⌘K
ENES
GitHub
conceptNetworking~5 min readUpdated May 01, 2026#cybersecurity#networking#express#forwarded-headers#trust-boundary

Header Trust in Node Express

Definition

Header trust in Node/Express is the decision of whether framework APIs such as req.ip, req.ips, req.hostname, and req.protocol should believe proxy-provided HTTP headers such as X-Forwarded-For, X-Forwarded-Host, X-Forwarded-Proto, or Forwarded.

Why it matters

Express applications often sit behind Nginx, Cloudflare, a load balancer, a CDN, or a platform router. In that shape, the TCP peer is the proxy, while the original client identity is carried in headers. If Express trusts those headers from the wrong network path, an attacker can spoof identity, bypass rate limits, poison logs, influence redirects, or reach IP-allowlisted behavior.

This is the common mistake the Mirror learning exercise exposed: old proxy code read cf-connecting-ip and x-forwarded-for as if they were facts, without proving the request arrived through the trusted proxy that should have authored them.

How it works

Express has 3 identity layers:

  1. Socket identity req.socket.remoteAddress is the network peer connected to the Node process. Behind a proxy, this is usually the proxy IP, not the user.

  2. Forwarded-header identity Headers such as X-Forwarded-For, X-Real-IP, CF-Connecting-IP, and RFC 7239 Forwarded claim the original client or previous hops. These are ordinary HTTP headers until a trusted proxy boundary gives them meaning.

  3. Framework-decided identity req.ip, req.ips, req.hostname, and req.protocol may change depending on Express trust proxy configuration. This value is convenient, but it is only as correct as the proxy trust model.

Unsafe pattern:

function getClientIp(req) {
  return (
    req.headers["cf-connecting-ip"] ||
    req.headers["x-forwarded-for"] ||
    req.socket.remoteAddress ||
    req.ip
  );
}

Safer shape:

app.set("trust proxy", "loopback, linklocal, uniquelocal");

app.use((req, res, next) => {
  res.locals.clientIp = req.ip;
  res.locals.tcpPeer = req.socket.remoteAddress;
  res.locals.forwardedFor = req.get("x-forwarded-for") || null;
  next();
});

The bug is not “using X-Forwarded-For.” The bug is believing a header before proving which network hop produced or normalized it.

Techniques / patterns

  • Find every use of req.ip, req.ips, req.hostname, req.protocol, req.headers["x-forwarded-for"], req.headers["x-real-ip"], CF-Connecting-IP, True-Client-IP, and Forwarded.
  • Identify the real deployment chain: internet → CDN → load balancer → app proxy → Express.
  • Check app.set("trust proxy", ...) and whether it matches the actual number or CIDR range of trusted proxy hops.
  • Compare raw headers, req.socket.remoteAddress, and Express-decided values in a local /whoami endpoint.
  • Look for security decisions based on client IP: admin allowlists, geo-blocking, fraud scoring, rate limiting, password reset throttling, and audit logs.
  • Treat direct-origin reachability as a separate test: if the origin is reachable without the intended proxy, forwarded-header trust usually collapses.

Variants and bypasses

Header trust in Express fails in 5 common ways.

1. Trust disabled but raw headers read manually

Express may be safely configured with trust proxy disabled, but application code bypasses the framework and reads req.headers["x-forwarded-for"] directly. This recreates the bug under a different name.

2. Boolean trust proxy: true

A boolean true tells Express to trust proxy headers broadly. If the app is reachable directly, the leftmost X-Forwarded-For value can become attacker-controlled identity.

3. Wrong hop count

trust proxy: 1 is correct only when exactly one trusted proxy sits in front of Express. Multi-CDN, service mesh, load balancer, and platform-router chains need explicit modeling. The wrong number chooses the wrong IP.

4. Header-name split brain

The edge proxy overwrites X-Forwarded-For, but the app reads CF-Connecting-IP, True-Client-IP, X-Client-IP, or Forwarded. Attackers look for the unsanitized sibling header.

5. Internal state hidden in headers

Middleware sets fake request headers such as x-is-proxy-header or x-country-not-allowed-header, then a later handler reads those names as internal state. It works until a future route, proxy, or middleware lets a client-controlled header collide with that namespace.

Impact

Ordered roughly by severity:

  • IP allowlist bypass. Admin paths or internal-only behavior become reachable by spoofing loopback, private, or trusted addresses.
  • Rate-limit bypass. Login, token, and API limits keyed only on req.ip can be rotated by changing forwarded headers.
  • Audit-log poisoning. Incident response follows a forged “client IP” rather than the observed TCP peer and proxy path.
  • Geo-policy bypass or false blocking. Country checks become attacker-controlled or unreliable.
  • Risk-scoring distortion. Fraud, bot, and abuse models ingest spoofed network identity.
  • Redirect and host confusion. X-Forwarded-Proto and forwarded host values can affect URL generation and callback logic when trusted too broadly.

Detection and defense

Ordered by effectiveness:

  1. Model the proxy chain explicitly Write down every trusted hop and whether it overwrites or appends forwarding headers. Express configuration should encode that exact model, not a vague “behind proxy” assumption.

  2. Use Express trust proxy narrowly Prefer exact hop counts or CIDR ranges for known proxies. Boolean true is only safe when direct access is impossible and the last proxy reliably strips/overwrites inbound forwarding headers.

  3. Stop reading raw forwarding headers for identity Centralize client-IP derivation once, after Express proxy trust is configured. Raw headers can still be logged as claims, but they should not become identity by themselves.

  4. Block direct origin reachability If the app should only receive traffic from a proxy or load balancer, enforce that with firewall/security-group rules and application-level rejection of unexpected TCP peers.

  5. Log observed and claimed identity separately Store tcp_peer, express_client_ip, and forwarded_for_claim as different fields. This preserves forensic truth without pretending all values have the same trust level.

  6. Keep internal state out of request headers Use res.locals, request-local symbols, typed metadata, or explicit error subclasses. Internal control signals should not reuse HTTP header names.

What does not work as a primary defense

  • Checking only that X-Forwarded-For exists. Existence proves nothing about who set it.
  • Trusting Cloudflare-style headers without checking Cloudflare source ranges. Header names do not authenticate the network path.
  • Using req.ip blindly. req.ip is a framework decision affected by trust proxy; it is not automatically the network truth.
  • Relying on a WAF rule that strips one header. Attackers can try sibling headers unless the whole forwarding namespace is normalized.
  • Treating IP allowlists as authentication. IP checks are a layer, not a substitute for user, device, or service identity.

Practical labs

Compare raw and framework identity

app.get("/whoami", (req, res) => {
  res.json({
    socketRemoteAddress: req.socket.remoteAddress,
    expressIp: req.ip,
    expressIps: req.ips,
    xForwardedFor: req.get("x-forwarded-for") || null,
    forwarded: req.get("forwarded") || null,
    cfConnectingIp: req.get("cf-connecting-ip") || null
  });
});

The result shows which values are observed network facts and which are client-supplied claims.

Test direct header spoofing

curl -s http://localhost:3000/whoami \
  -H "X-Forwarded-For: 127.0.0.1" \
  -H "CF-Connecting-IP: 10.0.0.5" | jq

If a security decision changes from this alone, the app is trusting headers from the wrong path.

Compare trust proxy modes

// Run the same /whoami route with each setting.
app.set("trust proxy", false);
app.set("trust proxy", 1);
app.set("trust proxy", "loopback, linklocal, uniquelocal");

The difference teaches why Express proxy trust is deployment-specific.

Search a codebase for raw header trust

rg -n "x-forwarded-for|x-real-ip|cf-connecting-ip|true-client-ip|forwarded|req\\.ip|trust proxy" src

Every match is either identity derivation, telemetry, or a bug candidate; classify it before changing behavior.

Store claims separately

request_id:
tcp_peer:
express_client_ip:
forwarded_for_claim:
trusted_proxy_config:
security_decision:

This evidence card prevents logs from flattening claims and facts into one misleading field.

Practical examples

  • An Express login limiter keys attempts by req.headers["x-forwarded-for"]; attackers rotate that header and bypass the per-IP limit.
  • A dashboard checks req.ip === "127.0.0.1" after trust proxy: true; a direct request with X-Forwarded-For: 127.0.0.1 reaches admin-only behavior.
  • A Cloudflare-fronted origin leaks its direct IP; attackers connect directly and send CF-Connecting-IP manually.
  • A geolocation middleware blocks or redirects users based on an attacker-supplied forwarded IP.
  • A middleware sets req.headers["x-blocked-user-agent-header"] internally; later refactors make it unclear whether that flag is client-supplied or server-owned.

Suggested future atomic notes

  • trust-proxy-configuration
  • forwarded-header-spec
  • origin-ip-discovery
  • express-security-middleware

References

  • Official Tool Docs: Express behind proxies — https://expressjs.com/en/guide/behind-proxies.html
  • Foundational: MDN X-Forwarded-For — https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/X-Forwarded-For
  • Foundational: RFC 7239: Forwarded HTTP Extension — https://www.rfc-editor.org/rfc/rfc7239

Explore nearby notes

NetworkingClient IP TrustClient IP trust is the question of **which IP an application treats as "the client"** when requests pass through any intermediary — reverse proxy, load balancer... NetworkingNAT and Private NetworksPrivate networks use IP address ranges that are **not routable on the public Internet**, by design. NAT (Network Address Translation) lets hosts in those ranges... NetworkingReverse ProxiesA reverse proxy is an HTTP intermediary that accepts requests on behalf of one or more backend services, applies a fixed set of transformations, and forwards the... NetworkingTLS and HTTPSHTTPS is HTTP carried over TLS. TLS (Transport Layer Security, formerly SSL) is the protocol that wraps a TCP connection — or a UDP-based QUIC connection in HTTP/3... NetworkingCaching and SecurityCaching is the reuse of previously generated responses by browsers, CDNs, reverse proxies, shared intermediaries, or application layers. Caching becomes a security... NetworkingCloud Instance Metadata EndpointsCloud-instance metadata endpoints are HTTP services hosted by the cloud provider on a **link-local address** (typically 169.254.169.254) that any process running...