Skip to main content
Vulnerability Database/CVE-2026-105452

CVE-2026-105452: Docker Sandboxes Information Disclosure

CVE-2026-105452 is an information disclosure vulnerability in Docker Sandboxes allowing credential forwarding to attacker-controlled accounts. This article covers the technical details, affected versions, impact, and mitigation.

Published:

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

CVE-2026-105452 Overview

CVE-2026-105452 affects Docker Sandboxes, where the host egress proxy can forward a client-supplied credential alongside a credential injected by the proxy itself. The proxy strips alternate credentials only when their values match known sentinel values. Untrusted code running inside an authorized sandbox can supply an unrecognized credential in another supported authentication header. For affected upstream services, the request authenticates to an attacker-controlled account and exposes any data included in the request. The weakness is categorized as [CWE-200] Information Exposure.

Critical Impact

Untrusted sandbox code can redirect authenticated requests to an attacker-controlled account, exposing request data sent to upstream services.

Affected Products

  • Docker Sandboxes (host egress proxy with credential injection)
  • Deployments relying on sentinel-value-based credential filtering
  • Upstream services accepting multiple authentication header types

Discovery Timeline

  • 2026-10-08 - CVE-2026-105452 published to NVD
  • 2026-10-08 - Last updated in NVD database

Technical Details for CVE-2026-105452

Vulnerability Analysis

Docker Sandboxes provide isolated execution environments that can reach external services through a host-managed egress proxy. The proxy injects credentials on behalf of the sandbox so untrusted code never sees the real secret. To prevent sandbox code from spoofing or overriding injected credentials, the proxy inspects outgoing requests and strips alternate credential headers. The stripping logic only matches on known sentinel values rather than removing all authentication headers unconditionally. An attacker running code inside the sandbox can set a different, unrecognized credential in a supported authentication header that the proxy does not clear.

Root Cause

The root cause is an allow-by-default sanitization pattern in the egress proxy. The proxy enumerates specific sentinel credential values and removes only matching headers, instead of enforcing a deny-by-default policy that strips all client-supplied authentication material before injecting its own. This is an information exposure flaw [CWE-200] driven by incomplete filtering of security-sensitive headers.

Attack Vector

Exploitation requires the attacker to run code inside an authorized Docker Sandbox. The attacker crafts an outbound HTTP request that includes both the proxy-injected credential and an additional authentication header carrying an attacker-controlled credential. The upstream service observes multiple valid authentication inputs and may bind the request to the attacker's account. Any request body, path parameters, or query data sent by the sandbox is then exposed to the attacker-controlled identity. The vector is local to the sandbox, but the resulting exposure crosses trust boundaries to upstream services. Refer to the Docker Credential Injection Guide for how the proxy injects credentials.

Detection Methods for CVE-2026-105452

Indicators of Compromise

  • Outbound sandbox requests containing more than one authentication header, such as both Authorization and a service-specific token header.
  • Upstream service audit logs showing authentications to accounts that were never provisioned for the sandbox tenant.
  • Egress proxy logs showing requests with unexpected credential header names or non-sentinel credential values.

Detection Strategies

  • Inspect proxy telemetry for requests where multiple credential-bearing headers coexist on a single outbound call.
  • Correlate sandbox identity with upstream service account identity to flag mismatches.
  • Alert when sandbox-originated traffic reaches endpoints or accounts outside the sandbox's declared scope.

Monitoring Recommendations

  • Centralize egress proxy and upstream service authentication logs for cross-correlation.
  • Baseline normal header sets per sandbox workload and alert on deviations.
  • Monitor rate and destinations of outbound requests from each sandbox for sudden shifts in upstream account usage.

How to Mitigate CVE-2026-105452

Immediate Actions Required

  • Update Docker Sandboxes and the associated host egress proxy to the latest patched release per Docker's advisory.
  • Audit upstream services for authentication events originating from sandbox egress IP ranges and investigate unexpected account use.
  • Restrict which upstream services sandboxes are allowed to reach while remediation is validated.

Patch Information

Apply updates provided by Docker for the sandbox host and egress proxy components. Review the Docker Sandboxes Overview for current supported versions and the Docker Credential Injection Guide for configuration guidance after patching.

Workarounds

  • Configure the egress proxy to strip all client-supplied authentication headers before injecting its own, rather than relying on sentinel-value matching.
  • Limit sandbox egress to upstream services that enforce a single, well-defined authentication header and reject requests carrying additional credential headers.
  • Disable credential injection for sandbox workloads that do not require authenticated upstream access.
bash
# Configuration example: deny-by-default header stripping at the egress proxy
# Remove any client-supplied auth headers before the proxy injects its credential
strip_request_headers = [
  "Authorization",
  "Proxy-Authorization",
  "X-Api-Key",
  "X-Auth-Token"
]
inject_credential_after_strip = true

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

Experience the Most Advanced Cybersecurity Platform

See how the world’s most intelligent, autonomous cybersecurity platform can protect your organization today and into the future.