feat(auth): throttle password login attempts

This commit is contained in:
2026-07-21 12:23:12 +02:00
parent e91935c03a
commit 8c74e360d2
6 changed files with 666 additions and 3 deletions

View File

@@ -1,6 +1,7 @@
from __future__ import annotations
from dataclasses import dataclass
from functools import lru_cache
from fastapi import APIRouter, Depends, HTTPException, Request, Response, status
from sqlalchemy.orm import Session
@@ -51,6 +52,11 @@ from govoplan_access.backend.semantic import collect_function_assignment_ids, co
from govoplan_access.backend.auth.tenant_context import AccessTenantContextSwitcher
from govoplan_access.backend.security.api_keys import authenticate_api_key
from govoplan_access.backend.security.passwords import DUMMY_PASSWORD_HASH, verify_password
from govoplan_access.backend.security.login_throttle import (
LoginThrottle,
LoginThrottleDecision,
build_login_throttle,
)
from govoplan_access.backend.security.sessions import (
authenticate_session_token,
collect_user_authorization_context,
@@ -277,6 +283,67 @@ def _resolve_login_user(session: Session, payload: LoginRequest) -> tuple[Accoun
return account, row[0], row[1]
@lru_cache(maxsize=1)
def _configured_login_throttle() -> LoginThrottle | None:
if not settings.auth_login_throttle_enabled:
return None
return build_login_throttle(
redis_url=settings.redis_url,
identity_limit=settings.auth_login_throttle_identity_limit,
client_limit=settings.auth_login_throttle_client_limit,
window_seconds=settings.auth_login_throttle_window_seconds,
redis_retry_seconds=settings.auth_login_throttle_redis_retry_seconds,
)
def _raise_login_throttled(decision: LoginThrottleDecision) -> None:
retry_after = max(1, decision.retry_after_seconds)
# The detail deliberately matches every credential failure. Status and
# Retry-After communicate endpoint throttling without disclosing whether
# the supplied identity exists or has an active membership.
raise HTTPException(
status_code=status.HTTP_429_TOO_MANY_REQUESTS,
detail="Invalid login",
headers={"Retry-After": str(retry_after)},
)
def _resolve_throttled_login_user(
session: Session,
payload: LoginRequest,
request: Request,
) -> tuple[Account, User, Tenant]:
throttle = _configured_login_throttle()
if throttle is None:
return _resolve_login_user(session, payload)
normalized_email = normalize_email(payload.email)
client_address = request.client.host if request.client else None
context = {
"normalized_email": normalized_email,
"tenant_slug": payload.tenant_slug,
"client_address": client_address,
}
decision = throttle.check(**context)
if not decision.allowed:
_raise_login_throttled(decision)
try:
resolved = _resolve_login_user(session, payload)
except HTTPException as exc:
if exc.status_code == status.HTTP_401_UNAUTHORIZED:
decision = throttle.record_failure(**context)
if not decision.allowed:
_raise_login_throttled(decision)
raise
throttle.record_success(
normalized_email=normalized_email,
tenant_slug=payload.tenant_slug,
)
return resolved
def _extract_auth_token(request: Request) -> tuple[str | None, str]:
x_api_key = request.headers.get("x-api-key")
if x_api_key:
@@ -620,7 +687,7 @@ def _me_response(
@router.post("/login", response_model=LoginResponse)
def login(payload: LoginRequest, request: Request, response: Response, session: Session = Depends(get_session)):
account, user, tenant = _resolve_login_user(session, payload)
account, user, tenant = _resolve_throttled_login_user(session, payload, request)
identity_directory = _identity_directory_from_request(request)
authorization_context = collect_user_authorization_context(
session,

View File

@@ -0,0 +1,358 @@
from __future__ import annotations
import hashlib
import logging
import threading
import time
from dataclasses import dataclass
from typing import Protocol
from redis import Redis
from redis.exceptions import RedisError
logger = logging.getLogger(__name__)
_REDIS_INCREMENT_SCRIPT = """
local count = redis.call('INCR', KEYS[1])
if count == 1 then
redis.call('EXPIRE', KEYS[1], ARGV[1])
end
local ttl = redis.call('TTL', KEYS[1])
return {count, ttl}
"""
@dataclass(frozen=True, slots=True)
class AttemptBucket:
count: int = 0
retry_after_seconds: int = 0
@dataclass(frozen=True, slots=True)
class LoginThrottleDecision:
allowed: bool
retry_after_seconds: int = 0
class LoginAttemptStore(Protocol):
def read(self, key: str) -> AttemptBucket: ...
def increment(self, key: str, *, window_seconds: int) -> AttemptBucket: ...
def delete(self, key: str) -> None: ...
class InMemoryLoginAttemptStore:
"""Bounded process-local fallback for development and Redis outages."""
def __init__(self, *, max_entries: int = 10_000) -> None:
self._entries: dict[str, tuple[int, float]] = {}
self._lock = threading.Lock()
self._max_entries = max(2, max_entries)
def read(self, key: str) -> AttemptBucket:
now = time.monotonic()
with self._lock:
entry = self._active_entry(key, now=now)
if entry is None:
return AttemptBucket()
count, expires_at = entry
return AttemptBucket(
count=count,
retry_after_seconds=max(1, int(expires_at - now + 0.999)),
)
def increment(self, key: str, *, window_seconds: int) -> AttemptBucket:
now = time.monotonic()
with self._lock:
entry = self._active_entry(key, now=now)
if entry is None:
self._make_room(now=now, incoming_key=key)
count = 1
expires_at = now + window_seconds
else:
count = entry[0] + 1
expires_at = entry[1]
self._entries[key] = (count, expires_at)
return AttemptBucket(
count=count,
retry_after_seconds=max(1, int(expires_at - now + 0.999)),
)
def delete(self, key: str) -> None:
with self._lock:
self._entries.pop(key, None)
def _active_entry(self, key: str, *, now: float) -> tuple[int, float] | None:
entry = self._entries.get(key)
if entry is None:
return None
if entry[1] <= now:
self._entries.pop(key, None)
return None
return entry
def _make_room(self, *, now: float, incoming_key: str) -> None:
if incoming_key in self._entries or len(self._entries) < self._max_entries:
return
expired = [key for key, (_, expires_at) in self._entries.items() if expires_at <= now]
for key in expired:
self._entries.pop(key, None)
while len(self._entries) >= self._max_entries:
self._entries.pop(next(iter(self._entries)))
class RedisLoginAttemptStore:
"""Redis-backed fixed-window counters shared by all API workers."""
def __init__(self, redis_url: str) -> None:
self._client = Redis.from_url(
redis_url,
decode_responses=True,
socket_connect_timeout=0.25,
socket_timeout=0.25,
health_check_interval=30,
)
def read(self, key: str) -> AttemptBucket:
pipeline = self._client.pipeline(transaction=False)
pipeline.get(key)
pipeline.ttl(key)
raw_count, raw_ttl = pipeline.execute()
count = int(raw_count or 0)
ttl = int(raw_ttl or 0)
return AttemptBucket(count=count, retry_after_seconds=max(0, ttl))
def increment(self, key: str, *, window_seconds: int) -> AttemptBucket:
result = self._client.eval(_REDIS_INCREMENT_SCRIPT, 1, key, window_seconds)
if not isinstance(result, (list, tuple)) or len(result) != 2:
raise RedisError("Unexpected login throttle response from Redis")
return AttemptBucket(
count=int(result[0]),
retry_after_seconds=max(1, int(result[1])),
)
def delete(self, key: str) -> None:
self._client.delete(key)
class ResilientLoginAttemptStore:
"""Prefer the distributed store and fail safely to a local bounded store."""
def __init__(
self,
primary: LoginAttemptStore | None,
fallback: LoginAttemptStore,
*,
retry_seconds: int = 30,
) -> None:
self._primary = primary
self._fallback = fallback
self._retry_seconds = max(1, retry_seconds)
self._primary_unavailable_until = 0.0
self._state_lock = threading.Lock()
def read(self, key: str) -> AttemptBucket:
fallback_result = self._fallback.read(key)
primary = self._available_primary()
if primary is None:
return fallback_result
try:
return _stricter_bucket(primary.read(key), fallback_result)
except (RedisError, OSError, TimeoutError, ConnectionError) as exc:
self._mark_primary_unavailable(exc)
return fallback_result
def increment(self, key: str, *, window_seconds: int) -> AttemptBucket:
primary = self._available_primary()
if primary is not None:
fallback_result = self._fallback.increment(key, window_seconds=window_seconds)
try:
# Mirror the active process's failures so a later Redis outage
# or recovery cannot restart its protection window from zero.
primary_result = primary.increment(key, window_seconds=window_seconds)
return _stricter_bucket(primary_result, fallback_result)
except (RedisError, OSError, TimeoutError, ConnectionError) as exc:
self._mark_primary_unavailable(exc)
return fallback_result
return self._fallback.increment(key, window_seconds=window_seconds)
def delete(self, key: str) -> None:
self._fallback.delete(key)
primary = self._available_primary()
if primary is None:
return
try:
primary.delete(key)
except (RedisError, OSError, TimeoutError, ConnectionError) as exc:
self._mark_primary_unavailable(exc)
def _available_primary(self) -> LoginAttemptStore | None:
if self._primary is None:
return None
with self._state_lock:
if time.monotonic() < self._primary_unavailable_until:
return None
return self._primary
def _mark_primary_unavailable(self, exc: Exception) -> None:
should_log = False
with self._state_lock:
now = time.monotonic()
if now >= self._primary_unavailable_until:
should_log = True
self._primary_unavailable_until = now + self._retry_seconds
if should_log:
logger.warning(
"Redis login throttling is unavailable; using the process-local fallback (%s)",
type(exc).__name__,
)
def _stricter_bucket(first: AttemptBucket, second: AttemptBucket) -> AttemptBucket:
return AttemptBucket(
count=max(first.count, second.count),
retry_after_seconds=max(first.retry_after_seconds, second.retry_after_seconds),
)
class LoginThrottle:
def __init__(
self,
store: LoginAttemptStore,
*,
identity_limit: int,
client_limit: int,
window_seconds: int,
key_prefix: str = "govoplan:access:login:v1",
) -> None:
self._store = store
self._identity_limit = max(1, identity_limit)
self._client_limit = max(1, client_limit)
self._window_seconds = max(1, window_seconds)
self._key_prefix = key_prefix.rstrip(":")
def check(
self,
*,
normalized_email: str,
tenant_slug: str | None,
client_address: str | None,
) -> LoginThrottleDecision:
return self._decision(
self._buckets(
normalized_email=normalized_email,
tenant_slug=tenant_slug,
client_address=client_address,
),
increment=False,
)
def record_failure(
self,
*,
normalized_email: str,
tenant_slug: str | None,
client_address: str | None,
) -> LoginThrottleDecision:
return self._decision(
self._buckets(
normalized_email=normalized_email,
tenant_slug=tenant_slug,
client_address=client_address,
),
increment=True,
)
def record_success(
self,
*,
normalized_email: str,
tenant_slug: str | None,
) -> None:
del tenant_slug
identity_key, _ = self._keys(
normalized_email=normalized_email,
client_address=None,
)
self._store.delete(identity_key)
def _decision(
self,
buckets: tuple[tuple[str, int], ...],
*,
increment: bool,
) -> LoginThrottleDecision:
blocked_retry_after = 0
for key, limit in buckets:
state = (
self._store.increment(key, window_seconds=self._window_seconds)
if increment
else self._store.read(key)
)
if state.count >= limit:
blocked_retry_after = max(
blocked_retry_after,
max(1, state.retry_after_seconds),
)
return LoginThrottleDecision(
allowed=blocked_retry_after == 0,
retry_after_seconds=blocked_retry_after,
)
def _buckets(
self,
*,
normalized_email: str,
tenant_slug: str | None,
client_address: str | None,
) -> tuple[tuple[str, int], ...]:
# Accounts and their passwords are global login identities. Do not put
# the caller-supplied tenant slug into the identity key: rotating fake
# slugs must not bypass the account-level limit.
del tenant_slug
identity_key, client_key = self._keys(
normalized_email=normalized_email,
client_address=client_address,
)
buckets = [(identity_key, self._identity_limit)]
if client_key is not None:
buckets.append((client_key, self._client_limit))
return tuple(buckets)
def _keys(
self,
*,
normalized_email: str,
client_address: str | None,
) -> tuple[str, str | None]:
identity = normalized_email.strip().casefold()
identity_digest = hashlib.sha256(identity.encode()).hexdigest()
identity_key = f"{self._key_prefix}:identity:{identity_digest}"
if not client_address:
return identity_key, None
client_digest = hashlib.sha256(client_address.strip().casefold().encode()).hexdigest()
return identity_key, f"{self._key_prefix}:client:{client_digest}"
def build_login_throttle(
*,
redis_url: str | None,
identity_limit: int,
client_limit: int,
window_seconds: int,
redis_retry_seconds: int,
) -> LoginThrottle:
redis_store = RedisLoginAttemptStore(redis_url) if redis_url and redis_url.strip() else None
resilient_store = ResilientLoginAttemptStore(
redis_store,
InMemoryLoginAttemptStore(),
retry_seconds=redis_retry_seconds,
)
return LoginThrottle(
resilient_store,
identity_limit=identity_limit,
client_limit=client_limit,
window_seconds=window_seconds,
)