppSecretsRepo: Send + Sync {
/// Fetch the app's realtime signing key, generating + persisting one
/// (32 random bytes, encrypted) if absent. Idempotent under
/// concurrency: a racing creator's `ON CONFLICT DO NOTHING` insert is
/// a no-op and the existing key is returned.
async fn get_or_create_signing_key(
&self,
app_id: AppId,
) -> Result<Vec<u8>, AppSecretsRepoError>;
/// Fetch the signing key if it exists, WITHOUT creating one. The SSE
/// verify path uses this: a missing key means no token was ever
/// minted for the app, so any presented token must be rejected.
async fn signing_key(&self, app_id: AppId) -> Result<Option<Vec<u8>>, AppSecretsRepoError>;
}
pub struct PostgresAppSecretsRepo {
pool: PgPool,
master_key: MasterKey,
}
impl PostgresAppSecretsRepo {
#[must_use]
pub fn new(pool: PgPool, master_key: MasterKey) -> Self {
Self { pool, master_key }
}
fn decode(
&self,
app_id: AppId,
encrypted: Option<Vec<u8>>,
nonce: Option<Vec<u8>>,
version: i16,
) -> Result<Option<Vec<u8>>, AppSecretsRepoError> {
decode_signing_key(&self.master_key, app_id, encrypted, nonce, version)
}
}
/// Audit 2026-06-11 H-D1 — AAD binding the realtime signing key to its
/// owning app, so a Postgres-write attacker can't swap one app's
/// ciphertext under another app's row.
fn signing_key_aad(app_id: AppId) -> Vec<u8> {
format!("app_secret:{app_id}:realtime_signing_key").into_bytes()
}
/// Current AAD-bound envelope version for the realtime signing key.
const SIGNING_KEY_ENVELOPE_V1: i16 = 1;
/// Resolve the signing key from a row's encrypted columns. The
/// plaintext fallback that existed in v1.1.7 is gone — v1.1.8 F1
/// drops the column outright (migration 0032 refuses to apply if any
/// row still needs encryption, guaranteeing the read path can't see
/// a row that only has a plaintext value).
///
/// Audit 2026-06-11 H-D1: dispatches on `version`. v0 = legacy no-AAD
/// (pre-audit rows); v1 = AAD bound to `app_secret:{app_id}:realtime_signing_key`.
fn decode_signing_key(
master_key: &MasterKey,
app_id: AppId,
encrypted: Option<Vec<u8>>,
nonce: Option<Vec<u8>>,
version: i16,
) -> Result<Option<Vec<u8>>, AppSecretsRepoError> {
match (encrypted, nonce) {
(Some(ct), Some(n)) => {
let key = match version {
0 => crypto::decrypt(&ct, &n, master_key.as_bytes()),
SIGNING_KEY_ENVELOPE_V1 => {
let aad = signing_key_aad(app_id);
crypto::decrypt_with_aad(&ct, &n, &aad, master_key.as_bytes())
}
_ => return Err(AppSecretsRepoError::Crypto),
}
.map_err(|_| AppSecretsRepoError::Crypto)?;
Ok(Some(key))
}
_ => Ok(None),
}
}
#[async_trait]
impl AppSecretsRepo for PostgresAppSecretsRepo {
async fn get_or_create_signing_key(
&self,
app_id: AppId,
) -> Result<Vec<u8>, AppSecretsRepoError> {
let mut fresh = vec![0u8; SIGNING_KEY_LEN];
rand::thread_rng().fill_bytes(&mut fresh);
// Audit 2026-06-11 H-D1 — new keys are AAD-bound (v1).
let aad = signing_key_aad(app_id);
let enc = crypto::encrypt_with_aad(&fresh, &aad, self.master_key.as_bytes());
// Insert-if-absent (encrypted-only). The racing-creator's insert
// is a no-op; the SELECT always returns the winning row.
sqlx::query(
"INSERT INTO app_secrets \
(app_id, realtime_signing_key_encrypted, realtime_signing_key_nonce, \
realtime_signing_key_version) \
VALUES ($1, $2, $3, $4) ON CONFLICT (app_id) DO NOTHING",
)
.bind(app_id.into_inner())
.bind(&enc.ciphertext)
.bind(&enc.nonce[..])
.bind(SIGNING_KEY_ENVELOPE_V1)
.execute(&self.pool)
.await?;
let row: (Option<Vec<u8>>, Option<Vec<u8>>, i16) = sqlx::query_as(
"SELECT realtime_signing_key_encrypted, realtime_signing_key_nonce, \
realtime_signing_key_version \
FROM app_secrets WHERE app_id = $1",
)
.bind(app_id.into_inner())
.fetch_one(&self.pool)
.await?;
// A row exists by construction, so a key must decode.
self.decode(app_id, row.0, row.1, row.2)?
.ok_or(AppSecretsRepoError::Crypto)
}
async fn signing_key(&self, app_id: AppId) -> Result<Option<Vec<u8>>, AppSecretsRepoError> {
let row: Option<(Option<Vec<u8>>, Option<Vec<u8>>, i16)> = sqlx::query_as(
"SELECT realtime_signing_key_encrypted, realtime_signing_key_nonce, \
realtime_signing_key_version \
FROM app_secrets WHERE app_id = $1",
)
.bind(app_id.into_inner())
.fetch_optional(&self.pool)
.await?;
match row {
Some((e, n, v)) => self.decode(app_id, e, n, v),
None => Ok(None),
}
}
}
#[cfg(test)]
mod tests {
use super::*;
fn key() -> MasterKey {
MasterKey::from_bytes([9u8; 32])
}
fn app() -> AppId {
AppId::new()
}
#[test]
fn legacy_v0_decodes() {
let mk = key();
let secret = vec![1u8, 2, 3, 4];
let enc = crypto::encrypt(&secret, mk.as_bytes());
let got = decode_signing_key(
&mk,
app(),
Some(enc.ciphertext),
Some(enc.nonce.to_vec()),
0,
)
.unwrap();
assert_eq!(got, Some(secret));
}
#[test]
fn v1_aad_round_trips() {
let mk = key();
let a = app();
let secret = vec![7u8; 32];
let aad = signing_key_aad(a);
let enc = crypto::encrypt_with_aad(&secret, &aad, mk.as_bytes());
let got = decode_signing_key(
&mk,
a,
Some(enc.ciphertext),
Some(enc.nonce.to_vec()),
SIGNING_KEY_ENVELOPE_V1,
)
.unwrap();
assert_eq!(got, Some(secret));
}
#[test]
fn v1_decode_under_wrong_app_fails() {
// Audit 2026-06-11 H-D1 — a row swapped to a different app_id
// can't be decrypted: the AAD no longer matches.
let mk = key();
let a = app();
let b = app();
let secret = vec![7u8; 32];
let aad = signing_key_aad(a);
let enc = crypto::encrypt_with_aad(&secret, &aad, mk.as_bytes());
let err = decode_signing_key(
&mk,
b,
Some(enc.ciphertext),
Some(enc.nonce.to_vec()),
SIGNING_KEY_ENVELOPE_V1,
)
.unwrap_err();
assert!(matches!(err, AppSecretsRepoError::Crypto));
}
#[test]
fn missing_columns_is_none() {
let got = decode_signing_key(&key(), app(), None, None, 0).unwrap();
assert_eq!(got, None);
}
#[test]
fn wrong_master_key_is_crypto_error() {
let secret = vec![3u8; 32];
let enc = crypto::encrypt(&secret, key().as_bytes());
let other = MasterKey::from_bytes([1u8; 32]);
let err = decode_signing_key(
&other,
app(),
Some(enc.ciphertext),
Some(enc.nonce.to_vec()),
0,
)
.unwrap_err();
assert!(matches!(err, AppSecretsRepoError::Crypto));
}
}