feat: throttle repeated operator sign-in failures (#270)

src/login_attempt_store.py

@@ -0,0 +1,149 @@
+"""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,
+    ) -> 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
+
+    @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
+                )
+                """
+            )
+            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 record_failure(self, source: str) -> 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,),
+                )
+        except (OSError, sqlite3.Error) as exc:
+            raise LoginAttemptStoreError(
+                "Sign-in throttling is temporarily unavailable"
+            ) from exc
+
+    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

src/main.py

@@ -28,6 +28,7 @@ from src.gitea_proxy import (
     repos,
 )
 from src.idempotency import IdempotencyLedger, IdempotencyLedgerBusy
+from src.login_attempt_store import LoginAttemptStore, LoginAttemptStoreError, client_source
 from src.models import Issue, Milestone, PullRequest, Repo, User
 from src.suggestion_engine import compute
 from src.views import router as frontend_router
@@ -129,6 +130,20 @@ class DashboardSignIn(BaseModel):
     access_token: str = Field(min_length=1, max_length=1_024)
 
 
+def _login_attempt_store() -> LoginAttemptStore:
+    state_dir = os.getenv("STACKCHAIN_STATE_DIR", ".stackchain-state")
+    return LoginAttemptStore(
+        os.getenv(
+            "STACKCHAIN_LOGIN_ATTEMPT_DB",
+            os.path.join(state_dir, "login-attempts.sqlite3"),
+        ),
+        clock=time.time,
+        max_failures=int(os.getenv("STACKCHAIN_LOGIN_MAX_FAILURES", "5")),
+        window_seconds=int(os.getenv("STACKCHAIN_LOGIN_WINDOW_SECONDS", "300")),
+        max_entries=int(os.getenv("STACKCHAIN_LOGIN_MAX_ENTRIES", "10000")),
+    )
+
+
 class NotificationReadBatch(BaseModel):
     ids: list[PositiveInt] = Field(min_length=1, max_length=50)
 
@@ -521,11 +536,51 @@ def health() -> dict[str, str]:
 
 @app.post("/api/v1/session")
 async def sign_in(payload: DashboardSignIn, request: Request, response: Response):
+    peer_host = request.client.host if request.client is not None else "unknown"
+    source = client_source(
+        peer_host,
+        request.headers.get("x-forwarded-for", ""),
+        os.getenv("STACKCHAIN_TRUSTED_PROXY_CIDRS", ""),
+    )
+    attempts = _login_attempt_store()
+    try:
+        retry_after = attempts.retry_after(source)
+    except LoginAttemptStoreError:
+        return JSONResponse(
+            {"detail": "Sign-in throttling is temporarily unavailable"},
+            status_code=503,
+            headers={"Cache-Control": "no-store"},
+        )
+    if retry_after:
+        return JSONResponse(
+            {"detail": "Too many sign-in attempts"},
+            status_code=429,
+            headers={
+                "Cache-Control": "no-store",
+                "Retry-After": str(retry_after),
+            },
+        )
     configured_token = dashboard_auth.access_token()
     if not configured_token or not hmac.compare_digest(
         payload.access_token, configured_token
     ):
+        try:
+            attempts.record_failure(source)
+        except LoginAttemptStoreError:
+            return JSONResponse(
+                {"detail": "Sign-in throttling is temporarily unavailable"},
+                status_code=503,
+                headers={"Cache-Control": "no-store"},
+            )
         raise HTTPException(status_code=401, detail="Invalid access token")
+    try:
+        attempts.clear(source)
+    except LoginAttemptStoreError:
+        return JSONResponse(
+            {"detail": "Sign-in throttling is temporarily unavailable"},
+            status_code=503,
+            headers={"Cache-Control": "no-store"},
+        )
     try:
         signed, session = dashboard_auth.issue_session()
     except dashboard_auth.SessionStoreError:

src/views.py

@@ -13,8 +13,8 @@ LOGIN_HTML = """<!doctype html>
 <title>Sign in · Stackchain Dashboard</title>
 <style>body{margin:0;background:#07111f;color:#eef6ff;font:16px system-ui;display:grid;min-height:100vh;place-items:center}main{width:min(90vw,24rem);padding:2rem;border:1px solid #29415d;border-radius:1rem;background:#0d1b2b}label,input,button{display:block;width:100%;box-sizing:border-box}input,button{min-height:48px;margin-top:.6rem;border-radius:.6rem;border:1px solid #49647f;padding:.75rem}button{margin-top:1rem;background:#55d6be;color:#06121b;font-weight:700}p{color:#a9bed3}</style></head>
 <body><main><h1>Operator sign in</h1><p>Enter the dashboard access token. It is exchanged for a private, short-lived session and is never stored on this device.</p>
-<form id="sign-in"><label>Access token<input name="access_token" type="password" autocomplete="current-password" required></label><button>Sign in</button><p id="status" role="status"></p></form></main>
-<script>document.getElementById('sign-in').addEventListener('submit',async(event)=>{event.preventDefault();const status=document.getElementById('status');status.textContent='Signing in…';const access_token=new FormData(event.currentTarget).get('access_token');const response=await fetch('api/v1/session',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({access_token})});if(response.ok){location.replace('./');return}status.textContent='Sign-in failed. Check the token and try again.';event.currentTarget.reset()});</script></body></html>"""
+<form id="sign-in"><label>Access token<input name="access_token" type="password" autocomplete="current-password" required></label><button id="submit-sign-in">Sign in</button><p id="status" role="status"></p></form></main>
+<script src="static/login.js"></script></body></html>"""
 
 
 class RevalidatingHTMLResponse(HTMLResponse):