fix(security): harden 2FA setup, OIDC default role, and rate-limit fallback (v0.17.3-beta)

Address the actionable findings from an adversarial review:

- 2FA enrollment now requires the current password (/totp/setup), so a
  hijacked session or XSS can't enable TOTP and lock the real user out.
  SSO-only accounts (no local password) are told to set one first. Frontend
  collects the password before starting setup.
- OIDC config rejects default_role='admin': auto-provisioning unmatched users
  straight to admin is almost never intended — admin must come from an explicit
  group → role mapping.
- Rate limiter no longer fails fully open on a Redis outage. Redis calls are
  bounded by an 800ms timeout (the client uses maxRetriesPerRequest:null, so
  commands would otherwise hang, not throw), and on failure a per-process
  in-memory sliding-window fallback keeps /login and other limited endpoints
  throttled. Verified with Redis stopped: attempts throttle (429) instead of
  hanging or passing unlimited.

Findings intentionally not actioned: the "API-token auth is fire-and-forget"
report was a false positive (Express awaits next(), which is called inside the
promise), and the webhook DB-pool and JWT-revocation items are either
overstated or a separate architectural decision.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

backend/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mikrotik-manager-backend",
-  "version": "0.17.2-beta",
+  "version": "0.17.3-beta",
   "description": "Mikrotik Network Management Platform - Backend API",
   "main": "dist/index.js",
   "scripts": {

backend/src/middleware/rateLimitRedis.ts

@@ -2,6 +2,40 @@ import { Request, Response, NextFunction } from 'express';
 import { redis } from '../config/redis';
 import { query } from '../config/database';
 
+// Per-process sliding-window fallback used ONLY when Redis is unavailable, so
+// throttling degrades gracefully (esp. for /login) instead of failing fully
+// open exactly when infrastructure is under stress. It's best-effort and not
+// shared across replicas, but far better than no limit during a Redis outage.
+const memBuckets = new Map<string, number[]>();
+let lastSweep = Date.now();
+
+// The Redis client uses maxRetriesPerRequest:null, so commands QUEUE (hang)
+// rather than throw when Redis is down. Bound each call so an outage fails fast
+// into the in-memory fallback instead of stalling the request.
+const REDIS_OP_TIMEOUT_MS = 800;
+function withTimeout<T>(p: Promise<T>): Promise<T> {
+  return Promise.race([
+    p,
+    new Promise<T>((_, reject) => setTimeout(() => reject(new Error('redis timeout')), REDIS_OP_TIMEOUT_MS)),
+  ]);
+}
+
+function memoryLimitExceeded(key: string, windowSec: number, max: number): boolean {
+  const now = Date.now();
+  const windowMs = windowSec * 1000;
+  if (now - lastSweep > 60_000) {
+    for (const [k, ts] of memBuckets) {
+      const kept = ts.filter((t) => now - t < windowMs);
+      if (kept.length) memBuckets.set(k, kept); else memBuckets.delete(k);
+    }
+    lastSweep = now;
+  }
+  const recent = (memBuckets.get(key) ?? []).filter((t) => now - t < windowMs);
+  recent.push(now);
+  memBuckets.set(key, recent);
+  return recent.length > max;
+}
+
 /**
  * Fixed-window rate limit using Redis INCR (shared across API replicas).
  */
@@ -21,19 +55,24 @@ export function rateLimitRedis(options: {
     const uid = req.user?.userId;
     const key = `rl:${keyPrefix}:${uid != null ? `u:${uid}` : `ip:${req.ip || 'unknown'}`}`;
     try {
-      const n = await redis.incr(key);
+      const n = await withTimeout(redis.incr(key));
       if (n === 1) {
-        await redis.expire(key, windowSec);
+        await withTimeout(redis.expire(key, windowSec));
       }
       if (n > max) {
-        const ttl = await redis.ttl(key);
+        const ttl = await withTimeout(redis.ttl(key));
         const retryAfterSec = Math.max(1, ttl > 0 ? ttl : windowSec);
         res.setHeader('Retry-After', String(retryAfterSec));
         res.status(429).json({ error: 'Too many requests. Please try again later.' });
         return;
       }
     } catch (e) {
-      console.warn('[rateLimitRedis] Redis error, allowing request:', (e as Error).message);
+      console.warn('[rateLimitRedis] Redis error, using in-memory fallback:', (e as Error).message);
+      if (memoryLimitExceeded(key, windowSec, max)) {
+        res.setHeader('Retry-After', String(windowSec));
+        res.status(429).json({ error: 'Too many requests. Please try again later.' });
+        return;
+      }
     }
     next();
   };
@@ -68,21 +107,28 @@ export function loginRateLimit(): (req: Request, res: Response, next: NextFuncti
   return async (req: Request, res: Response, next: NextFunction): Promise<void> => {
     const ip = req.ip || 'unknown';
     const key = `rl:login:ip:${ip}`;
+    // Resolved outside the try so the fallback can reuse the limits on Redis failure
+    // (getLoginLimits has its own catch and never throws).
+    const { windowSec, max } = await getLoginLimits();
     try {
-      const { windowSec, max } = await getLoginLimits();
-      const n = await redis.incr(key);
+      const n = await withTimeout(redis.incr(key));
       if (n === 1) {
-        await redis.expire(key, windowSec);
+        await withTimeout(redis.expire(key, windowSec));
       }
       if (n > max) {
-        const ttl = await redis.ttl(key);
+        const ttl = await withTimeout(redis.ttl(key));
         const retryAfterSec = Math.max(1, ttl > 0 ? ttl : windowSec);
         res.setHeader('Retry-After', String(retryAfterSec));
         res.status(429).json({ error: 'Too many login attempts. Please try again later.' });
         return;
       }
     } catch (e) {
-      console.warn('[loginRateLimit] Redis error, allowing request:', (e as Error).message);
+      console.warn('[loginRateLimit] Redis error, using in-memory fallback:', (e as Error).message);
+      if (memoryLimitExceeded(key, windowSec, max)) {
+        res.setHeader('Retry-After', String(windowSec));
+        res.status(429).json({ error: 'Too many login attempts. Please try again later.' });
+        return;
+      }
     }
     next();
   };

backend/src/routes/auth.ts

@@ -99,8 +99,24 @@ router.post('/totp/verify', rateLimitRedis({ windowSec: 60, max: 5, keyPrefix: '
   return res.json({ token, user: { id: user.id, username: user.username, role: user.role } });
 });
 
-// Generate a new TOTP secret + QR code for the current user (does not enable yet)
+// Generate a new TOTP secret + QR code for the current user (does not enable yet).
+// Requires the current password so a hijacked session can't enroll 2FA and lock
+// the legitimate user out.
 router.post('/totp/setup', requireAuth, async (req: Request, res: Response) => {
+  const { password } = req.body as { password?: string };
+  if (!password) return res.status(400).json({ error: 'Current password is required to set up 2FA' });
+
+  const account = await queryOne<{ password_hash: string | null }>(
+    `SELECT password_hash FROM users WHERE id = $1`,
+    [req.user!.userId]
+  );
+  if (!account?.password_hash) {
+    return res.status(400).json({ error: 'This account has no local password (SSO). Set a password before enabling 2FA.' });
+  }
+  if (!(await bcrypt.compare(password, account.password_hash))) {
+    return res.status(401).json({ error: 'Incorrect password' });
+  }
+
   const secret = new OTPAuth.Secret({ size: 20 });
   const totp = new OTPAuth.TOTP({
     issuer: 'MikroTik Manager',

backend/src/routes/oidc.ts

@@ -86,8 +86,15 @@ router.put('/config', requireAuth, requireAdmin, async (req: Request, res: Respo
   const body = req.body as Partial<OidcConfig> & { client_secret?: string };
 
   const validRole = (r: unknown): r is AppRole => APP_ROLES.includes(r as AppRole);
-  if (body.default_role !== undefined && !validRole(body.default_role)) {
-    return res.status(400).json({ error: 'Invalid default_role' });
+  if (body.default_role !== undefined) {
+    if (!validRole(body.default_role)) {
+      return res.status(400).json({ error: 'Invalid default_role' });
+    }
+    // Auto-provisioning strangers straight to admin is almost never intended —
+    // admin must be granted via an explicit group mapping, not the fallback.
+    if (body.default_role === 'admin') {
+      return res.status(400).json({ error: 'default_role cannot be admin — grant admin via an explicit group → role mapping instead' });
+    }
   }
   if (body.group_role_map !== undefined) {
     if (typeof body.group_role_map !== 'object' || body.group_role_map === null) {