stackchain-dashboard/src/login_attempt_store.py
timmy 350bfc067d
All checks were successful
CI / lint (pull_request) Successful in 1m10s
CI / build-release (pull_request) Successful in 4s
CI / release-candidate (pull_request) Has been skipped
feat: surface failed token sign-in alerts (Closes #543)
2026-08-11 05:35:35 +00:00

295 lines
12 KiB
Python

"""Durable throttling state for operator sign-in attempts."""
import hashlib
import ipaddress
import math
import sqlite3
from pathlib import Path
from typing import Callable
class LoginAttemptStoreError(RuntimeError):
"""Raised when sign-in throttling state cannot be accessed safely."""
def client_source(peer_host: str, forwarded_for: str, trusted_proxy_cidrs: str) -> str:
"""Resolve a canonical client IP without trusting arbitrary forwarding headers."""
try:
peer = ipaddress.ip_address(peer_host)
trusted = [
ipaddress.ip_network(value.strip())
for value in trusted_proxy_cidrs.split(",")
if value.strip()
]
except ValueError:
return peer_host
if not any(peer in network for network in trusted):
return str(peer)
try:
forwarded = [
ipaddress.ip_address(value.strip())
for value in forwarded_for.split(",")
if value.strip()
]
except ValueError:
return str(peer)
for address in reversed(forwarded):
if not any(address in network for network in trusted):
return str(address)
return str(forwarded[0]) if forwarded else str(peer)
class LoginAttemptStore:
def __init__(
self,
path: str | Path,
*,
clock: Callable[[], float],
max_failures: int,
window_seconds: int,
max_entries: int = 10_000,
lock_timeout_seconds: float = 0.1,
alert_bucket_seconds: int = 60 * 60,
alert_retention_seconds: int = 30 * 24 * 60 * 60,
max_alert_buckets: int = 720,
) -> None:
self.path = Path(path)
self.clock = clock
self.max_failures = max(1, max_failures)
self.window_seconds = max(1, window_seconds)
self.max_entries = max(1, max_entries)
self.lock_timeout_seconds = lock_timeout_seconds
self.alert_bucket_seconds = max(1, alert_bucket_seconds)
self.alert_retention_seconds = max(1, alert_retention_seconds)
self.max_alert_buckets = max(1, max_alert_buckets)
@staticmethod
def _digest(source: str) -> str:
return hashlib.sha256(source.encode()).hexdigest()
def _connect(self) -> sqlite3.Connection:
try:
self.path.parent.mkdir(parents=True, exist_ok=True)
connection = sqlite3.connect(self.path, timeout=self.lock_timeout_seconds)
connection.execute(
"""
CREATE TABLE IF NOT EXISTS login_attempts (
source_hash TEXT PRIMARY KEY,
failures INTEGER NOT NULL,
window_started_at REAL NOT NULL
)
"""
)
connection.execute(
"""
CREATE TABLE IF NOT EXISTS login_admissions (
bucket TEXT NOT NULL,
source_hash TEXT NOT NULL,
attempts INTEGER NOT NULL,
window_started_at REAL NOT NULL,
PRIMARY KEY (bucket, source_hash)
)
"""
)
connection.execute(
"""
CREATE TABLE IF NOT EXISTS login_alerts (
method TEXT NOT NULL,
bucket_started_at INTEGER NOT NULL,
failed_count INTEGER NOT NULL DEFAULT 0,
blocked_count INTEGER NOT NULL DEFAULT 0,
first_at INTEGER NOT NULL,
last_at INTEGER NOT NULL,
PRIMARY KEY (method, bucket_started_at)
)
"""
)
return connection
except (OSError, sqlite3.Error) as exc:
raise LoginAttemptStoreError(
"Sign-in throttling is temporarily unavailable"
) from exc
def retry_after(self, source: str) -> int:
now = self.clock()
try:
with self._connect() as connection:
row = connection.execute(
"SELECT failures, window_started_at FROM login_attempts WHERE source_hash = ?",
(self._digest(source),),
).fetchone()
except (OSError, sqlite3.Error) as exc:
raise LoginAttemptStoreError(
"Sign-in throttling is temporarily unavailable"
) from exc
if row is None or row[0] < self.max_failures:
return 0
remaining = row[1] + self.window_seconds - now
return max(0, math.ceil(remaining))
def admit(self, bucket: str, source: str, *, limit: int) -> int:
"""Consume one fixed-window admission or return seconds until retry."""
now = self.clock()
source_hash = self._digest(source)
limit = max(1, limit)
try:
with self._connect() as connection:
connection.execute("BEGIN IMMEDIATE")
connection.execute(
"DELETE FROM login_admissions WHERE window_started_at + ? <= ?",
(self.window_seconds, now),
)
row = connection.execute(
"SELECT attempts, window_started_at FROM login_admissions "
"WHERE bucket = ? AND source_hash = ?",
(bucket, source_hash),
).fetchone()
if row is not None and row[0] >= limit:
return max(1, math.ceil(row[1] + self.window_seconds - now))
if row is None:
connection.execute(
"INSERT INTO login_admissions VALUES (?, ?, 1, ?)",
(bucket, source_hash, now),
)
else:
connection.execute(
"UPDATE login_admissions SET attempts = attempts + 1 "
"WHERE bucket = ? AND source_hash = ?",
(bucket, source_hash),
)
connection.execute(
"""
DELETE FROM login_admissions
WHERE (bucket, source_hash) NOT IN (
SELECT bucket, source_hash FROM login_admissions
ORDER BY window_started_at DESC, rowid DESC LIMIT ?
)
""",
(self.max_entries,),
)
return 0
except (OSError, sqlite3.Error) as exc:
raise LoginAttemptStoreError(
"Sign-in throttling is temporarily unavailable"
) from exc
def _record_alert(
self, connection: sqlite3.Connection, method: str, *, blocked: bool, now: float
) -> None:
safe_method = method if method in {"token", "passkey"} else "unknown"
occurred_at = int(now)
bucket = occurred_at - (occurred_at % self.alert_bucket_seconds)
failed = 0 if blocked else 1
blocked_count = 1 if blocked else 0
connection.execute(
"""
INSERT INTO login_alerts(
method, bucket_started_at, failed_count, blocked_count, first_at, last_at
) VALUES (?, ?, ?, ?, ?, ?)
ON CONFLICT(method, bucket_started_at) DO UPDATE SET
failed_count = failed_count + excluded.failed_count,
blocked_count = blocked_count + excluded.blocked_count,
first_at = MIN(first_at, excluded.first_at),
last_at = MAX(last_at, excluded.last_at)
""",
(safe_method, bucket, failed, blocked_count, occurred_at, occurred_at),
)
connection.execute(
"DELETE FROM login_alerts WHERE (method, bucket_started_at) NOT IN "
"(SELECT method, bucket_started_at FROM login_alerts "
"ORDER BY bucket_started_at DESC LIMIT ?)",
(self.max_alert_buckets,),
)
def record_failure(self, source: str, *, method: str = "token") -> None:
now = self.clock()
source_hash = self._digest(source)
try:
with self._connect() as connection:
connection.execute("BEGIN IMMEDIATE")
connection.execute(
"DELETE FROM login_attempts WHERE window_started_at + ? <= ?",
(self.window_seconds, now),
)
row = connection.execute(
"SELECT failures, window_started_at FROM login_attempts WHERE source_hash = ?",
(source_hash,),
).fetchone()
if row is None or row[1] + self.window_seconds <= now:
connection.execute(
"INSERT OR REPLACE INTO login_attempts VALUES (?, 1, ?)",
(source_hash, now),
)
else:
connection.execute(
"UPDATE login_attempts SET failures = failures + 1 WHERE source_hash = ?",
(source_hash,),
)
connection.execute(
"""
DELETE FROM login_attempts
WHERE source_hash NOT IN (
SELECT source_hash FROM login_attempts
ORDER BY window_started_at DESC, rowid DESC LIMIT ?
)
""",
(self.max_entries,),
)
self._record_alert(connection, method, blocked=False, now=now)
except (OSError, sqlite3.Error) as exc:
raise LoginAttemptStoreError(
"Sign-in throttling is temporarily unavailable"
) from exc
def record_blocked(self, *, method: str = "token") -> None:
now = self.clock()
try:
with self._connect() as connection:
connection.execute("BEGIN IMMEDIATE")
self._record_alert(connection, method, blocked=True, now=now)
except (OSError, sqlite3.Error) as exc:
raise LoginAttemptStoreError(
"Sign-in throttling is temporarily unavailable"
) from exc
def list_alerts(self, *, limit: int = 24) -> list[dict[str, int | str]]:
bounded_limit = min(100, max(1, limit))
now = int(self.clock())
try:
with self._connect() as connection:
connection.execute(
"DELETE FROM login_alerts WHERE last_at < ?",
(now - self.alert_retention_seconds,),
)
rows = connection.execute(
"SELECT method, failed_count, blocked_count, first_at, last_at "
"FROM login_alerts ORDER BY bucket_started_at DESC LIMIT ?",
(bounded_limit,),
).fetchall()
except (OSError, sqlite3.Error) as exc:
raise LoginAttemptStoreError(
"Sign-in throttling is temporarily unavailable"
) from exc
return [
{
"method": row[0],
"failed_count": row[1],
"blocked_count": row[2],
"first_at": row[3],
"last_at": row[4],
}
for row in rows
]
def clear(self, source: str) -> None:
try:
with self._connect() as connection:
connection.execute(
"DELETE FROM login_attempts WHERE source_hash = ?",
(self._digest(source),),
)
except (OSError, sqlite3.Error) as exc:
raise LoginAttemptStoreError(
"Sign-in throttling is temporarily unavailable"
) from exc