/

CVE-2026-65370

Severity:

HIGH RISK

(7.5)

HTTP Request Smuggling (CWE-444) in ServiceTalk HTTP/1.x

CVE-2026-65370

Analyzed

Vulnerability type

Category:

HTTP Request Smuggling

CWE:

CWE-444

Impact:

Request smuggling, request routing confusion, and potential access-control bypass

Severity:

HIGH RISK

(7.5)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

A malformed Transfer-Encoding handling flaw in ServiceTalk HTTP/1.x can cause different components to interpret HTTP request boundaries inconsistently. An attacker who can send crafted HTTP requests may bypass front-end controls, interfere with request routing, or access data intended for another request.

Affected context

The vulnerability occurs in ServiceTalk deployments that process HTTP/1.x traffic and encounter malformed or ambiguous Transfer-Encoding headers. Risk is greater when ServiceTalk is positioned behind a reverse proxy, load balancer, gateway, or other intermediary that parses request boundaries differently from the backend service.

Exploit preconditions

  • Attack surface: HTTP/1.x request-processing endpoints
  • Required access: Ability to send HTTP requests to an exposed ServiceTalk service or an intermediary forwarding requests to it
  • Trigger point: Parsing of malformed or ambiguous Transfer-Encoding headers
  • Why exploitation is possible: Inconsistent interpretation of message framing can cause front-end and back-end components to disagree about where one request ends and another begins

Why this happens

The underlying issue is improper handling and normalization of malformed Transfer-Encoding values during HTTP/1.x message framing. When a server or intermediary accepts ambiguous framing instead of applying consistent parsing and rejection rules, separate components may desynchronize their request boundaries. This behavior maps to CWE-444, HTTP Request/Response Smuggling, because attacker-controlled framing data can influence how multiple HTTP components interpret the same connection.

Potential impact

  • An attacker may bypass security controls enforced by an HTTP intermediary
  • Requests could be routed to unintended backend handlers
  • Sensitive responses may be exposed to an attacker
  • Session or authorization boundaries could be confused
  • Backend services may process attacker-controlled or queued requests unexpectedly

General mitigation guidance

Mitigation generally focuses on making HTTP message-framing behavior consistent

across every proxy, gateway, server, and application component.

Common defensive patterns include:

  • Rejecting malformed, duplicated, conflicting, or ambiguous Transfer-Encoding headers
  • Applying strict, standards-aligned HTTP/1.x parsing and canonicalization
  • Ensuring intermediaries and backend services use compatible request-framing rules
  • Prefering HTTP/2 or another framing model where appropriate, while validating protocol transitions
  • Limiting unnecessary intermediary hops and isolating backend services from direct public access
  • Applying least-privilege network policies and minimizing sensitive actions available through affected routes

How teams detect issues like this

Organizations commonly detect request-smuggling risks through:

  • Static and dynamic analysis of HTTP parser behavior around Content-Length and Transfer-Encoding combinations
  • Code review checks for permissive header parsing, inconsistent canonicalization, or acceptance of malformed framing
  • Integration tests that compare request-boundary interpretation across proxies, gateways, and backend services
  • Web application firewall and gateway policies that reject ambiguous or invalid transfer-coding syntax
  • Network monitoring for unusual connection reuse, desynchronized requests, or unexpected backend responses

Automated security testing tools can help identify parser differentials in controlled environments.

Sources

SourceLink
github.comgithub.com/apple/servicetalk/security/advisories/GHSA-56h2-h3h4-m9x6

Frequently asked

questions

Potentially. Exploitation may be possible over the network when an attacker can send HTTP/1.x requests through a vulnerable ServiceTalk endpoint or an intermediary that forwards traffic to it.

Not necessarily. The key requirement is usually access to an HTTP request path where malformed Transfer-Encoding data reaches the vulnerable parser. Authentication and network filtering may limit the practical scope.

Use strict, consistent HTTP message-framing validation and reject malformed or ambiguous Transfer-Encoding input before it reaches application routing logic.

Test all proxy-to-backend request paths for parser differentials, standardize HTTP parsing policies, keep intermediary behavior aligned, and monitor for abnormal connection reuse and request-boundary anomalies.

Find out what's already
exploitable in your codebase.