feat(web): harden login gate — throttle, scoped TLS, token-derived seal key

Remediates the two web-console residuals from the 2026-07-05 posture audit: - Brute-force throttle (loginThrottle.ts): per-IP exponential backoff after 5 free attempts, plus a global floor for spread-out floods, keyed on the socket peer IP (not spoofable X-Forwarded-For) with a size-capped map. The constant-time compare already stopped the timing leak; this bounds guess *volume* against a by-design LAN-exposed console. - Session seal key now derives from the high-entropy mgmt token instead of the low-entropy login password, so a captured cookie is no longer an offline password oracle. Falls back to the password only when no token is configured (dev/local). Rotating the token now invalidates sessions. - Replace the process-wide NODE_TLS_REJECT_UNAUTHORIZED=0 with per-request Bun TLS scoped to the loopback proxy hop; a non-loopback mgmt URL now verifies normally. Dropped the env var from the systemd unit, Steam Deck installer, Windows run scripts, env examples, and web README. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-06 07:28:39 +02:00
parent ab56536842
commit bda5556d37
11 changed files with 204 additions and 34 deletions
@@ -3,8 +3,9 @@
 # Installed by the punktfunk-web .deb to /usr/lib/systemd/user/. AUTO-WIRED — no env editing:
 # it sources the host's mgmt token + the generated login password, serves HTTPS (HTTP/1.1 over TLS)
 # with the host's own identity cert (~/.config/punktfunk/{cert,key}.pem), and points the /api proxy
-# at the host's loopback HTTPS mgmt API (self-signed cert → NODE_TLS_REJECT_UNAUTHORIZED for the
-# proxy's only outbound hop, which is loopback). Enable per user:
+# at the host's loopback HTTPS mgmt API. The self-signed cert is accepted only for that loopback hop,
+# scoped inside the proxy code (Bun per-request TLS) — no process-wide NODE_TLS_REJECT_UNAUTHORIZED.
+# Enable per user:
 #   systemctl --user enable --now punktfunk-web
 [Unit]
 Description=punktfunk management web console
@@ -20,7 +21,6 @@ Type=simple
 EnvironmentFile=%h/.config/punktfunk/mgmt-token
 EnvironmentFile=-%h/.config/punktfunk/web-password
 Environment=PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990
-Environment=NODE_TLS_REJECT_UNAUTHORIZED=0
 Environment=PORT=47992
 Environment=HOST=0.0.0.0
 # Serve HTTPS (HTTP/1.1 over TLS) with the host's own identity cert; mark the
@@ -208,7 +208,7 @@ Description=punktfunk management web console
 After=punktfunk-host.service

 [Service]
-ExecStart=$DISTROBOX enter $BOX -- bash -lc 'cd $SRC/web; set -a; . $CONFIG/mgmt-token; . $CONFIG/web.env; set +a; export PUNKTFUNK_MGMT_URL=https://127.0.0.1:$MGMT_PORT NODE_TLS_REJECT_UNAUTHORIZED=0 PORT=$WEB_PORT HOST=0.0.0.0 NITRO_PORT=$WEB_PORT NITRO_HOST=0.0.0.0 PUNKTFUNK_UI_TLS_CERT=$CONFIG/cert.pem PUNKTFUNK_UI_TLS_KEY=$CONFIG/key.pem PUNKTFUNK_UI_SECURE=1; exec bun .output/server/index.mjs'
+ExecStart=$DISTROBOX enter $BOX -- bash -lc 'cd $SRC/web; set -a; . $CONFIG/mgmt-token; . $CONFIG/web.env; set +a; export PUNKTFUNK_MGMT_URL=https://127.0.0.1:$MGMT_PORT PORT=$WEB_PORT HOST=0.0.0.0 NITRO_PORT=$WEB_PORT NITRO_HOST=0.0.0.0 PUNKTFUNK_UI_TLS_CERT=$CONFIG/cert.pem PUNKTFUNK_UI_TLS_KEY=$CONFIG/key.pem PUNKTFUNK_UI_SECURE=1; exec bun .output/server/index.mjs'
 Restart=on-failure
 RestartSec=3

@@ -37,7 +37,8 @@ rem Fixed deployment wiring (the Windows analogue of scripts/punktfunk-web.servi
 set "PORT=47992"
 set "HOST=0.0.0.0"
 set "PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990"
-set "NODE_TLS_REJECT_UNAUTHORIZED=0"
+rem No NODE_TLS_REJECT_UNAUTHORIZED: the host's self-signed cert is accepted only for the loopback
+rem proxy hop, scoped inside the proxy code (Bun per-request TLS), not process-wide.
 rem Serve HTTPS (HTTP/1.1 over TLS) with the host's identity cert; mark the session cookie Secure.
 set "PUNKTFUNK_UI_TLS_CERT=%CERTFILE%"
 set "PUNKTFUNK_UI_TLS_KEY=%KEYFILE%"
@@ -19,13 +19,15 @@ PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990
 # the proxy gets 401.
 PUNKTFUNK_MGMT_TOKEN=

-# REQUIRED with the HTTPS mgmt API: the host presents a SELF-SIGNED identity cert on loopback,
-# which the proxy's fetch would otherwise reject (→ 502). The web server makes no other outbound
-# TLS calls, so disabling verification here only affects the loopback hop to the host's own cert.
-NODE_TLS_REJECT_UNAUTHORIZED=0
+# NOTE: NODE_TLS_REJECT_UNAUTHORIZED is intentionally NOT set. The host's self-signed loopback cert
+# is accepted only for the /api proxy's loopback hop — scoped inside the proxy code (Bun per-request
+# TLS: server/routes/api/[...].ts), so it can never silently unverify some other outbound TLS. A
+# NON-loopback PUNKTFUNK_MGMT_URL is verified normally (present a valid chain).

-# OPTIONAL: explicit cookie-sealing secret (>= 32 chars). If unset, a stable key is derived
-# from PUNKTFUNK_UI_PASSWORD (changing the password then invalidates sessions).
+# OPTIONAL: explicit cookie-sealing secret (>= 32 chars). If unset, the key is derived from the
+# high-entropy PUNKTFUNK_MGMT_TOKEN (so a captured cookie is NOT an offline password oracle); only if
+# no token is configured (dev/local) does it fall back to deriving from PUNKTFUNK_UI_PASSWORD.
+# Rotating the mgmt token invalidates existing sessions.
 # PUNKTFUNK_UI_SECRET=

 # TLS: serve the console over HTTPS (HTTP/1.1 over TLS) using the HOST's own identity cert (the cert
@@ -52,13 +52,14 @@ LAN console.)
 bun run build             # → .output/  (Nitro `bun` preset + our Bun.serve TLS entry)
 PORT=47992 HOST=0.0.0.0 \
  PUNKTFUNK_UI_PASSWORD=… PUNKTFUNK_MGMT_TOKEN=… \
-  PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990 NODE_TLS_REJECT_UNAUTHORIZED=0 \
+  PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990 \
  PUNKTFUNK_UI_TLS_CERT=~/.config/punktfunk/cert.pem \
  PUNKTFUNK_UI_TLS_KEY=~/.config/punktfunk/key.pem PUNKTFUNK_UI_SECURE=1 \
  bun run start           # = bun run .output/server/index.mjs
 # PUNKTFUNK_UI_TLS_* unset ⇒ plain HTTP (local dev); both set ⇒ HTTPS (HTTP/1.1 over TLS).
-# NODE_TLS_REJECT_UNAUTHORIZED=0 is only for the proxy's loopback fetch to the host's self-signed
-# mgmt cert; the console makes no other outbound TLS calls. See .env.example.
+# The host's self-signed mgmt cert is accepted only for the proxy's loopback hop, scoped in code
+# (Bun per-request TLS: server/routes/api/[...].ts) — no process-wide NODE_TLS_REJECT_UNAUTHORIZED.
+# See .env.example.
 bun run lint              # tsc --noEmit
 ```

@@ -1,13 +1,26 @@
 // POST /_auth/login {password} — verify the shared password (constant-time), then seal an
 // authenticated session cookie. Public (allowlisted in the gate) so an unauthenticated user
-// can actually log in.
-import { createError, defineEventHandler, readBody, useSession } from "h3";
+// can actually log in. Brute force is bounded by an in-memory per-IP throttle (loginThrottle):
+// the constant-time compare stops a timing leak, the throttle stops guessing at volume.
+import {
+	createError,
+	defineEventHandler,
+	getRequestIP,
+	readBody,
+	setResponseHeader,
+	useSession,
+} from "h3";
 import {
 	type SessionData,
 	sessionConfig,
 	timingSafeEqual,
 	uiPassword,
 } from "../../util/auth";
+import {
+	recordLoginFailure,
+	recordLoginSuccess,
+	throttleRetryAfterMs,
+} from "../../util/loginThrottle";

 export default defineEventHandler(async (event) => {
 	const expected = uiPassword();
@@ -17,11 +30,28 @@ export default defineEventHandler(async (event) => {
 			statusMessage: "auth not configured",
 		});
 	}
+	// The socket peer address — deliberately NOT trusting X-Forwarded-For (spoofable unless we sit
+	// behind a known proxy, which the packaged console does not). Falls back to a single shared bucket
+	// if the address is somehow unavailable, so the throttle still applies.
+	const ip = getRequestIP(event) ?? "unknown";
+
+	// Throttle BEFORE touching the password so a locked-out client can't keep the guess loop spinning.
+	const wait = throttleRetryAfterMs(ip);
+	if (wait > 0) {
+		setResponseHeader(event, "Retry-After", Math.ceil(wait / 1000));
+		throw createError({
+			statusCode: 429,
+			statusMessage: "too many attempts — try again shortly",
+		});
+	}
+
 	const body = await readBody<{ password?: string }>(event);
 	const password = String(body?.password ?? "");
 	if (!timingSafeEqual(password, expected)) {
+		recordLoginFailure(ip);
 		throw createError({ statusCode: 401, statusMessage: "invalid password" });
 	}
+	recordLoginSuccess(ip);
 	const session = await useSession<SessionData>(event, sessionConfig());
 	await session.update({ authenticated: true });
 	return { ok: true };
@@ -9,11 +9,12 @@ import {
 	proxyRequest,
 	setResponseStatus,
 } from "h3";
-import { mgmtToken, mgmtUrl } from "../../util/auth";
+import { isLoopbackUrl, mgmtToken, mgmtUrl } from "../../util/auth";

 export default defineEventHandler((event) => {
 	const { pathname, search } = getRequestURL(event);
-	const target = `${mgmtUrl()}${pathname}${search}`;
+	const base = mgmtUrl();
+	const target = `${base}${pathname}${search}`;
 	const token = mgmtToken();
 	// The mgmt API now requires a token always. Without one configured, forwarding an empty bearer
 	// would just bounce as 401 — fail fast and legibly instead (the packaged service sources the
@@ -25,7 +26,20 @@ export default defineEventHandler((event) => {
 				"management token not configured (PUNKTFUNK_MGMT_TOKEN / ~/.config/punktfunk/mgmt-token)",
 		};
 	}
+	// TLS scoping (replaces the old process-wide NODE_TLS_REJECT_UNAUTHORIZED=0): the host presents a
+	// SELF-SIGNED, no-SAN identity cert on loopback, which normal verification rejects. We relax
+	// verification ONLY for this one loopback hop, via Bun's per-request `tls` option — so any OTHER
+	// outbound TLS the process ever makes still verifies normally (the global env unverified
+	// everything). If the operator points PUNKTFUNK_MGMT_URL at a NON-loopback host, we do NOT relax:
+	// a remote mgmt API must present a valid chain, which is stricter than the old blanket accept.
+	const fetchOptions = isLoopbackUrl(base)
+		? // `tls` is a Bun.fetch extension (the console runs on bun — Bun.serve/`bun .output/...`), not
+			// in the standard RequestInit type, so cast through unknown.
+			({ tls: { rejectUnauthorized: false } } as unknown as RequestInit)
+		: undefined;
+
 	return proxyRequest(event, target, {
+		fetchOptions,
 		headers: {
 			// Overwrite, not append: the host-held token replaces anything the browser sent.
 			authorization: `Bearer ${token}`,
@@ -29,22 +29,59 @@ export function mgmtToken(): string {
 	return process.env.PUNKTFUNK_MGMT_TOKEN ?? "";
 }

+/** Whether `url`'s host is a loopback address — the only place the proxy relaxes TLS verification
+ * for the host's self-signed cert. IPv4 127.0.0.0/8, IPv6 ::1, and the `localhost` name. */
+export function isLoopbackUrl(url: string): boolean {
+	let host: string;
+	try {
+		host = new URL(url).hostname;
+	} catch {
+		return false;
+	}
+	// URL wraps IPv6 in brackets in .host but strips them in .hostname; normalize anyway.
+	const h = host.replace(/^\[|\]$/g, "").toLowerCase();
+	if (h === "localhost" || h === "::1") return true;
+	return /^127\.\d{1,3}\.\d{1,3}\.\d{1,3}$/.test(h);
+}
+
 /**
- * The cookie-sealing key for h3 `useSession` (must be ≥ 32 chars). Use PUNKTFUNK_UI_SECRET
- * if set; otherwise derive a stable 64-hex key from the password so single-var config works
- * (changing the password then invalidates existing sessions, which is fine).
+ * The cookie-sealing key for h3 `useSession` (must be ≥ 32 chars). Precedence:
+ *   1. PUNKTFUNK_UI_SECRET — explicit operator override.
+ *   2. Derived from the MANAGEMENT TOKEN (a 32-byte / 64-hex CSPRNG value) — the packaged deployment
+ *      always has one, so the seal key is high-entropy without any extra config.
+ *   3. Only as a last resort (dev/local with no token) derive from the password.
+ *
+ * Why not (2)→password by default: the password is low-entropy (a human picks it), so a key DERIVED
+ * from it turns any captured session cookie into an OFFLINE dictionary oracle — an attacker unseals
+ * candidate cookies locally, no server round-trips, so the login throttle can't help. The mgmt token
+ * is unguessable, so a cookie sealed under it leaks nothing about the password. (Deriving from the
+ * token instead of the password also means changing the password no longer silently invalidates
+ * sessions; rotating the mgmt token does — the correct, security-relevant trigger.)
 */
 export function sessionConfig(): SessionConfig {
-	const secret = process.env.PUNKTFUNK_UI_SECRET;
-	const password =
-		secret && secret.length >= 32
-			? secret
-			: createHash("sha256")
+	const explicit = process.env.PUNKTFUNK_UI_SECRET;
+	const token = mgmtToken();
+	let secret: string;
+	if (explicit && explicit.length >= 32) {
+		secret = explicit;
+	} else if (token) {
+		// High-entropy source: the CSPRNG mgmt token. Hash it (never use the raw admin token as the
+		// seal key) with a distinct label so the two uses can't be conflated.
+		secret = createHash("sha256")
+			.update(`punktfunk-session-v1:token:${token}`)
+			.digest("hex");
+	} else {
+		// Last resort (no token configured — dev/local only). No worse than before; a real deployment
+		// always has a token and never reaches here.
+		secret = createHash("sha256")
 			.update(`punktfunk-session-v1:${uiPassword()}`)
 			.digest("hex");
+	}
 	return {
 		name: SESSION_NAME,
-		password,
+		// h3's `useSession` calls this seal key `password` (it's the iron/AES-GCM key, not the login
+		// password — see the derivation above).
+		password: secret,
 		// Bounds a stolen/replayed cookie's lifetime (sets the cookie Max-Age AND the iron
 		// seal TTL). 7 days for a single-user console.
 		maxAge: 60 * 60 * 24 * 7,
@@ -0,0 +1,85 @@
+// In-memory brute-force throttle for the login gate. The password compare is constant-time (no
+// timing leak), but nothing bounded the *volume* of guesses — a LAN peer could hammer
+// PUNKTFUNK_UI_PASSWORD at network speed against a console that is, by design, LAN-exposed. This is
+// the console analogue of the host's PAIRING_COOLDOWN / single-use pairing PIN: every other secret
+// gate in the project is rate-limited, so this one is too.
+//
+// State is a module-level Map — it lives for the lifetime of the single Bun/Nitro server process,
+// which is exactly the scope we want (a restart clearing it is fine; lockouts are short). Keyed on
+// the socket peer IP (NOT a spoofable X-Forwarded-For — see the caller). The map is size-capped so a
+// distinct-IP flood can't grow it without bound, and a global counter adds a mild floor so a
+// botnet-style spread across many IPs still can't get unlimited aggregate guesses.
+
+/** Backoff schedule. After `FREE_ATTEMPTS` failures from an IP, each further failure locks that IP
+ * for `BASE_LOCKOUT_MS * 2^(n)` up to `MAX_LOCKOUT_MS`. Generous enough never to bother a human who
+ * fat-fingered their password a couple of times, punishing enough to make brute force hopeless. */
+const FREE_ATTEMPTS = 5;
+const BASE_LOCKOUT_MS = 1_000;
+const MAX_LOCKOUT_MS = 5 * 60_000; // 5 minutes
+/** Forget an IP that's been quiet this long (also the idle-eviction horizon). */
+const ENTRY_TTL_MS = 15 * 60_000;
+/** Cap the tracking map so a distinct-IP flood can't grow memory without bound. */
+const MAX_TRACKED_IPS = 4_096;
+/** Global floor: once this many failures accumulate inside GLOBAL_WINDOW_MS across ALL IPs, every
+ * login attempt waits out a small fixed delay too, so a spread-out flood is still bounded. */
+const GLOBAL_FAIL_THRESHOLD = 100;
+const GLOBAL_WINDOW_MS = 60_000;
+const GLOBAL_LOCK_MS = 1_000;
+
+interface Entry {
+	fails: number;
+	/** Epoch ms before which this IP may not attempt again. */
+	lockedUntil: number;
+	/** Last activity (for TTL eviction). */
+	seen: number;
+}
+
+const byIp = new Map<string, Entry>();
+let globalFails: number[] = []; // epoch-ms timestamps of recent failures (any IP)
+
+function sweep(now: number): void {
+	if (byIp.size <= MAX_TRACKED_IPS) {
+		// Cheap path: only drop clearly-stale entries when we're not under memory pressure.
+		for (const [ip, e] of byIp) {
+			if (now - e.seen > ENTRY_TTL_MS) byIp.delete(ip);
+		}
+		return;
+	}
+	// Over the cap: evict oldest-seen first until back under it.
+	const oldest = [...byIp.entries()].sort((a, b) => a[1].seen - b[1].seen);
+	for (const [ip, e] of oldest) {
+		if (byIp.size <= MAX_TRACKED_IPS && now - e.seen <= ENTRY_TTL_MS) break;
+		byIp.delete(ip);
+	}
+}
+
+/** Call BEFORE checking the password. If the IP (or the global floor) is currently locked out,
+ * returns the number of ms the caller should tell the client to wait; otherwise 0 (proceed). */
+export function throttleRetryAfterMs(ip: string, now = Date.now()): number {
+	const e = byIp.get(ip);
+	if (e && now < e.lockedUntil) return e.lockedUntil - now;
+	globalFails = globalFails.filter((t) => now - t < GLOBAL_WINDOW_MS);
+	if (globalFails.length >= GLOBAL_FAIL_THRESHOLD) return GLOBAL_LOCK_MS;
+	return 0;
+}
+
+/** Record a failed attempt and arm/extend this IP's lockout with exponential backoff. */
+export function recordLoginFailure(ip: string, now = Date.now()): void {
+	sweep(now);
+	const e = byIp.get(ip) ?? { fails: 0, lockedUntil: 0, seen: now };
+	e.fails += 1;
+	e.seen = now;
+	const over = e.fails - FREE_ATTEMPTS;
+	if (over > 0) {
+		const lock = Math.min(MAX_LOCKOUT_MS, BASE_LOCKOUT_MS * 2 ** (over - 1));
+		e.lockedUntil = now + lock;
+	}
+	byIp.set(ip, e);
+	globalFails.push(now);
+}
+
+/** Record a successful login — clears the IP's failure state so a legitimate user isn't penalized
+ * for earlier typos. */
+export function recordLoginSuccess(ip: string): void {
+	byIp.delete(ip);
+}
@@ -38,7 +38,8 @@ rem Fixed deployment wiring (the Windows analogue of scripts/punktfunk-web.servi
 set "PORT=47992"
 set "HOST=0.0.0.0"
 set "PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990"
-set "NODE_TLS_REJECT_UNAUTHORIZED=0"
+rem No NODE_TLS_REJECT_UNAUTHORIZED: the host's self-signed cert is accepted only for the loopback
+rem proxy hop, scoped inside the proxy code (Bun per-request TLS), not process-wide.
 rem Serve HTTPS (HTTP/1.1 over TLS) with the host's identity cert; mark the session cookie Secure.
 set "PUNKTFUNK_UI_TLS_CERT=%CERTFILE%"
 set "PUNKTFUNK_UI_TLS_KEY=%KEYFILE%"
@@ -2,7 +2,7 @@
 #
 # On a `apt install punktfunk-web` install you DO NOT edit anything: the systemd --user units wire
 # everything automatically —
-#   punktfunk-web.service sets PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990, NODE_TLS_REJECT_UNAUTHORIZED=0,
+#   punktfunk-web.service sets PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990,
 #   PORT=47992, HOST=0.0.0.0, the PUNKTFUNK_UI_TLS_* cert paths + PUNKTFUNK_UI_SECURE=1, and sources:
 #     ~/.config/punktfunk/mgmt-token    (written by the host's `serve` — the shared bearer token)
 #     ~/.config/punktfunk/web-password  (written by punktfunk-web-init — the console login password)
@@ -10,10 +10,9 @@
 #
 # This file documents the variables for a MANUAL deploy (running `bun .output/server/index.mjs`
 # yourself — the console runs on bun: `Bun.serve` is a Bun API, node can't run it). The mgmt API is
-# HTTPS with the host's self-signed loopback cert, so the proxy needs NODE_TLS_REJECT_UNAUTHORIZED=0
-# (its only outbound TLS hop is that loopback connection).
+# HTTPS with the host's self-signed loopback cert; the proxy accepts it ONLY for that loopback hop,
+# scoped in code (Bun per-request TLS) — so NODE_TLS_REJECT_UNAUTHORIZED is deliberately unset.
 PUNKTFUNK_MGMT_URL=https://127.0.0.1:47990
-NODE_TLS_REJECT_UNAUTHORIZED=0
 PORT=47992
 HOST=0.0.0.0