Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
HTTP request smuggling exploits a disagreement between two servers about where one request ends, letting an attacker prepend hidden content to another user’s request.

What Is HTTP Request Smuggling?
Nearly every public website is served by at least two HTTP implementations: a content delivery network, load balancer, or reverse proxy at the front, and an application server behind it. Both parse the same byte stream, and both decide independently where one request stops and the next begins. HTTP request smuggling is what happens when they reach different answers. The attacker supplies a request the front end reads as complete and the back end reads as complete plus a leftover fragment, and that fragment is then prepended to whatever request arrives next on the reused connection, which usually belongs to another user.
HTTP/1.1 offers two ways to state a body length: the Content-Length header and Transfer-Encoding: chunked. The specification says chunked wins when both are present, and the trouble begins when one of the two servers disagrees or can be persuaded to overlook a header.
Content-Length while the back end honours Transfer-Encoding, so bytes the front end counted as body are treated by the back end as the start of a new request.James Kettle brought the technique back into general awareness at Black Hat USA in August 2019 with “HTTP Desync Attacks: Smashing into the Cell Next Door”, having built the automated detection that made it practical to find at scale. Multiple implementations were patched as a result, among them CVE-2019-18277 in HAProxy and a set of fixes in Node.js and Go.
HTTP/2 uses an explicit length field per frame, which removes the ambiguity within the protocol. The ambiguity returns at the boundary, because many front ends downgrade HTTP/2 requests to HTTP/1.1 before passing them to the back end, and the translation has to invent the headers that HTTP/1.1 needs. Kettle’s 2021 research, “HTTP/2: The Sequel Is Always Worse”, showed H2.CL and H2.TE variants, plus request tunnelling where a smuggled request affects only the attacker’s own connection but still reaches internal endpoints. Later work on client-side desync and the CL.0 pattern, where a server ignores the body of certain requests entirely, widened the class again.
/admin are enforced at the proxy, a smuggled request reaches the back end without passing them.Content-Length and Transfer-Encoding, or a malformed chunked body, should produce a 400 response and a closed connection.Timing-based probes are the standard method: a request crafted so that a desync causes the back end to wait for bytes that never arrive produces a measurable delay. Burp Suite’s HTTP Request Smuggler extension automates the permutations, and differential responses confirm the finding. Testing requires care, because a successful probe can affect genuine users sharing the connection, so it belongs in a controlled window with the service owner informed. Our penetration test services include desync testing for layered architectures on that basis.
Possibly more so, because a CDN guarantees at least two HTTP implementations in the chain and heavy connection reuse to your origin. Major providers patch known desync patterns quickly, but the risk depends on how your origin parses requests as much as on the provider. Test the combination you actually run rather than assuming the provider covers it.
Only if HTTP/2 is used for the whole path. Most deployments terminate HTTP/2 at the edge and forward HTTP/1.1 to the back end, and that downgrade recreates the ambiguity the newer protocol removed. End-to-end HTTP/2, or strict rejection of ambiguous messages at the downgrade point, is what actually closes the gap.
Usually high or critical. The attack affects other users rather than only the attacker, captures session tokens without any interaction from the victim, and bypasses controls enforced at the edge. It is also difficult to detect after the fact, because the malicious bytes appear in logs as part of a legitimate user’s request rather than as a separate entry.
A firewall can reject requests containing conflicting length headers, which blocks the simplest variants. It cannot resolve a disagreement between two servers about parsing, and if the firewall itself is one of the implementations in the chain it may be part of the problem. Fix the parsing behaviour rather than filtering the symptom.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.